Python 2.7 quick reference

Python 2.7 quick reference John W. Shipman

tcc-doc@nmt.edu

$Revision: 1.21 $ $Date: 2013/04/24 19:51:32 $ A reference guide to most of the common features of the Python programming language, version 2.7. This publication is available in Web form and also as a PDF document. Please forward any comments to tcc-doc@nmt.edu.

Introduction: What is Python? Python is a recent, general-purpose, high-level programming language. It is freely available and runs pretty much everywhere. This document is a reference guide, not a tutorial. If you are new to Python programming, see A Python programming tutorial. Complete documentation and free installs are available from the python.org homepage. This document does not describe every single feature of Python 2.7. A few interesting features that 99% of Python users will never need, such as metaclasses, are not described here. Refer to the official documentation for the full feature set.

Python 2.7 and Python 3.x At this writing, both Python 2.7 and Python 3.2 are officially maintained implementations. The 3.0 release marked the first release in the development of Python that a new version was incompatible with the old one. If you are using 2.x releases of Python, there is no hurry to convert to the 3.x series. Release 2.7 is guaranteed to be around for many years. Furthermore, there are tools to help you automate much of the conversion process. Notes throughout this document will discuss specific features of 2.7 that are intended to ease the transition. For a discussion of the changes between 2.7 and 3.0, see What's New in Python. To see what changes must be made in your program to allow automatic conversion to Python 3.x, run Python with this flag: python -3 yourprogram To convert your program to Python 3.x, first make a copy of the original program, then run this command: python3-2to3 -w yourprogram The -w flag replaces yourprogram with the converted 3.x version, and moves the original to “yourprogram.bak” For full documentation of the Python 3.2 version, see the online documentation.

Starting Python You can use Python in two different ways: In “calculator” or “conversational mode”, Python will prompt you for input with three greater-than signs (>>>). Type a line and Python will print the result. Here's an example: >>> 2+2 4 >>> 1.0 / 7.0 0.14285714285714285 You can also use Python to write a program, sometimes called a script.

Using Python in Windows If you are using Python at the NM Tech Computer Center (TCC), you can get conversational mode from Start → All Programs → ActiveState ActivePython 2.6 → Python Interactive Shell. To write a program: Start → All Programs → ActiveState ActivePython 2.6 → PythonWin Editor. Use File → New, select Python Script in the pop-up menu, and click OK. This will bring up an edit window. Write your Python program in the edit window, then use File → Save As... to save it under some file name that ends in “.py”. Use your U: drive. This drive is mounted everywhere at the TCC, and contains your personal files. It is backed up regularly. To run your program, use File → Run. In the “Run Script” popup, enter the name of your program in the field labeled Script File, then click OK. The output will appear in the “Interactive Window”. You may also run a Python script by double-clicking on it, provided that its name ends with “.py”.

Using Python in Linux To enter conversational mode on a Linux system, type this command: python Type Control-D to terminate the session. If you write a Python script named filename.py, you can execute it using the command python filename.py Under Unix, you can also make a script self-executing by placing this line at the top: #!/usr/bin/env python You must also tell Linux that the file is executable by using the command “chmod +x filename”. For example, if your script is called hello.py, you would type this command: chmod +x hello.py

Line syntax The comment character is “#”; comments are terminated by end of line. Long lines may be continued by ending the line with a backslash (\), but this is not necessary if there is at least one open “(”, “[”, or “{”.

Names and keywords Python names (also called identifiers) can be any length and follow these rules: The first or only character must be a letter (uppercase or lowercase) or the underbar character, “_”. Any additional characters may be letters, underbars, or digits. Examples: coconuts, sirRobin, blanche_hickey_869, __secretWord. Case is significant in Python. The name “Robin” is not the same name as “robin”. The names below are keywords, also known as reserved words. They have special meaning in Python and cannot be used as names or identifiers. and def finally in print yield as del for is raise assert elif from lambda return break else global not try class except if or with continue exec import pass while

Basic types In programming, you manipulate values using operators. For example, in the expression “1+2”, the addition operator (+) is operating on the values 1 and 2 to produce the sum, 3. The Python operators are described in , but let's look first at Python's way of representing values. Every Python value must have a type. For example, the type of the whole number 1 is int, short for “integer.” Here is a table summarizing most of the commonly-used Python types. Python's common types Type name Values Examples int Integers in the range [-2147483648, 2147483647]. See . 42, -3, 1000000 long Integers of any size, limited only by the available memory. See . 42L, -3L, 100000000000000L bool The two Boolean values True and False. See . True, False float Floating-point numbers; see . 3.14159, -1.0, 6.0235e23 complex Complex numbers. If the idea of computing with the square root of -1 bothers you, just ignore this type, otherwise see . (3.2+4.9j), (0+3.42e-3j) str Strings of 8-bit characters; see . Strings can be empty: write such as a string as “""” or “''”. 'Sir Robin', "xyz", "I'd've" unicode Strings of 32-bit Unicode characters; see . u'Fred', u'\u03fa' list A mutable sequence of values; see . ['dot', 'dash']; [] tuple An immutable sequence of values; see . ('dot', 'dash'); (); ("singleton",) dict Use dict values (dictionaries) to structure data as look-up tables; see . {'go':1, 'stop':2}; {} bytearray A mutable sequence of 8-bit bytes; see . bytearray('Bletchley') file A file being read or written; see . open('/etc/motd') None A special, unique value that may be used where a value is required but there is no obvious value. See . None

Numeric types Python has a number of different types used for representing numbers.

Type <code >int</code >: Integers Python values of type int represent integers, that is, whole numbers in the range [-231, 231-1], roughly plus or minus two billion. You can represent a value in octal (base 8) by preceding it with “0o”. Similarly, use a leading “0x” to represent a value in hexadecimal (base 16), or “0b” for binary. Examples in conversational mode: >>> 999+1 1000 >>> 0o77 63 >>> 0xff 255 >>> 0b1001 9 The 0o and 0b prefixes work only in Python versions 2.6 and later. In 2.5 and earlier versions, any number starting with “0” was considered to be octal. This functionality is retained in the 2.6+ versions, but will not work in the Python 3.x versions. To convert other numbers or character strings to type int, see . If you perform operations on int values that result in numbers that are too large, Python automatically converts them to long type; see .

Type <code >long</code >: Extended-precision integers Values of long type represent whole numbers, but they may have many more than the nine or ten digits allowed by int type. In practice, the number of digits in a long value is limited only by processor memory size. To write a long-type constant, use the same syntax as for int-type constants, but place a letter L immediately after the last digit. Also, if an operation on int values results in a number too large to represent as an int, Python will automatically converted it to type long. >>> 100 * 100 10000 >>> 100L * 100L 10000L >>> 1000000000*1000000000 1000000000000000000L >>> 0xffffL 65535L To convert a value of a different numeric type or a string of characters to a long value, see .

Type <code >bool</code >: Boolean truth values A value of bool type represents a Boolean (true or false) value. There are only two values, written in Python as “True” and “False”. Internally, True is represented as 1 and False as 0, and they can be used in numeric expressions as those values. Here's an example. In Python, the expression “

a <
        b

” compares two values a and b, and returns True if a is less than b, False is a is greater than or equal to b. >>> 2 < 3 True >>> 3 < 2 False >>> True+4 5 >>> False * False 0 These values are considered False wherever true/false values are expected, such as in an if statement: The bool value False. Any numeric zero: the int value 0, the float value 0.0, the long value 0L, or the complex value 0.0j. Any empty sequence: the str value '', the unicode value u'', the empty list value [], or the empty tuple value (). Any empty mapping, such as the empty dict (dictionary) value {}. The special value None. All other values are considered True. To convert any value to a Boolean, see .

Type <code >float</code >: Floating-point numbers Values of this type represent real numbers, with the usual limitations of IEEE-754 floating point type: it cannot represent very large or very small numbers, and the precision is limited to only about 15 digits. For complete details on the IEEE-754 standard and its limitations, see the Wikipedia article. A floating-point constant may be preceded by a “+” or “-” sign, followed by a string of one or more digits containing a decimal point (“.”). For very large or small numbers, you may express the number in exponential notation by appending a letter “e” followed by a power of ten (which may be preceded by a sign). For example, Avogadro's Number gives the number of atoms of carbon in 12 grams of carbon12, and is written as 6.0221418×1023. In Python that would be “6.0221418e23”. Please note that calculations involving float type are approximations. In calculator mode, Python will display the numbers to their full precision, so you may see a number that is very close to what you expect, but not exact. >>> 1.0/7.0 0.14285714285714285 >>> -2*-4.2e37 8.4000000000000004e+37

Type <code >complex</code >: Imaginary numbers Mathematically, a complex number is a number of the form A+Bi where i is the imaginary number, equal to the square root of -1. Complex numbers are quite commonly used in electrical engineering. In that field, however, because the symbol i is used to represent current, they use the symbol j for the square root of -1. Python adheres to this convention: a number followed by “j” is treated as an imaginary number. Python displays complex numbers in parentheses when they have a nonzero real part. >>> 5j 5j >>> 1+2.56j (1+2.5600000000000001j) >>> (1+2.56j)*(-1-3.44j) (7.8064-6j) Unlike Python's other numeric types, complex numbers are a composite quantity made of two parts: the real part and the imaginary part, both of which are represented internally as float values. You can retrieve the two components using attribute references. For a complex number C: C.real is the real part. C.imag is the imaginary part as a float, not as a complex value. >>> a=(1+2.56j)*(-1-3.44j) >>> a (7.8064-6j) >>> a.real 7.8064 >>> a.imag -6.0 To construct a complex value from two float values, see .

Sequence types The next four types described (str, unicode, list and tuple) are collectively referred to as sequence types. Each sequence value represents an ordered set in the mathematical sense, that is, a collection of things in a specific order. Python distinguishes between mutable and immutable sequences: An immutable sequence can be created or destroyed, but the number, sequence, and values of its elements cannot change. The values of a mutable sequence can be changed. Any element can be replaced or deleted, and new elements can be added at the beginning, the end, or in the middle. There are four sequence types, but they share most of the same operations. . (immutable). (immutable). (mutable). (immutable).

Operations common to all the sequence types These functions work on values of the four sequence types: int, unicode, tuple, and list. . . . These operators apply to sequences. S1+S2 Concatenation—for two sequences S1 and S2 of the same type, a new sequence containing all the elements from S1 followed by all the elements of S2. >>> "vi" + "car" 'vicar' >>> [1,2,3]+[5,7,11,13]+[15] [1, 2, 3, 5, 7, 11, 13, 15] >>> ('roy', 'g')+('biv',) ('roy', 'g', 'biv') S*n For a sequence S and a positive integer n, the result is a new sequence containing all the elements of S repeated n times. >>> 'worra'*8 'worraworraworraworraworraworraworraworra' >>> [0]*4 [0, 0, 0, 0] >>> (True, False)*5 (True, False, True, False, True, False, True, False, True, False) x in S Is any element of a sequence S equal to x? For convenience in searching for substrings, if the sequence to be searched is a string, the x operand can be a multi-character string. In that case, the operation returns True if x is found anywhere in S. >>> 1 in [2,4,6,0,8,0] False >>> 0 in [2,4,6,0,8,0] True >>> 'a' in 'banana' True >>> 3.0 in (2.5, 3.0, 3.5) True >>> "baz" in "rowrbazzle" True x not in S Are all the elements of a sequence S not equal to x? >>> 'a' not in 'banana' False >>> 'x' not in 'banana' True S[i] Subscripting: retrieve the ith element of s, counting from zero. If i is greater than or equal to the number of elements of S, an IndexError exception is raised. >>> 'Perth'[0] 'P' >>> 'Perth'[1] 'e' >>> 'Perth'[4] 'h' >>> 'Perth'[5] Traceback (most recent call last): File "<stdin>", line 1, in <module> IndexError: string index out of range >>> ('red', 'yellow', 'green')[2] 'green' S[i:j] Slicing: For a sequence S and two integers i and j, return a new sequence with copies of the elements of S between positions i and j. The values used in slicing refer to the positions between elements, where position zero is the position before the first element; position 1 is between the first and second element; and so on. You can also specify positions relative to the end of a sequence. Position -1 is the position before the last element; -2 is the position before the second-to-last element; and so on. You can omit the starting position to obtain a slice starting at the beginning. You can omit the ending position to get all the elements through the last. For example, here is a diagram showing three slices of the string 'abcdef'. >>> 'abcdef'[2:5] 'cde' >>> 'abcdef'[:3] 'abc' >>> 'abcdef'[3:] 'def' >>> (90, 91, 92, 93, 94, 95)[2:5] (92, 93, 94) S[i:j:k] You can use a slice expression like this to select every kth element. Examples: >>> teens = range(13,20) >>> teens [13, 14, 15, 16, 17, 18, 19] >>> teens[::2] [13, 15, 17, 19] >>> teens[1::2] [14, 16, 18] >>> teens[1:5] [14, 15, 16, 17] >>> teens[1:5:2] [14, 16]

Type <code >str</code >: Strings of 8-bit characters Python has two string types. Type str holds strings of zero or more 8-bit characters, while unicode strings provide full support of the expanded Unicode character set; see . In addition to the functions described in , these functions apply to strings: . . .

String constants There are many forms for string constants: '...': You may enclose the string in single-quotes. "...": You may instead enclose it in double-quotes. Internally, there is absolutely no difference. To include a double-quote character inside the string, use the escape sequence “\"”. In conversational mode, Python will generally display values using single-quotes. If the string contains single-quotes but no double-quotes, Python will display the value using double-quotes. If the string contains both, the value will be displayed in single-quotes, with single-quote characters inside the value displayed as the escape sequence “\'”. '''...''': You may enclose your string between three single quotes in a row. The difference is that you can continue such a string over multiple lines, and the line breaks will be included in the string as newline characters. """...""": You can use three sets of double quotes. As with three sets of single quotes, line breaks are allowed and preserved as "\n" characters. If you use these triply-quoted strings in conversational mode, continuation lines will prompt you with “... ”. >>> 'Penguin' 'Penguin' >>> "ha'penny" "ha'penny" >>> "Single ' and double\" quotes" 'Single \' and double" quotes' >>> '' '' >>> "" '' >>> s='''This string ... contains two lines.''' >>> t="""This string ... contains ... three lines.""" In addition, you can use any of these escape sequences inside a string constant (see Wikipedia for more information on the ASCII code). String escape sequences \newline A backslash at the end of a line is ignored. \\ Backslash (\) \' Closing single quote (') \" Double-quote character (") \n Newline (ASCII LF or linefeed) \b Backspace (in ASCII, the BS character) \f Formfeed (ASCII FF) \r Carriage return (ASCII CR) \t Horizontal tab (ASCII HT) \v Vertical tab (ASCII VT) \ooo The character with octal code ooo, e.g., '\177'. \xhh The character with hexadecimal value hh, e.g., '\xFF'.

Raw strings: If you need to use a lot of backslashes inside a string constant, and doubling them is too confusing, you can prefix any string with the letter r to suppress the interpretation of escape sequences. For example, '\\\\' contains two backslashes, but r'\\\\' contains four. Raw strings are particularly useful with .

Definition of “whitespace” In Python, these characters are considered whitespace: Escape sequence ASCII name English name ' ' SP space '\n' NL newline '\r' CR carriage return '\t' HT horizontal tab '\f' FF form feed '\v' VT vertical tab

Methods on <code >str</code > values These methods are available on any string value S. S.capitalize() Return S with its first character capitalized (if a letter). >>> 'e e cummings'.capitalize() 'E e cummings' >>> '---abc---'.capitalize() '---abc---' S.center(w) Return S centered in a string of width w, padded with spaces. If w<=len(S), the result is a copy of S. If the number of spaces of padding is odd, the extra space will placed after the centered value. Example: >>> 'x'.center(4) ' x ' S.count(t[,start[,end]]) Return the number of times string t occurs in S. To search only a slice S[start:end] of S, supply start and end arguments. >>> 'banana'.count('a') 3 >>> 'bananana'.count('na') 3 >>> 'banana'.count('a', 3) 2 >>> 'banana'.count('a', 3, 5) 1

S.decode ( encoding )

If S contains an encoded Unicode string, this method will return the corresponding value as unicode type. The encoding argument specifies which decoder to use; typically this will be the string 'utf_8' for the UTF-8 encoding. For discussion and examples, see . S.endswith(t[,start[,end]]) Predicate to test whether S ends with string t. If you supply the optional start and end arguments, it tests whether the slice S[start:end] ends with t. >>> 'bishop'.endswith('shop') True >>> 'bishop'.endswith('bath and wells') False >>> 'bishop'[3:5] 'ho' >>> 'bishop'.endswith('o', 3, 5) True S.expandtabs([tabsize]) Returns a copy of S with all tabs replaced by one or more spaces. Each tab is interpreted as a request to move to the next “tab stop”. The optional tabsize argument specifies the number of spaces between tab stops; the default is 8. Here is how the function actually works. The characters of S are copied to a new string T one at a time. If the character is a tab, it is replaced by enough tabs so the new length of T is a multiple of the tab size (but always at least one space). >>> 'X\tY\tZ'.expandtabs() 'X Y Z' >>> 'X\tY\tZ'.expandtabs(4) 'X Y Z' >>> 'a\tbb\tccc\tdddd\teeeee\tfffff'.expandtabs(4) 'a bb ccc dddd eeeee fffff' S.find(t[,start[,end]]) If string t is not found in S, return -1; otherwise return the index of the first position in S that matches t. The optional start and end arguments restrict the search to slice S[start:end]. >>> 'banana'.find('an') 1 >>> 'banana'.find('ape') -1 >>> 'banana'.find('n', 3) 4 >>> 'council'.find('c', 1, 4) -1

.format(*p,
            **kw)

See . S.index(t[,start[,end]]) Works like .find(), but if t is not found, it raises a ValueError exception. >>> 'council'.index('co') 0 >>> 'council'.index('phd') Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: substring not found S.isalnum() Predicate that tests whether S is nonempty and all its characters are alphanumeric. >>> ''.isalnum() False >>> 'abc123'.isalnum() True >>> '&*$#&*()abc123'.isalnum() False S.isalpha() Predicate that tests whether S is nonempty and all its characters are letters. >>> 'abc123'.isalpha() False >>> 'MaryRecruiting'.isalpha() True >>> ''.isalpha() False S.isdigit() Predicate that tests whether S is nonempty and all its characters are digits. >>> 'abc123'.isdigit() False >>> ''.isdigit() False >>> '2415'.isdigit() True S.islower() Predicate that tests whether S is nonempty and all its letters are lowercase (non-letter characters are ignored). >>> ''.islower() False >>> 'abc123'.islower() True >>> 'ABC123'.islower() False S.isspace() Predicate that tests whether S is nonempty and all its characters are whitespace characters. >>> ''.isspace() False >>> ' \t\n\r'.isspace() True >>> 'killer \t \n rabbit'.isspace() False S.istitle() A predicate that tests whether S has “title case”. In a title-cased string, uppercase characters may appear only at the beginning of the string or after some character that is not a letter. Lowercase characters may appear only after an uppercase letter. >>> 'abc def GHI'.istitle() False >>> 'Abc Def Ghi'.istitle() True S.isupper() Predicate that tests whether S is nonempty and all its letters are uppercase letters (non-letter characters are ignored). >>> 'abcDEF'.isupper() False >>> '123GHI'.isupper() True >>> ''.isupper() False S.join(L) L must be an iterable that produces a sequence of strings. The returned value is a string containing the members of the sequence with copies of the delimiter string S inserted between them. One quite common operation is to use the empty string as the delimiter to concatenate the elements of a sequence. Examples: >>> '/'.join(['never', 'pay', 'plan']) 'never/pay/plan' >>> '(***)'.join ( ('Property', 'of', 'the', 'zoo') ) 'Property(***)of(***)the(***)zoo' >>> ''.join(['anti', 'dis', 'establish', 'ment', 'arian', 'ism']) 'antidisestablishmentarianism' S.ljust(w) Return a copy of S left-justified in a field of width w, padded with spaces. If w<=len(S), the result is a copy of S. >>> "Ni".ljust(4) 'Ni ' S.lower() Returns a copy of S with all uppercase letters replaced by their lowercase equivalent. >>> "I like SHOUTING!".lower() 'i like shouting!' S.lstrip([c]) Return S with all leading characters from string c removed. The default value for c is a string containing all the whitespace characters. >>> ' \t \n Run \t \n away ! \n \t '.lstrip() 'Run \t \n away ! \n \t ' "***Done***".lstrip('*') 'Done***' >>> "(*)(*)(*Undone*)".lstrip ( ")(*" ) 'Undone*)' S.partition(d) Searches string S for the first occurrence of some delimiter string d. If S contains the delimiter, it returns a tuple (pre, d, post), where pre is the part of S before the delimiter, d is the delimiter itself, and post is the part of S after the delimiter. If the delimiter is not found, this method returns a 3-tuple (S, '', ''). >>> "Daffy English kniggets!".partition(' ') ('Daffy', ' ', 'English kniggets!') >>> "Daffy English kniggets!".partition('/') ('Daffy English kniggets!', '', '') >>> "a*b***c*d".partition("**") ('a*b', '**', '*c*d') S.replace(old,new[,max]) Return a copy of S with all occurrences of string old replaced by string new. Normally, all occurrences are replaced; if you want to limit the number of replacements, pass that limit as the max argument. >>> 'Frenetic'.replace('e', 'x') 'Frxnxtic' >>> 'Frenetic'.replace('e', '###') 'Fr###n###tic' >>> 'banana'.replace('an', 'erzerk') 'berzerkerzerka' >>> 'banana'.replace('a', 'x', 2) 'bxnxna' S.rfind(t[,start[,end]]) Like .find(), but if t occurs in S, this method returns the highest starting index. >>> 'banana'.find('a') 1 >>> 'banana'.rfind('a') 5 S.rindex(t[,start[,end]]) Similar to S.index(), but it returns the last index in S where string t is found. It will raise a ValueError exception if the string is not found. >>> "Just a flesh wound.".index('s') 2 >>> "Just a flesh wound.".rindex('s') 10 >>> "Just a flesh wound.".rindex('xx') Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: substring not found S.rjust(w[,fill]) Return a copy of S right-justified in a field of width w, padded with spaces. If w<=len(S), the result is a copy of S. To pad values with some character other than a space, pass that character as the optional second argument. >>> '123'.rjust(5) ' 123' >>> '123'.rjust(5,'*') '**123' S.rpartition(d) Similar to S.partition(), except that it finds the last occurrence of the delimiter. >>> "Daffy English kniggets!".rpartition(' ') ('Daffy English', ' ', 'kniggets!') >>> "a*b***c*d".rpartition("**") ('a*b*', '**', 'c*d') S.rsplit(d[,max]) Similar to S.split(d[,max]), except that if there are more fields than max, the split fields are taken from the end of the string instead of from the beginning. >>> "I am Zoot's identical twin sister, Dingo.".rsplit(None, 2) ["I am Zoot's identical twin", 'sister,', 'Dingo.'] S.rstrip([c]) Return S with all trailing characters from string c removed. The default value for c is a string containing all the whitespace characters. >>> ' \t \n Run \t \n away ! \n \t '.rstrip() ' \t \n Run \t \n away !' S.split([d[,max]]) Returns a list of strings

[s0,
              s1,
              ...]

made by splitting S into pieces wherever the delimiter string d is found. The default is to split up S into pieces wherever clumps of one or more whitespace characters are found. >>> "I'd annex \t \r the Sudetenland" .split() ["I'd", 'annex', 'the', 'Sudetenland'] >>> '3/crunchy frog/ Bath & Wells'.split('/') ['3', 'crunchy frog', ' Bath & Wells'] >>> '//Norwegian Blue/'.split('/') ['', '', 'Norwegian Blue', ''] >>> 'never<*>pay<*>plan<*>'.split('<*>') ['never', 'pay', 'plan', ''] The optional max argument limits the number of pieces removed from the front of S. The resulting list will have no more than max+1 elements. To use the max argument while splitting the string on clumps of whitespace, pass None as the first argument. >>> 'a/b/c/d/e'.split('/', 2) ['a', 'b', 'c/d/e'] >>> 'a/b'.split('/', 2) ['a', 'b'] >>> "I am Zoot's identical twin sister, Dingo.".split(None, 2) ['I', 'am', "Zoot's identical twin sister, Dingo."] S.splitlines([keepends]) Splits S into lines and returns a list of the lines as strings. Discards the line separators unless the optional keepends arguments is true. >>> """Is that ... an ocarina?""".splitlines() ['Is that', 'an ocarina?'] >>> """What is your name? ... Sir Robin of Camelot.""".splitlines(True) ['What is your name?\n', 'Sir Robin of Camelot.'] S.startswith(t[,start[,end]]) Predicate to test whether S starts with string t. Otherwise similar to .endswith(). >>> "bishop".startswith('bish') True >>> "bishop".startswith('The') False S.strip([c]) Return S with all leading and trailing characters from string c removed. The default value for c is a string containing all the whitespace characters. >>> ' \t \n Run \t \n away ! \n \t '.strip() 'Run \t \n away !' S.swapcase() Return a copy of S with each lowercase character replaced by its uppercase equivalent, and vice versa. >>> "abcDEF".swapcase() 'ABCdef' S.title() Returns the characters of S, except that the first letter of each word is uppercased, and other letters are lowercased. >>> "huge...tracts of land".title() 'Huge...Tracts Of Land' S.translate(new[,drop]) This function is used to translate or remove each character of S. The new argument is a string of exactly 256 characters, and each character x of the result is replaced by new[ord(x)]. If you would like certain characters removed from S before the translation, provide a string of those characters as the drop argument. For your convenience in building the special 256-character strings used here, see the definition of the maketrans() function of , where you will find examples. S.upper() Return a copy of S with all lowercase characters replaced by their uppercase equivalents. >>> 'I like shouting'.upper() 'I LIKE SHOUTING' S.zfill(w) Return a copy of S left-filled with '0' characters to width w. >>> '12'.zfill(9) '000000012'

The string <code >.format()</code > method The .format() method of the str type is an extremely convenient way to format text exactly the way you want it. This method was added in Python 2.6. Quite often, we want to embed data values in some explanatory text. For example, if we are displaying the number of nematodes in a hectare, it is a lot more meaningful to display it as

"There were 37.9
        nematodes per hectare

" than just "37.9". So what we need is a way to mix constant text like "nematodes per hectare" with values from elsewhere in your program. Here is the general form: template.format(p0, p1, ..., k0=v0, k1=v1, ...) The template is a string containing a mixture of one or more format codes embedded in constant text. The format method uses its arguments to substitute an appropriate value for each format code in the template. The arguments to the .format() method are of two types. The list starts with zero or more positional arguments pi, followed by zero or more keyword arguments of the form ki=vi, where each ki is a name with an associated value vi. Just to give you the general flavor of how this works, here's a simple conversational example. In this example, the format code “{0}” is replaced by the first positional argument (49), and “{1}” is replaced by the second positional argument, the string "okra". >>> "We have {0} hectares planted to {1}.".format(49, "okra") 'We have 49 hectares planted to okra.' >>> In the next example, we supply the values using keyword arguments. The arguments may be supplied in any order. The keyword names must be valid Python names (see ). >>> "{monster} has now eaten {city}".format( ... city='Tokyo', monster='Mothra') 'Mothra has now eaten Tokyo' You may mix references to positional and keyword arguments: >>> "The {structure} sank {0} times in {1} years.".format( ... 3, 2, structure='castle') 'The castle sank 3 times in 2 years.' If you need to include actual “{” and “}” characters in the result, double them, like this: >>> "There are {0} members in set {{a}}.".format(15) 'There are 15 members in set {a}.'

General form of a format code Here is the general form of a format code, where optional parts in [brackets], and actual characters are in "double quotes": "{" [name] ["!" conversion] [":" spec] "}" For the name portion, see . For the conversion part, see . For the spec part, see .

The <replaceable >name</replaceable > part The name part of a format code specifies the source of the value to be formatted here. Numbers refer to positional arguments passed to the .format() method, starting at 0 for the first argument. You may also use any Python name to refer to one of the keyword arguments. If the associated argument is an iterable, you may append an expression of this form to retrieve one of its elements: "[" index "]" For example: >>> signal=['red', 'yellow', 'green'] >>> signal[2] 'green' >>> "The light is {0[2]}!".format(signal) 'The light is green!' If the associated argument has attributes, you can append an expression of this form to refer to that attribute: "."name For example: >>> import string >>> string.digits '0123456789' >>> "Our digits are '{s.digits}'.".format(s=string) "Our digits are '0123456789'." In general, you can use any combination of these features. For example: >>> "The sixth digit is '{s.digits[5]}'".format(s=string) "The sixth digit is '5'" Starting with Python 2.7, you may omit all of the numbers that refer to positional arguments, and they will be used in the sequence they occur. For example: >>> "The date is {}-{}-{}.".format(2012, 5, 1) 'The date is 2012-5-1.' If you use this convention, you must omit all those numbers. You can, however, omit all the numbers and still use the keyword names feature: >>> "Can I have {} pounds to {excuse}?".format( ... 50, excuse='mend the shed') 'Can I have 50 pounds to mend the shed?'

The <replaceable >conversion</replaceable > part Following the name part of a format code, you can use one of these two forms to force the value to be converted by a standard function: !s str() !r repr() Here's an example: >>> "{}".format('Don\'t') "Don't" >>> "{!r}".format('Don\'t') '"Don\'t"'

The <replaceable >spec</replaceable > part After the name and conversion parts of a format code, you may use a colon (“:”) and a format specifier to supply more details about how to format the related value. Here is the general form of a format specifier. ":" [[fill] align] [sign] ["#"] ["0"] [width] [","] ["." prec] [type] fill You may specify any fill character except “}”. This character is used to pad a short value to the specified length. It may be specified only in combination with an align character. align Specifies how to align values that are not long enough to occupy the specified length. There are four values: < Left-justify the value. This is the default alignment for string values. > Right-justify the value. This is the default alignment for numbers. ^ Center the value. = For numbers using a sign specifier, add the padding between the sign and the rest of the value. Here are some examples of the use of fill and align. >>> "{:>8}".format(13) ' 13' >>> "{:>8}".format('abc') ' abc' >>> "{:*>8}".format('abc') '*****abc' >>> "{:*<8}".format('abc') 'abc*****' >>> "{:>5d}".format(14) ' 14' >>> "{:#>5d}".format(14) '###14' >>> "{:<6}".format('Git') 'Git ' >>> "{:*<6}".format('Git') 'Git***' >>> "{:=^8}".format('Git') '==Git===' >>> "{:*=-9d}".format(-3) '-*******3' sign This option controls whether an arithmetic sign is displayed. There are three possible values: + Always display a sign: + for positive, - for negative. - Display - only for negative values. (one space) Display one space for positive values, - for negative. Here are some examples of use of the sign options. >>> '{} {}'.format(17, -17) '17 -17' >>> '{:5} {:5}'.format(17, -17) ' 17 -17' >>> '{:<5} {:<5}'.format(17, -17) '17 -17 ' >>> '{:@<5} {:@<5}'.format(17, -17) '17@@@ -17@@' >>> '{:@>5} {:@>5}'.format(17, -17) '@@@17 @@-17' >>> '{:@^5} {:@^5}'.format(17, -17) '@17@@ @-17@' >>> '{:@^+5} {:@^+5}'.format(17, -17) '@+17@ @-17@' >>> '{:@^-5} {:@^-5}'.format(17, -17) '@17@@ @-17@' >>> '{:@^ 5} {:@^ 5}'.format(17, -17) '@ 17@ @-17@' "#" This option selects the “alternate form” of output for some types. When formatting integers as binary, octal, or hexadecimal, the alternate form adds “0b”, “0o”, or “0x” before the value, to show the radix explicitly. >>> "{:4x}".format(255) ' ff' >>> "{:#4x}".format(255) '0xff' >>> "{:9b}".format(62) ' 111110' >>> "{:#9b}".format(62) ' 0b111110' >>> "{:<#9b}".format(62) '0b111110 ' When formatting float, complex, or Decimal values, the “#” option forces the result to contain a decimal point, even if it is a whole number. >>> "{:5.0f}".format(36) ' 36' >>> "{:#5.0f}".format(36) ' 36.' >>> from decimal import Decimal >>> w=Decimal(36) >>> "{:g}".format(w) '36' >>> "{:#g}".format(w) '36.' "0" To fill the field with left zeroes, place a “0” at this position in your format code. >>> "{:5d}".format(36) ' 36' >>> "{:05d}".format(36) '00036' >>> "{:021.15}".format(1.0/7.0) '00000.142857142857143' width Place a number at this position to specify the total width of the displayed value. >>> "Beware the {}!".format('Penguin') 'Beware the Penguin!' >>> "Beware the {:11}!".format('Penguin') 'Beware the Penguin !' >>> "Beware the {:>11}!".format('Penguin') 'Beware the Penguin!' "," Place a comma at this position in your format code to display commas between groups of three digits in whole numbers. This feature was added in Python 2.7. >>> "{:,d}".format(12345678901234) '12,345,678,901,234' >>> "{:,f}".format(1234567890123.456789) '1,234,567,890,123.456787' >>> "{:25,f}".format(98765432.10987) ' 98,765,432.109870' "." precision Use this part to specify the number of digits after the decimal point. >>> from math import pi >>> "{}".format(pi) '3.141592653589793' >>> "{:.3}".format(pi) '3.14' >>> "{:25,.3f}".format(1234567890123.456789) ' 1,234,567,890,123.457' type This code specifies the general type of format used. The default is to convert the value of a string as if using the str() function. Refer to the table below for allowed values. b Format an integer in binary. c Given a number, display the character that has that code. d Display a number in decimal (base 10). e Display a float value using the exponential format. E Same as e, but use a capital “E” in the exponent. f Format a number in fixed-point form. g General numeric format: use either f or g, whichever is appropriate. G Same as “g”, but uses a capital “E” in the exponential form. n For formatting numbers, this format uses the current local setting to insert separator characters. For example, a number that Americans would show as “1,234.56”, Europeans would show it as “1.234,56”. o Display an integer in octal format. x Display an integer in hexadecimal (base 16). Digits greater than 9 are displayed as lowercase characters. X Display an integer in hexadecimal (base 16). Digits greater than 9 are displayed as uppercase characters. % Display a number as a percentage: its value is multiplied by 100, followed by a “%” character. Examples: >>> "{:b}".format(9) '1001' >>> "{:08b}".format(9) '00001001' >>> "{:c}".format(97) 'a' >>> "{:d}".format(0xff) '255' >>> from math import pi >>> "{:e}".format(pi*1e10) '3.141593e+10' >>> "{:E}".format(pi*1e10) '3.141593E+10' >>> "{:f}".format(pi) '3.141593' >>> "{:g}".format(pi) '3.14159' >>> "{:g}".format(pi*1e37) '3.14159e+37' >>> "{:G}".format(pi*1e37) '3.14159E+37' >>> "{:o}".format(255) '377' >>> "{:#o}".format(255) '0o377' >>> "{:x}".format(105199) '19aef' >>> "{:X}".format(105199) '19AEF' >>> "{:<#9X}".format(105199) '0X19AEF ' >>> "{:%}".format(0.6789) '67.890000%' >>> "{:15.3%}".format(0.6789) ' 67.890%'

Formatting a field of variable length Sometimes you need to format a field using a length that is available only once the program is running. To do this, you can use a number or name in {braces} inside a format code at the width position. This item then refers to either a positional or keyword argument to the .format() method as usual. Here's an example. Suppose you want to format a number n using d digits. Here are examples showing this with and without left-zero fill: >>> n = 42 >>> d = 8 >>> "{0:{1}d}".format(42, 8) ' 42' >>> "{0:0{1}d}".format(42, 8) '00000042' >>> You can, of course, also use keyword arguments to specify the field width. This trick also works for variable precision. "{count:0{width}d}".format(width=8, count=42) '00000042' >>> The same technique applies to substituting any of the pieces of a format code. >>> "{:&<14,d}".format(123456) '123,456&&&&&&&' >>> "{1:{0}{2}{3},{4}}".format('&', 123456, '<', 14, 'd') '123,456&&&&&&&' >>> "{:@^14,d}".format(1234567) '@@1,234,567@@@' >>> "{n:{fil}{al}{w},{kind}}".format( ... kind='d', w=14, al='^', fil='@', n=1234567) '@@1,234,567@@@'

The older string format operator Python versions before 2.6 did not have the string .format() method described in . Instead, string formatting used this general form: f % v where f is a template string and v specifies the value or values to be formatted using that template. If multiple values are to be formatted, v must be a tuple. The template string may contain any mixture of ordinary text and format codes. A format code always starts with a percent (%) symbol. See . The result of a format operation consists of the ordinary characters from the template with values substituted within them wherever a format code occurs. A conversational example: >>> print "We have %d pallets of %s today." % (49, "kiwis") We have 49 pallets of kiwis today. In the above example, there are two format codes. Code “%d” means “substitute a decimal number here,” and code “%s” means “substitute a string value here”. The number 49 is substituted for the first format code, and the string "kiwis" replaces the second format code. In general, format codes have this form: %[p][m[.n]]c Parts of the format operator p An optional prefix; see . m Specifies the total desired field width. The result will never be shorter than this value, but may be longer if the value doesn't fit; so, "%5d" % 1234 yields

"
                    1234"

, but "%2d" % 1234 yields "1234". If the value is negative, values are left-aligned in the field whenever they don't fill the entire width. n For float values, this specifies the number of digits after the decimal point. c Indicates the type of formatting.

Here are the format type codes, c in the general expression above: Format codes %s Format a string. For example,

'%-3s' %
                  'xy'

yields 'xy '; the width (-3) forces left alignment. %d Decimal conversion. For example,

'%3d' %
                  -4

yields the string ' -4'. %e Exponential format; allow four characters for the exponent. Examples:

'%08.1e' %
                  1.9783

yields '0002.0e+00'. %E Same as %e, but the exponent is shown as an uppercase E. %f For float type. E.g.,

'%4.1f' %
                  1.9783

yields ' 2.0'. %g General numeric format. Use %f if it fits, otherwise use %e. %G Same as %G, but an uppercase E is used for the exponent if there is one. %o Octal (base 8). For example,

'%o' %
                  13

yields '15'. %x Hexadecimal (base 16). For example,

'%x'
                  % 247

yields 'f7'. %X Same as %x, but capital letters are used for the digits A-F. For example,

'%04X'
                  % 247

yields '00F7'; the leading zero in the length (04) requests that Python fill up any empty leading positions with zeroes. %c Convert an integer to the character with the corresponding ASCII code. For example,

'%c'
                  % 0x61

yields the string 'a'. %% Places a percent sign (%) in the result. Does not require a corresponding value. Example: "Energy at %d%%." % 88 yields the value 'Energy at 88%.'.

Format code prefixes + For numeric types, forces the sign to appear even for positive values. - Left-justifies the value in the field. 0 For numeric types, use zero fill. For example, '%04d' % 2 produces the value '0002'. # With the %o (octal) format, append a leading "0"; with the %x (hexadecimal) format, append a leading "0x"; with the %g (general numeric) format, append all trailing zeroes. Examples: >>> '%4o' % 127 ' 177' >>> '%#4o' % 127 '0177' >>> '%x' % 127 '7f' >>> '%#x' % 127 '0x7f' >>> '%10.5g' % 0.5 ' 0.5' >>> '%#10.5g' % 0.5 ' 0.50000'

You can also use the string format operator % to format a set of values from a dictionary D (see ): f % D In this form, the general form for a format code is: %(k)[p][m[.n]]c where k is a key in dictionary D, and the rest of the format code is as in the usual string format operator. For each format code, the value of D[k] is used. Example: >>> named = {'last': 'Poe', 'first': 'Aloysius'} >>> 'Dear %(first)s %(last)s:' % named 'Dear Aloysius Poe:'

Type <code >unicode</code >: Strings of 32-bit characters With the advent of the Web as medium for worldwide information interchange, the Unicode character set has become vital. For general background on this character set, see the Unicode homepage. To get a Unicode string, prefix the string with u. For example: u'klarn' is a five-character Unicode string. To include one of the special Unicode characters in a string constant, use these escape sequences: \xHH For a code with the 8-bit hexadecimal value HH. \uHHHH For a code with the 16-bit hexadecimal value HHHH. \UHHHHHHHH For a code with the 32-bit hexadecimal value HHHHHHHH. Examples: >>> u'Klarn.' u'Klarn.' >>> u'Non-breaking-\xa0-space.' u'Non-breaking-\xa0-space.' >>> u'Less-than-or-equal symbol: \u2264' u'Less-than-or-equal symbol: \u2264' >>> u"Phoenician letter 'wau': \U00010905" u"Phoenician letter 'wau': \U00010905" >>> len(u'\U00010905') 1 All the operators and methods of str type are available with unicode values. Additionally, for a Unicode value U, use this method to encode its value as a string of type str:

U.encode (
            encoding[, error )

Return the value of U as type str. The encoding argument is a string that specifies the encoding method. In most cases, this will be 'utf_8'. For discussion and examples, see . The optional error string specifies what to do with characters that do not have exact equivalents. For example, if you are converting to the ASCII character set, the encoding argument is 'ascii'. Values of the error argument are given in the table below. 'strict' Raise a UnicodeError exception if any character has no ASCII equivalent. This is the default behavior. 'ignore' Leave out characters that have no equivalent. 'replace' Substitute a '?' for each character that has no equivalent. 'xmlcharrefreplace' Use the XML character entity escape sequence for characters with no ASCII equivalent. The general form of this sequence is "&#N;", where N is the decimal value of the character's code point. This feature is very handy for generating internationalized Web pages. 'backslashreplace' Use Python backslash escape sequences to represent characters with no equivalent. Here are some examples to demonstrate error argument values. >>> s = u"a\u262ez" >>> len(s) 3 >>> s u'a\u262ez' >>> s.encode('ascii') Traceback (most recent call last): File "<stdin>", line 1, in <module> UnicodeEncodeError: 'ascii' codec can't encode character u'\u262e' in position 1: ordinal not in range(128) >>> s.encode('ascii', 'ignore') 'az' >>> s.encode('ascii', 'replace') 'a?z' >>> s.encode('ascii', 'xmlcharrefreplace') 'a☮z' >>> hex(9774) '0x262e' >>> t = s.encode('ascii', 'backslashreplace') >>> t 'a\\u262eb' >>> print t a\u262eb >>> len(t) 8 >>> t[1] '\\'

The UTF-8 encoding How, you might ask, do we pack 32-bit Unicode characters into 8-bit bytes? Quite prevalent on the Web and the Internet generally is the UTF-8 encoding, which allows any of the Unicode characters to be represented as a string of one or more 8-bit bytes. First, some definitions: A code point is a number representing a unique member of the Unicode character set. The Unicode code points are visualized as a three-dimensional structure made of planes, each of which has a range of 65536 code points organized as 256 rows of 256 columns each. The low-order eight bits of the code point select the column; the next eight more significant bits select the row; and the remaining most significant bits select the plane. This diagram shows how UTF-8 encoding works. The first 128 code points (hexadecimal 00 through 7F) are encoded as in normal 7-bit ASCII, with the high-order bit always 0. For code points above hex 7F, all bytes have the high-order (0x80) bit set, and the bits of the code point are distributed through two, three, or four bytes, depending on the number of bits needed to represent the code point value. To encode a Unicode string U, use this method: U.encode('utf_8') To decode a regular str value S that contains a UTF-8 encoded value, use this method: S.decode('utf_8') Examples: >>> tilde='~' >>> tilde.encode('utf_8') '~' >>> u16 = u'\u0456' >>> s = u16.encode('utf_8') >>> s '\xd1\x96' >>> s.decode('utf_8') u'\u0456' >>> u32 = u'\U000E1234' >>> s = u32.encode('utf_8') >>> s '\xf3\xa1\x88\xb4' >>> s.decode('utf_8') u'\U000e1234' UTF-8 is not the only encoding method. For more details, consult the documentation for the Python module codecs.

Type <code >list</code >: Mutable sequences To form values into a sequence, use Python's list type if you are going to change, delete, or add values to the sequence. For a discussion of when to use list and when to use tuple, see . To create a list, enclose a list of zero or more comma-separated values inside square brackets, “[...]”. Examples: [] ["baked beans"] [23, 30.9, 'x'] You can also create a list by performing specific operations on each element of some sequence; see . Lists support all the operations described under . Methods available on lists are discussed in . There are a number of functions that can be used with lists as well: . . . . . . . . . . . . . . . . .

Methods on lists For any list value L, these methods are available. L.append(x) Append a new element x to the end of list L. Does not return a value. >>> colors = ['red', 'green', 'blue'] >>> colors.append('indigo') >>> colors ['red', 'green', 'blue', 'indigo'] L.count(x) Return the number of elements of L that compare equal to x. >>> [59, 0, 0, 0, 63, 0, 0].count(0) 5 >>> ['x', 'y'].count('Fomalhaut') 0 L.extend(S) Append another sequence S to L. >>> colors ['red', 'green', 'blue', 'indigo'] >>> colors.extend(['violet', 'pale puce']) >>> colors ['red', 'green', 'blue', 'indigo', 'violet', 'pale puce']

L.index(x[,
              start[, end]])

If L contains any elements that equal x, return the index of the first such element, otherwise raise a ValueError exception. The optional start and end arguments can be used to search only positions within the slice L[start:end]. >>> colors ['red', 'green', 'blue', 'indigo', 'violet', 'pale puce'] >>> colors.index('blue') 2 >>> colors.index('taupe') Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: list.index(x): x not in list >>> M=[0, 0, 3, 0, 0, 3, 3, 0, 0, 3] >>> M.index(3) 2 >>> M.index(3, 4, 8) 5 >>> M.index(3, 0, 2) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: list.index(x): x not in list L.insert(i,x) Insert a new element x into list L just before the ith element, shifting all higher-number elements to the right. No value is returned. >>> colors ['red', 'green', 'blue', 'indigo', 'violet', 'pale puce'] >>> colors[1] 'green' >>> colors.insert(1, "yellow") >>> colors ['red', 'yellow', 'green', 'blue', 'indigo', 'violet', 'pale puce'] L.pop([i]) Remove and return the element with index i from L. The default value for i is -1, so if you pass no argument, the last element is removed. >>> colors ['red', 'yellow', 'green', 'blue', 'indigo', 'violet', 'pale puce'] >>> tos = colors.pop() >>> tos 'pale puce' >>> colors ['red', 'yellow', 'green', 'blue', 'indigo', 'violet'] >>> colors[4] 'indigo' >>> dye = colors.pop(4) >>> dye 'indigo' >>> colors ['red', 'yellow', 'green', 'blue', 'violet'] L.remove(x) Remove the first element of L that is equal to x. If there aren't any such elements, raises ValueError. >>> colors ['red', 'yellow', 'green', 'blue', 'violet'] >>> colors.remove('yellow') >>> colors ['red', 'green', 'blue', 'violet'] >>> colors.remove('cornflower') Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: list.remove(x): x not in list >>> notMuch = [0, 0, 3, 0] >>> notMuch.remove(0) >>> notMuch [0, 3, 0] >>> notMuch.remove(0) >>> notMuch [3, 0] >>> notMuch.remove(0) >>> notMuch [3] >>> notMuch.remove(0) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: list.remove(x): x not in list L.reverse() Reverses the elements of L in place. Does not return a result. Compare . >>> colors ['red', 'green', 'blue', 'violet'] >>> colors.reverse() >>> colors ['violet', 'blue', 'green', 'red'] L.sort(cmp[,key[,reverse]]]) Sort list L in place. Does not return a result. Compare . The reordering is guaranteed to be stable—that is, if two elements are considered equal, their order after sorting will not change. While sorting, Python will use the built-in cmp() function to compare elements; see . You may provide, as the first argument to the .sort() method, your own comparator function to compare elements. This function must have the same calling sequence and return value convention as the built-in cmp() function: it must take two arguments, and return a negative number of the first argument precedes the second, a positive number if the second argument precedes the first, or zero if they are considered equal. You may also provide a “key extractor function” that is applied to each element to determine its key. This function must take one argument and return the value to be used as the sort key. If you want to provide a key extractor function but not a comparator function, pass None as the first argument to the method. Additionally, you may provide a third argument of True to sort the sequence in descending order; the default behavior is to sort into ascending order. >>> temps=[67, 73, 85, 93, 92, 78, 95, 100, 104] >>> temps.sort() >>> temps [67, 73, 78, 85, 92, 93, 95, 100, 104] >>> def reverser(n1, n2): ... '''Comparison function to use reverse order. ... ''' ... return cmp(n2, n1) ... >>> temps.sort(reverser) >>> temps [104, 100, 95, 93, 92, 85, 78, 73, 67] >>> def unitsDigit(n): ... '''Returns only the units digit of n. ... ''' ... return n % 10 ... >>> temps.sort(None, unitsDigit) >>> temps [100, 92, 93, 73, 104, 95, 85, 67, 78] >>> temps.sort(None, None, True) >>> temps [104, 100, 95, 93, 92, 85, 78, 73, 67]

List comprehensions You can use a form called a list comprehension to create a list. The general form is: [ e for v1 in s1 for v2 in s2 ... if c ] where e is some expression, followed by one or more for clauses, optionally followed by an if clause. The result is a list containing all the values of expression e after all the nested for loops have been run; the for loops have the same structure as in . If there is an “if” clause, it determines which values of e are added to the list: if the if condition is true, the value is added, otherwise it is not added. This is perhaps easiest to explain with a few examples. In the first example, we construct a list containing the cubes of the numbers from 1 to 10, inclusive. The for loop generates the numbers 1, 2, ..., 10, and then the expression “x**3” cubes each one and appends it to the resulting list. >>> range(1, 11) [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] >>> [ x**3 for x in range(1,11 ) ] [1, 8, 27, 64, 125, 216, 343, 512, 729, 1000] In the next example, we use two for loops. The outer loop generates the sequence [1, 2, 3], and the inner loop generates the sequence [50, 51]. The expression “x*1000 + y” is computed for each of the resulting six value sets for x and y, and the result is appended to the list being built. >>> [ x*1000 + y ... for x in range(1,4) ... for y in range(50, 52) ] [1050, 1051, 2050, 2051, 3050, 3051] In the next example, there are two nested loops, each generating the sequence [0, 1, 2]. For each of the nine trips through the inner loop, we test the values of x and y and discard the cases where they are equal. The expression “(y, x)” combines the two values into a 2-tuple. >>> [ (y, x) ... for y in range(3) ... for x in range(3) ... if x != y ] [(0, 1), (0, 2), (1, 0), (1, 2), (2, 0), (2, 1)]

Type <code >tuple</code >: Immutable sequences For representing a sequence of values, Python has two similar container types: list and tuple. An instance of a container type is basically a value that has other values inside it; we call the contained values elements. So, when should you use a list, and when a tuple? In many contexts, either will work. However, there are important differences. Values of type list are mutable; that is, you can delete or add elements, or change the value of any of the elements inside the list. Lists cannot be used in certain contexts. For example, you can't use a list as the key in a dictionary. >>> d={} >>> d[(23,59)] = 'hike' >>> d[[46,19]] = 'hut' Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: list objects are unhashable Values of type tuple are immutable. Once you have assembled a tuple, you cannot add or delete elements or change the value of an element inside the tuple. Among the reasons to use a tuple instead of a list: Tuples are allowed in certain contexts where lists are not, such as the right-hand argument of the string format operator, or as a key in a dictionary. If your program is in danger of running out of processor memory, tuples are slightly more efficient in their memory usage. Write a literal tuple as a sequence of values in parentheses, separated by commas. There may be any number of values of any type or any mixture of types. There may be zero values. To write a tuple with exactly one value, you must use this special syntax: (value,) That is, you must provide a comma before the closing “)”, in order to show that it is a tuple, and not just a parenthesized expression. Note especially the last two examples below: >>> () () >>> ('farcical', 'aquatic', 'ceremony') ('farcical', 'aquatic', 'ceremony') >>> ('Ni',) ('Ni',) >>> ('Ni') 'Ni' You may also convert any iterable into a tuple using . The tuple type does not have comprehensions (see ), but you can get the equivalent by applying the tuple() function to a list comprehension. Here is an example: >>> tuple([x**2 for x in (2,3,8)]) (4, 9, 64) >>> Tuples also support the .index() and .count() methods as described in .

The <code >bytes</code > type To understand why Python version 2.6 and beyond have a bytes type, it is necessary to review a little history. Most early computing used 7- and 8-bit character codes, but these character sets are very limited. In particular, life was difficult for Francophone countries when “è” and “é” are very different letters. The 32-bit character set of the Unicode standard is the current preferred practice, and provides enough characters to last a good while into the future. Text handling in the Python 2.x releases was awkward due to the presence of two different types for representing character data: str and unicode. Consequently, in the upcoming major incompatible 3.x releases, all character data will be represented internally by 32-bit characters. Therefore, in Python 2.6 the bytes type was added to aid transition to the 3.0 family, which has a separate bytes type for 8-bit character strings. In the 3.x versions, a bytes value is a sequence of zero or more unsigned 8-bit integers, each in the range 0–255, inclusive. In Python 2.6 and subsequent versions, the bytes type is a synonym for str. The bytes() function works exactly like the str() function. >>> s=bytes(987) >>> s '987' >>> type(s) <type 'str'> Use this type where your program expects 8-bit characters, and it will ease the transition to Python 3.x, because the semi-automated translation process will know that values of bytes type are intended for sequences of 8-bit characters.

Using the <code >bytes</code > type in 3.x conversion Versions 2.6+ support a new notation: to create a literal of type bytes, place a “b” just before the opening quote. >>> s = b'abc' >>> s 'abc' >>> type(s) <type 'str'> Such literals are exactly like regular string literals. The difference comes when you convert your program to the 3.x versions. In Python 3.x, a string of the form b'...' will have type bytes, which will be different than the str (32-bit character) type in 3.x. One step in converting your 2.x programs to 3.x is to add this import before all the other imports in your program: from __future__ import unicode_literals In programs that start with this declaration, all string literals will automatically be considered unicode type without using the u'...' prefix. This means you may also include escape sequences of the form '\uXXXX', each of which designates a 16-bit Unicode code point as four hexadecimal digits XXXX. Here is a demonstration of the difference. Before the import, the \u escape is not recognized, and the value has type str. Afterwards, the return value is type unicode >>> s = '\u2672' >>> len(s) 6 >>> s '\\u2672' >>> type(s) <type 'str'> >>> from __future__ import unicode_literals >>> t = '\u2672' >>> len(t) 1 >>> type(t) <type 'unicode'> >>> t u'\u2672'

The <code >bytearray</code > type New in Python 2.6 is the bytearray type. Each instance is a sequence of 8-bit bytes, each of which is an unsigned integer in the range 0 <= 255. Unlike the str type, however, bytearray values are mutable: you can delete, insert, or replace arbitrary values or slices. As with the features described in , this type is intended to ease the transition to Python 3.x versions. Use it for situations where you are handling sequences of 8-bit bytes that are not intended as textual representations, such as raw binary data. Values of this type support almost all of the operators and methods of the str type (with the exception of .encode() and .format() methods). They also support these methods of the list type: .extend(), .insert(), .pop(), .remove(), .reverse(). You can also replace values using either integers or the b'...' (bytes) literals. Some examples: >>> s=bytearray('abcdef') >>> s bytearray(b'abcdef') >>> type(s) <type 'bytearray'> >>> s[3] 100 >>> s.insert(0, b'^') >>> s bytearray(b'^abcdef') >>> s.reverse() >>> s bytearray(b'fedcba^') >>> s[2:6] bytearray(b'dcba') >>> s[2:6] = b'#' >>> s bytearray(b'fe#^') >>> s[0]=63 >>> s bytearray(b'?e#^') >>> The bytearray type also has a static method named .fromhex() that creates a bytearray value from a Unicode string containing hexadecimal characters (which may be separated by spaces for legibility). >>> ao = bytearray.fromhex(u'00 ff') >>> ao bytearray(b'\x00\xff') >>> ao[1] 255

Types <code >set</code > and <code >frozenset</code >: Set types Mathematically speaking, a set is an unordered collection of zero or more distinct elements. Python has two set types that represent this mathematical abstraction. Use these types when you care only about whether something is a member of the set or not, and you don't need them to be in any specific order. The elements of a Python set must be immutable. In particular, you can't have list or dictionary elements in a set. Most operations on sets work with both set and frozenset types. Values of type set are mutable: you can add or delete members. There are two ways to create a mutable set. In all Python versions of the 2.x series, the set(S) function operates on a sequence S and returns a mutable set containing the unique elements of S. The argument is optional; if omitted, you get a new, empty set. >>> s1 = set([1, 1, 1, 9, 1, 8, 9, 8, 3]) set([8, 1, 3, 9]) >>> s1 = set([1, 1, 1, 9, 1, 8, 9, 8, 3]) >>> s2 = set() >>> s1 set([8, 1, 3, 9]) >>> s2 set([]) >>> print len(s1), len(s2) 4 0 >>> s3 = set("notlob bolton") >>> s3 s3 set([' ', 'b', 'l', 'o', 'n', 't']) Starting in Python 2.7, you can create a set by simply enclosing one or more elements within braces {...} separated by commas. s1 = {1, 1, 1, 9, 1, 8, 9, 8, 3} >>> s1 set([8, 9, 3, 1]) Note the wording “one or more:” an empty pair of braces “{}” is an empty dictionary, not an empty set. A frozenset value is immutable: you can't change the membership, but you can use a frozenset value in contexts where set values are not allowed. For example, you can use a frozenset as a key in a dictionary, but you can't use a set value as a dictionary key. To create a set or frozenset, see and . A number of functions that work on sequences also work on sets. In each case, the set is converted to a list before being passed to the function. . Predicate to test whether all members of a set are True. . Predicate to test whether any member of a set is true. . Returns a list of the elements that pass through a filtering function. . Returns an iterator that will visit every element of the set. . Returns the length (cardinality) of the set. . Returns the elements of the set as a list. . Returns a list containing the result of the application of a function to each element of a set. . Returns the largest element of a set. . Returns the smallest element of a set. . Returns the result of the application of a given function pairwise to all the elements of a set. . Returns a list containing the sorted elements of the set. Another new feature in Python 2.7 is the set comprehension. This is similar to the feature described in . Here is the general form: { e for v1 in s1 for v2 in s2 ... if c } As with a list comprehension, you use one or more for clauses to iterate over sets of values, and the expression e is evaluated for every combination of the values in the sequences si. If there is no “if” clause, or if the “if” condition evaluates as True, the value is added to the sequence from which a set is then constructed. Here is an example. Function takeUppers() takes one string argument and returns a set of the unique letters in that string, uppercased. The for clause iterates over the characters in the argument s; the if clause discards characters that aren't letters; and the .upper() method converts lowercase letters to uppercase. >>> def takeUpper(s): ... return { c.upper() ... for c in s ... if c.isalpha() } ... >>> takeUpper("A a|ccCc^#zZ") set(['A', 'C', 'Z'])

Operations on mutable and immutable sets These operations are supported by both set and frozenset types: x in S Predicate that tests whether element x is a member of set S. >>> 1 in set([0,1,4]) True >>> 99 in set([0,1,4]) False x not in S Predicate that tests whether element x is not a member of set S. >>> 1 not in set([0,1,4]) False >>> 99 not in set([0,1,4]) True S1 == S2 Predicate that tests whether sets S1 and S2 have exactly the same members. >>> set('bedac') == set('abcde') True >>> set('bedac') == set('bedack') False S1 != S2 Predicate that tests whether sets S1 and S2 have different members. >>> set ( 'bedac' ) != set ( 'abcde' ) False >>> set('bedac')!=set('bedack') True S1 < S2 Predicate that tests whether S1 is a proper subset of S2; that is, all the elements of S1 are also members of S2, but there is at least one element of S2 that is not in S1. >>> set('ab') < set('ab') False >>> set('ab') < set('abcde') True S1 > S2 Predicate that tests whether S1 is a proper superset of S2; that is, all the elements of S2 are also members of S1, but there is at least one element of S1 that is not in S2. >>> set('ab') > set('ab') False >>> set('abcde') > set('cd') True S.copy() Return a new set of the same type as S, containing all the same elements. >>> s1=set('aeiou') >>> s2=s1 >>> s3=s1.copy() >>> s1.add('y') >>> s1 set(['a', 'e', 'i', 'o', 'u', 'y']) >>> s2 set(['a', 'e', 'i', 'o', 'u', 'y']) >>> s3 set(['a', 'i', 'e', 'u', 'o']) S1.difference(S2) Returns a new set of the same type as S1, containing only those values found in S1 but not found in S2. The S2 argument may be a set or a sequence. >>> set('roygbiv').difference('rgb') set(['i', 'o', 'v', 'y']) S1 - S2 Same as S1.difference(S2), except that S2 must be a set. >>> set('roygbiv') - set('rgb') set(['i', 'y', 'o', 'v']) S1.intersection(S2) Returns a new set, of the same type as S1, containing only the elements found both in S1 and S2. S2 may be a set or a sequence. >>> set([1,2,3,5,7,11]).intersection(set([1,3,5,7,9])) set([1, 3, 5, 7]) >>> set([1,3,5]).intersection( (2,4,6,8) ) set([]) S1 & S2 Same as S1.intersection(S2), but S2 must be a set. S1.issubset(S2) Predicate that tests whether every element of S1 is also in S2. S2 may be a set or a sequence. >>> set([1,2]).issubset(set([2,4,1,8])) True >>> set([2,4,1,8]).issubset(set([1,2])) False >>> set(['r', 'g', 'b']) <= set(['r', 'o', 'y', 'g', 'b', 'i', 'v']) True S1 <= S2 Same as S1.issubset(S2), but S2 must be a set. S1.issuperset(S2) Predicate that tests whether every element of S2 is also in S1. S2 may be a set or a sequence. >>> set([1,2]).issuperset(set([2,4,1,8])) False >>> set([2,4,1,8]).issuperset(set([1,2])) True S1 >= S2 Same as S1.issuperset(S2). S1.symmetric_difference(S2) Returns a new set of the same type as S1, containing only elements found in S1 or S2, but not found in both. The S2 argument may be a set or a sequence. >>> set('aeiou').symmetric_difference('etaoin') set(['n', 'u', 't']) S1 ^ S2 Same as S1.symmetric_difference(S2), but S2 must be a set. S1.union(S2) Returns a new set, with the same type as S1, containing all the elements found in either S1 or S2. The S2 argument may be a set or a sequence. >>> set([1,2]).union(set([1,3,7])) set([1, 2, 3, 7]) >>> set([1,2]).union( (8,2,4,5) ) set([8, 1, 2, 4, 5]) S1 | S2 Same as S1.union(S2).

Operations on mutable sets The operations described in this section apply to set (mutable) values, but may not be used with frozenset (immutable) values. S.add(x) Add element x to set S. Duplicate elements will be ignored. >>> pbr=set(['USA', 'Brazil', 'Canada']) >>> pbr.add('Australia') >>> pbr set(['Brazil', 'Canada', 'Australia', 'USA']) >>> pbr.add('USA') >>> pbr set(['Brazil', 'Canada', 'Australia', 'USA']) S.clear() Remove all the elements from set S. >>> pbr set(['Brazil', 'USA']) >>> pbr.clear() >>> pbr set([]) S.discard(x) If set S contains element x, remove that element from S. If x is not in S, it is not considered an error; compare S.remove(x). >>> pbr set(['Brazil', 'Australia', 'USA']) >>> pbr.discard('Swaziland') >>> pbr set(['Brazil', 'Australia', 'USA']) >>> pbr.discard('Australia') >>> pbr set(['Brazil', 'USA']) S1.difference_update(S2) Modify set S1 by removing any values found in S2. Value S2 may be a set or a sequence. >>> s1=set('roygbiv') >>> s1.difference_update('rgb') >>> s1 set(['i', 'o', 'v', 'y']) S1 -= S2 Same as S1.difference_update(S2), but S2 must be a set. S1.intersection_update(S2) Modify set S1 so that it contains only values found in both S1 and S2. >>> s1=set('roygbiv') >>> s1 set(['b', 'g', 'i', 'o', 'r', 'v', 'y']) >>> s1.intersection_update('roy') >>> s1 set(['y', 'r', 'o']) S1 &= S2 Same as S1.intersection_update(S2), but S2 must be a set. S.remove(x) If element x is in set S, remove that element from S. If x is not an element of S, this operation will raise a KeyError exception. >>> pbr set(['Brazil', 'Canada', 'Australia', 'USA']) >>> pbr.remove('Canada') >>> pbr set(['Brazil', 'Australia', 'USA']) >>> pbr.remove('Swaziland') Traceback (most recent call last): File "<stdin>", line 1, in <module> KeyError: 'Swaziland' S1.symmetric_difference_update(S2) Remove from S1 any elements found in both S1 and S2. Value S2 may be a set or a sequence. >>> s1=set('abcd') >>> s1.symmetric_difference_update('cdefg') >>> s1 set(['a', 'b', 'e', 'g', 'f']) S1 ^= S2 Same as S1.symmetric_difference_update(S2), but S2 must be a set. S1.update(S2) Add to S1 any elements of S2 not found in S1. The S2 argument may be a set or a sequence. >>> s1=set('rgb') >>> s1 set(['r', 'b', 'g']) >>> s1.update('roygbiv') >>> s1 set(['b', 'g', 'i', 'o', 'r', 'v', 'y']) S1 |= S2 Same as S1.update(S2), but S2 must be a set.

Type <code >dict</code >: Dictionaries Python dictionaries are one of its more powerful built-in types. They are generally used for look-up tables and many similar applications. A Python dictionary represents a set of zero or more ordered pairs (ki, vi) such that: Each ki value is called a key; each key is unique and immutable; and the associated value vi can be of any type. Another term for this structure is mapping, since it maps the set of keys onto the set of values (in the algebraic sense). To create a new dictionary, use this general form: { k0: v0, k1: v1, ... } There can be any number of key-value pairs (including zero). Each key-value has the form “ki:vi”, and pairs are separated by commas. Here are some examples of dictionaries: {} {'Bolton': 'Notlob', 'Ipswich': 'Esher'} {(1,1):48, (8,20): 52} For efficiency reasons, the order of the pairs in a dictionary is arbitrary: it is essentially an unordered set of ordered pairs. If you display a dictionary, the pairs may be shown in a different order than you used when you created it. >>> signals = {0:'red', 1: 'yellow', 2:'green'} >>> signals {2: 'green', 0: 'red', 1: 'yellow'}

Operations on dictionaries These operations are available on any dictionary object D: len(D) Returns the number of key-value pairs in D. D[k] If dictionary D has a key whose value is equal to k, this operation returns the corresponding value for that key. If there is no matching key, it raises a KeyError exception. >>> signals = {0: 'red', 1: 'yellow', 2: 'green'} >>> signals[2] 'green' >>> signals[88] Traceback (most recent call last): File "<stdin>", line 1, in <module> KeyError: 88 D[k] = v If dictionary D does not have a key-value pair whose key equals k, a new pair is added with key k and value v. If D already has a key-value pair whose key equals k, the value of that pair is replaced by v. k in D A predicate that tests whether D has a key equal to k. >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> 3 in roster True >>> 88 in roster False k not in D A predicate that tests whether D does not have a key equal to k. >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> 3 not in roster False >>> 88 not in roster True del D[k] In Python, del is a statement, not a function; see . If dictionary D has a key-value pair whose key equals k, that key-value pair is deleted from D. If there is no matching key-value pair, the statement will raise a KeyError exception. >>> rgb = {'red':'#ff0000', 'green':'#00ff00', 'blue':'#0000ff'} >>> del rgb['red'] >>> rgb {'blue': '#0000ff', 'green': '#00ff00'} >>> del rgb['cerise'] Traceback (most recent call last): File "<stdin>", line 1, in <module> KeyError: 'cerise' D.get(k, x) If dictionary D has a key equal to x, it returns the corresponding value, that is, it is the same as the expression “D[x]”. However, if D has no key-value pair for key k, this method returns the default value x. The second argument is optional; if omitted, and D has no key equal to k, it returns None. >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> roster.get(2) 'Ray' >>> v = roster.get(8) >>> print v None >>> roster.get(2, 'Not found') 'Ray' >>> roster.get(8, 'Not found') 'Not found' D.has_key(k) A predicate that returns True if D has a key k. >>> signals = {0: 'red', 1: 'yellow', 2: 'green'} >>> signals.has_key(1) True >>> signals.has_key(88) False D.items() Returns the contents of dictionary D as a list of two-element tuples

(k,
              v)

, in no particular order. >>> signals = {0: 'red', 1: 'yellow', 2: 'green'} >>> signals.items() [(0, 'red'), (1, 'yellow'), (2, 'green')] D.iteritems() Returns an iterator that generates the values from dictionary D as a sequence of two-element tuples (k, v). See . >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> rosterScan = roster.iteritems() >>> for n, name in rosterScan: ... print "{0:04d}: {1}".format(n, name) ... 0001: Pat 0002: Ray 0003: Min D.iterkeys() Returns an iterator that generates the keys from dictionary D. See . >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> nScan = roster.iterkeys() >>> for n in nScan: ... print n, ... 1 2 3 D.itervalues() Returns an iterator that generates the values from dictionary D. See . >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> nameScan = roster.itervalues() >>> for name in nameScan: ... print name, ... Pat Ray Min D.keys() Returns a list of the key values in dictionary D, in no particular order. >>> signals = {0: 'red', 1: 'yellow', 2: 'green'} >>> signals.keys() [1, 0, 2] D.popitem() Returns an arbitrary entry from dictionary D as a (key, value) tuple, and also removes that entry. If D is empty, raises a KeyError exception. >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> roster.popitem() (1, 'Pat') >>> roster {2: 'Ray', 3: 'Min'} D.setdefault(k, x) If dictionary D has a key equal to k, this method returns the corresponding value D[k]. If D has no key equal to k, the method returns the default value x. However, unlike the .get() method, it also creates a new key-value pair (k, x) in D. As with the .get() method, the second argument is optional, and defaults to the value None. D.values() Returns a list of the values from key-value pairs in dictionary D, in no particular order. However, if you call both the .items() and .values() methods of a dictionary without changing that dictionary's contents between those calls, Python guarantees that the ordering of the two results will be the same. >>> signals = {0: 'red', 1: 'yellow', 2: 'green'} >>> signals.values() ['yellow', 'red', 'green'] >>> signals.keys() [1, 0, 2] D.update(D2) Merge the contents of dictionary D2 into dictionary D. For any key-value pairs that have the same key in both D and D2, the value for that key in D after this operation will be the value from D2, not the value from D. >>> roster={1:'Pat', 2:'Ray', 3:'Min'} >>> newer={3:'Bev', 4:'Wes'} >>> roster.update(newer) >>> roster {1: 'Pat', 4: 'Wes', 2: 'Ray', 3: 'Bev'} >>> newer {3: 'Bev', 4: 'Wes'}

Dictionary comprehensions New in Python 2.7 are dictionary comprehensions: a construct that allows you to build a dictionary dynamically, somewhat like . Here is the general form: { ek: ev for v1 in s1 for v2 in s2 ... if c } As with list comprehensions, you provide one or more for clauses and an optional if clause. For all possible combinations of the values in the for clauses that have a true value for the if clause, two expressions ek and ev are evaluated, and a new dictionary entry is added with key ek and value ev. Here is an example. The Wikipedia article on the game of Scrabble gives the Scrabble score for each letter of the alphabet. What we would like is a dictionary whose keys are letters, and each related value is the score. However, the Wikipedia article shows the score values grouped by score: the 1's together, the 2's together, and so on. So, to make it easy to check that we have entered the right score values and letters, we can use a list of tuples, where the first element of each tuple is the score and the second element is a string of all the letters with that score. We can then convert that list to the desired dictionary using a dictionary comprehension. >>> scrabbleTuples = [ (1, "EAOINRTLSU"), (2, "DG"), (3, "BCMP"), ... (4, "FHVWY"), (5, "K"), (8, "JX"), (10, "QZ")] >>> scrabbleMap = { letter: score ... for score, letterList in scrabbleTuples ... for letter in letterList } >>> scrabbleMap['A'] 1 >>> scrabbleMap['Z'] 10 Evaluating the set comprehension proceeds as follows. In the first for clause, the first tuple is unpacked, setting score to 1 and letterList to "EAOINRTLSU". In the second for clause, letter is set to "E". A new entry is added to the dictionary, with key "E" and value 1. In the second for clause, letter is set to "A". A new entry is added with key "A" and value 1. Execution proceeds in this manner until all the for clauses are complete. Then the name scrabbleMap is bound to the resulting dictionary.

Type <code >file</code >: Input and output files To open a file, use this general form: f = open(name[,mode[,bufsize]]]) name The path name of the file to be opened, as a string. mode An optional string specifying what you plan to do with the file. If omitted, you will get read access to the file. In general the value consists of three parts: General mode, one of: r Read access. The file must already exist. You will not be allowed to write to it. w Write access. If there is no file by this name, a new one will be created. If there is an existing file, it will be deleted! a Append access. If there is a file by this name, your initial position will be at the end of the file, and you will be allowed to write (and read). If there is no file by this name, a new one will be created. On some systems, all writes to a file with append mode are added at the end of the file, regardless of the current file position. If you plan to modify the file, append a “+” next. For example, mode “r+” puts you at the beginning of an existing file and allows you to write to the file anywhere. Mode “w+” is the same as “w”: it deletes an existing file if there is any, then creates a new file and gives you write access. Mode “a+” allows you to write new data at the end of an existing file; if no file by this name exists, it will create a new one. If you are handling binary data, as opposed to lines of text, add “b” at the end of the mode string. For modes beginning with 'r', you may append a capital 'U' to request universal newline treatment. This is handy when you are reading files made on a platform with different line termination conventions. When reading lines from a file opened in this way, any line terminator ('\n', '\r', or '\r\n') will appear in the return value as the standard '\n'. Also, files so opened will have an attribute named .newlines; this attribute will be None initially, but after any line terminators have been read, it will be a tuple containing all the different line terminator strings seen so far. bufsize Buffer size: this affects when physical device writes are done, compared to write operations that your program performs. In most cases you will probably want to omit this argument. The default is to use line buffering for terminal-type devices, or some system default for other devices. Use 0 to force unbuffered operation. This may be inefficient, but any file writes are performed immediately. Use 1 for line buffering: output lines are written whenever you write a line terminator such as '\n'. Use larger values to specify the actual size of the buffer. Use a negative value to request the system defaults. If you are reading text files, and you don't want to worry about the variety of line termination protocols, you may use a mode value of “U” for “universal line terminator mode.” In this mode, input lines may be terminated with either carriage return ('\r'), newline ('\n'), or both, but the lines you receive will always be terminated with a single newline. (Exception: If the last line is unterminated, the string you get will also be unterminated.) There are a number of potential error conditions. For modes starting with “r”, the file must exist before you open it. Also, you must have access according to the underlying operating system. For example, in Linux environments, you must have read access to read a file, and you must have write access to modify or delete a file. These sorts of failures will raise an IOError exception. A file is its own iterator (see ). Hence, if you have a file inFile opened for reading, you can use a for loop that looks like this to iterate over the lines of the file: for line in inFile: ... The variable line will be set to each line of the file in turn. The line terminator character (if any) will be present in that string. Other aspects of files: Every open file has a current position. Initially, this will 0L if you opened it for reading or writing, or the size of the file if you opened it with append access. Each write or read operation moves this position by the amount read or written. You can also query or set the file position; see . Files may use a technique called buffering. Because physical access to some storage media (such as disk drives) takes a relatively long time, Python may employ a storage area called a buffer as a holding area for data being input or output. For example, if you are writing data to a disk file, Python may keep the data in the file's buffer area until the buffer is full and only then actually write it to the physical disk. There are various techniques for controlling this behavior; see .

Methods on <code >file</code > objects Use these methods on an open file instance F. F.close() Close file F. Any unwritten data in the buffer will be flushed. No further operations will be allowed on the file unless it is reopened with the open() function. F.flush() For buffered files, you can use this method to make sure that all data written to the file has been physically transmitted to the storage medium. Closing a file will also flush the buffers. Avoid using this method unless you really need it, as it may make your program less efficient. F.isatty() A predicate that tests whether F is a terminal; “tty” is an ancient term that originally meant “Teletype”, but has come to mean any terminal or simulated terminal. F.read(n ) Read the next n characters from F and return them as a string. If there are fewer than n characters remaining after your current position, you will get all remaining characters. If you are at the end of the file, you will get back an empty string (''). The argument is optional. If omitted, you will get the entire remaining contents of the file as a string. F.readline(maxlen) Read the next text line from F and return it as a string, including the line terminator if any. If you need to limit the maximum size of incoming lines, pass that size limit as the optional maxlen argument. The default is to return a line of any size (subject to memory limitations). If the line exceeds maxlen, the file position will be left pointing to the first unread character of that line. F.readlines() Read all remaining lines from F and return them as a list of strings, including line terminator characters if any. F.seek(offset, whence) Change the file's current position. The value of whence determines how the offset value is used to change the position: 0: Set the position to offset bytes after the beginning of the file. 1: Move the current position offset bytes toward the end of the file. 2: Move the current position offset bytes relative to the end of the file. For example, for a file f, this operation would set the position to 4 bytes before the end of the file: f.seek(-4, 2) F.tell() This method returns the current file position relative to the beginning as a long value. >>> f=open('roundtable', 'w') >>> f.write('Bedevere') >>> f.tell() 8L >>> f.seek(2L) >>> f.tell() 2L F.truncate([pos]) Remove any contents of F past position pos, which defaults to the current position. F.write(s) Write the contents of string s to file F. This operation will not add terminator characters; if you want newlines in your file, include them in the string s. F.writelines(S) For a sequence S containing strings, write all those strings to F. No line terminators will be added; you must provide them explicitly if you want them.

<code >None</code >: The special placeholder value Python has a unique value called None. This special null value can be used as a placeholder, or to signify that some value is unknown. In conversational mode, any expression that evaluates to None is not printed. However, if a value of None is converted to a string, the result is the string 'None'; this may happen, for example, in a print statement. >>> x = None >>> x >>> print x None The value None is returned from any function that executes a return statement with no value, or any function after it executes its last line if that last line is not a return statement. >>> def useless(): ... print "Useless!" ... >>> useless() Useless! >>> z=useless() Useless! >>> z >>> print z None >>> def boatAnchor(): ... pass ... >>> x=boatAnchor() >>> x >>> print x None

Operators and expressions Python's operators are shown here from highest precedence to lowest, with a ruled line separating groups of operators with equal precedence: Python operator precedence (E) Parenthesized expression or tuple. [E, ...] List. {key:value, ...} Dictionary or set. `...` Convert to string representation. x.attribute Attribute reference. x[...] Subscript or slice; see . f(...) Call function f. x**y x to the y power. -x Negation. ~x Bitwise not (one's complement). x*y Multiplication. x/y, x//y Division. The “//” form discards the fraction from the result. For example, “13.9//5.0” returns the value 2.0. x%y Modulo (remainder of x/y). x+y Addition, concatenation. x-y Subtraction. x<<y x shifted left ybits. x>>y x shifted right ybits. x&y Bitwise and. x^y Bitwise exclusive or. x|y Bitwise or. x<y, x<=y, x>y, x>=y, x!=y, x==y Comparisons. These operators are all predicates; see . x in y, x not in y Test for membership. x is y, x is not y Test for identity. not x Boolean “not.” x and y Boolean “and.” x or y Boolean “or.”

What is a predicate? We use the term predicate to mean any Python function that tests some condition and returns a Boolean value. For example, x < y is a predicate that tests whether x is less than y. For example, 5 < 500 returns True, while 5 >= 500 returns False.

What is an iterable? To iterate over a sequence means to visit each element of the sequence, and do some operation for each element. In Python, we say that a value is an iterable when your program can iterate over it. In short, an iterable is a value that represents a sequence of one more values. All instances of Python's sequence types are iterables. These types may be referred to as container types: a unicode string string is a container for 32-bit characters, and lists and tuples are general-purpose containers that can contain any sequence. One of the most common uses for an iterable is in a for statement, where you want to perform some operation on a sequence of values. For example, if you have a tuple named celsiuses containing Celsius temperatures, and you want to print them with their Fahrenheit equivalents, and you have written a function cToF() that converts Celsius to Fahrenheit, this code does it: >>> def cToF(c): return c*9.0/5.0 + 32.0 ... >>> celsiuses = (0, 20, 23.6, 100) >>> for celsius in celsiuses: ... print "{0:.1f} C = {1:.1f} F".format(celsius, cToF(celsius)) ... 0.0 C = 32.0 F 20.0 C = 68.0 F 23.6 C = 74.5 F 100.0 C = 212.0 F However, Python also supports mechanisms for lazy evaluation: a piece of program that acts like a sequence, but produces its contained values one at a time. Keep in mind that the above code works exactly the same if celsiuses is an iterator (see ). You may find many uses for iterators in your programs. For example, celsiuses might be a system that goes off and reads an actual thermometer and returns the readings every ten seconds. In this application, the code above doesn't care where celsiuses gets the values, it cares only about how to convert and print them.

Duck typing, or: what is an interface?

James Whitcomb Riley When I see a bird that walks like a duck and swims like a duck and quacks like a duck, I call that bird a duck.

The term duck typing comes from this quote. In programming terms, this means that the important thing about a value is what it can do, not its type. As the excellent Wikipedia article on duck typing says, “Simply stated: provided you can perform the job, we don't care who your parents are.” One common example of duck typing is in the Python term “file-like object”. If you open a file for reading using the open() function, you get back a value of type file: >>> inFile = open('input') >>> type(inFile) <type 'file'> Let's suppose that you write a function called numberIt() that takes a readable file as an argument and prints the lines from a file preceded by five-digit line numbers. Here's the function and an example of its use: >>> def numberIt(f): ... for lineNo, line in enumerate(f): ... print "{0:05d} {1}".format(lineNo, line.rstrip()) ... >>> numberIt(inFile) 00000 Kant 00001 Heidegger 00002 Hume The way you have written the numberIt() function, it works for files, but it also works for any iterable. Thus, when you see the statement that some Python feature works with a “file-like object,” that means that the object must have an interface like that of the file type; Python doesn't care about the type, just the operations that it supports. In practice, the enumerate() function works with any iterable, so your function will also work with any iterable: >>> numberIt(['Kant', 'Heidegger', 'Hume']) 00000 Kant 00001 Heidegger 00002 Hume So in Python when we say that we expect some value to have an interface, we mean that it must provide certain methods or functions, but the actual type of the value is immaterial. More formally, when we say that a value supports the iterable interface, that value must provide either of the following features: A .__getitem__() method as described in . A .__iter__() method as described in .

What is the locale? In order to accommodate different character encodings, your system may have a locale setting that specifies a preferred character set. In the USA, most systems use the ASCII encoding. Good application code should be written in a way that does not depend on this encoding to deal with cultural issues. For general information on handling locale issues, see the documentation for the locale module.

Basic functions Python has a lot of built-in functions. This section describes the ones that most people use most of the time. If you are interested in exploring some of the more remote corners of Python, see .

<code >abs()</code >: Absolute value To find the absolute value of a number x: abs(x) If x is negative, the function returns -x; otherwise it returns x. For complex values, the function returns the magnitude, that is, the square root of (x.real**2+x.imag**2). >>> abs(-33) 33 >>> abs(33) 33 >>> abs(0) 0 >>> abs(complex(1,5)) 5.0990195135927845

<code >all()</code >: Are all the elements of an iterable true? A predicate that tests whether all the elements of some iterable are considered True. If any elements are not already type bool, they are converted to Boolean values using the bool() built-in function. >>> all([True, 14, (88,99)]) True >>> all([True, 14, (88,99), None]) False

<code >any()</code >: Are any of the members of an iterable true? This function, applied to some iterable, is a predicate that tests whether any of the elements of that iterable are True. If any element is not already type bool, it is converted to a Boolean value using the bool() built-in function. >>> noneTrue = (0, 0.0, (), None) >>> any(noneTrue) False >>> someTrue = (0, 0.0, (88,), 'penguin') >>> any(someTrue) True

<code >bin()</code >: Convert to binary This function takes an integer argument and returns a string that represents that number in binary (base 2) starting with '0b'. >>> bin(7) '0b111' >>> bin(257) '0b100000001'

<code >bool()</code >: Convert to Boolean This function takes any value x and converts it to a Boolean (true or false) value. For the list of values that are considered True or False, see . Examples: >>> bool(0) False >>> bool(0.0) False >>> bool(0L) False >>> bool(0j) False >>> bool('') False >>> bool([]) False >>> bool(()) False >>> bool({}) False >>> bool(None) False >>> bool(1) True >>> bool(15.9) True >>> bool([0]) True >>> bool((None,)) True >>> bool({None: False}) True

<code >bytearray()</code >: Create a byte array See .

<code >chr()</code >: Get the character with a given code For arguments n in the range 0 to 255, this function returns a one-character string containing the character that has code n. Compare . >>> chr(65) 'A' >>> chr(0) '\x00'

<code >cmp()</code >: Compare two values This function compares the values of two arguments x and y: cmp(x, y) The return value is: A negative number if x is less than y. Zero if x is equal to y. A positive number if x is greater than y. The built-in cmp() function will typically return only the values -1, 0, or 1. However, there are other places that expect functions with the same calling sequence, and those functions may return other values. It is best to observe only the sign of the result. >>> cmp(2,5) -1 >>> cmp(5,5) 0 >>> cmp(5,2) 1 >>> cmp('aardvark', 'aardwolf') -1

<code >complex()</code >: Convert to <code >complex</code > type To create a complex number from a real part R and a complex part I: complex(R, I) Both arguments are optional. With two arguments, both arguments must be numbers of any numeric type. With one numeric argument, it returns a complex number with real part R and an imaginary part of zero. To convert a complex number in string form, pass that string as the only argument. If the string is not a valid complex number, the function raises a ValueError exception. If called with no arguments, it returns a complex zero. Examples: >>> complex(-4.04, 3.173) (-4.04+3.173j) >>> complex(-4.04) (-4.04+0j) >>> complex() 0j >>> c1=4+5j >>> c2=6+7j >>> complex(c1, c2) # Equals (4+5j)+(6+7j)j = 4+5j+6j-7 (-3+11j) >>> c1 + c2*1.0j (-3+11j)

<code >dict()</code >: Convert to a dictionary This function creates a new dictionary from its arguments. The general form is: dict(v, k0=v0, k1=v1, ...) That is, there may be one optional positional argument or any number of keyword arguments. If you supply no arguments, you get a new, empty dictionary. If one positional argument is supplied, it must be a iterable containing two-element iterables. Each two-element iterable becomes one key-value pair of the result. >>> dict() {} >>> dict ( [ (0, 'stop'), (1, 'go') ] ) {0: 'stop', 1: 'go'} >>> dict((('y', 'boy'), ('x', 'girl'))) {'y': 'boy', 'x': 'girl'} If you supply any keyword arguments, each keyword becomes a key in the resulting dictionary, and that argument's value becomes the corresponding value of that key-value pair. >>> dict(bricks='sleep', keith='maniac', rj='gumby') {'bricks': 'sleep', 'keith': 'maniac', 'rj': 'gumby'}

<code >divmod()</code >: Quotient and remainder divmod(x, y) Sometimes you want both the quotient and remainder when dividing x by y. This function returns a tuple (q, r), where q is the quotient and r is the remainder. If either x or y is a float, the returned value q is the whole part of the quotient, and the returned r is computed as x-(r*d). Examples: >>> divmod(13, 5) (2, 3) >>> divmod(1.6, 0.5) (3.0, 0.10000000000000009)

<code >enumerate()</code >: Step through indices and values of an iterable Given an iterable S, enumerate(S) produces an iterator that iterates over the pairs of values

(i,
        S[i])

, for i having the values in range(len(S)). For more information on iterators, see . >>> L = ['Ministry', 'of', 'Silly', 'Walks'] >>> for where, what in enumerate(L): ... print "[{0}] {1}".format(where, what) ... [0] Ministry [1] of [2] Silly [3] Walks If you would like the numbers to start at a different origin, pass that origin as the second argument to the enumerate() function. You will still get all the elements of the sequence, but the numbers will start at the value you provide. (Python 2.6 and later versions only.) >>> for where, what in enumerate(L, 1): ... print "[{0}] {1}".format(where, what) ... [1] Ministry [2] of [3] Silly [4] Walks

<code >file()</code >: Open a file This function is identical to the open() function; for details, see .

<code >filter()</code >: Extract qualifying elements from an iterable This function is useful for removing some of the elements of an iterable. You must provide a filtering function that takes one argument and returns a bool value. Here is the calling sequence: filter(f, S) The filtering function f is the first argument. It is applied to every element of some iterable S. The result is a new sequence containing only those elements x of S for which f(x) returned True. If f is a string or tuple, the result has the same type, otherwise the result is a list. If f is None, you get a sequence of the true elements of S. In this case, the filtering function is effectively the bool() function. >>> def isOdd(x): ... if (x%2) == 1: return True ... else: return False ... >>> filter(isOdd, [88, 43, 65, -11, 202]) [43, 65, -11] >>> filter(isOdd, (1, 2, 4, 6, 9, 3, 3)) (1, 9, 3, 3) >>> def isLetter(c): ... return c.isalpha() ... >>> filter(isLetter, "01234abcdeFGHIJ*(&!^") 'abcdeFGHIJ' >>> maybes = [0, 1, (), (2,), 0.0, 0.25] >>> filter(None, maybes) [1, (2,), 0.25] >>> filter(bool, maybes) [1, (2,), 0.25]

<code >float()</code >: Convert to <code >float</code > type Converts a value to type float. The argument must be a number, or a string containing a numeric value in string form (possibly surrounded by whitespace). If the argument is not a valid number, this function will raise a ValueError exception. If no argument is given, it will return 0.0. >>> float() 0.0 >>> float(17) 17.0 >>> float(' 3.1415 ') 3.1415000000000002 >>> print float('6.0221418e23') 6.0221418e+23 >>> float('142x') Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: invalid literal for float(): 142x

<code >format()</code >: Format a value This function converts a value to a formatted representation. The general form is: format(value[, spec]) For built-in types, the spec has the syntax as that described in between “:” and the closing “}”. >>> x = 34.56 >>> format(x, '9.4f') ' 34.5600' >>> '{0:9.4f}'.format(x) ' 34.5600' You can define how this function works for a user-defined class by defining the special method described in . If the spec argument is omitted, the result is the same as str(value).

<code >frozenset()</code >: Create a frozen set This function is used to create a new frozenset value: an immutable set. General form: frozenset(S) This function converts an existing iterable S to a frozenset. The argument is optional; if omitted, you get a frozen empty set. >>> frozenset() frozenset([]) >>> frozenset('aeiou') frozenset(['a', 'i', 'e', 'u', 'o']) >>> frozenset([0, 0, 0, 44, 0, 44, 18]) frozenset([0, 18, 44]) For more information, see .

<code >hex()</code >: Convert to base 16 Given an integer, this function returns a string displaying that value in hexadecimal (base 16). >>> hex(15) '0xf' >>> hex(255) '0xff' >>> hex(256) '0x100' >>> hex(1325178541275812780L) '0x1263fadcb8b713ac' See also : hexadecimal conversion is supported by specifying a type code of 'x' or 'X'.

<code >int()</code >: Convert to <code >int</code > type To convert a number of a different type to int type, or to convert a string of characters that represents a number: int(ns) where ns is the value to be converted. If ns is a float, the value will be truncated, discarding the fraction. If you want to convert a character string s, expressed in a radix (base) other than 10, to an int, use this form, where b is an integer in the range [2, 36] that specifies the radix. int(s, b) Examples: >>> int(43L) 43 >>> int(True) 1 >>> int(False) 0 >>> int(43.89) 43 >>> int("69") 69 >>> int('77', 8) 63 >>> int('7ff', 16) 2047 >>> int('10101', 2) 21

<code >input()</code >: Read an expression from the user This function asks the user to type something, then evaluates it as a Python expression. Here is the general form: input([prompt]) If you supply a string as the optional prompt argument, that string will be written to the user before the input is read. In any case, the result is the value of the expression. Of course, if the user types something that isn't a valid Python expression, the function will raise an exception. >>> input() 2+2 4 >>> print "The answer was '{0}'.".format(input()) 2+3*4 The answer was '14'. >>> print "The answer was '{0}'.".format(input("Type an expression:")) Type an expression:2+3*4 The answer was '14'. >>> print input() 1/0 Traceback (most recent call last): File "<stdin>", line 1, in <module> File "<string>", line 1, in <module> ZeroDivisionError: integer division or modulo by zero

<code >iter()</code >: Produce an iterator over a sequence Given a sequence S, this function returns an iterator that generates the elements of the sequence. For more information, see . >>> listWalker = iter ( [23, 47, 'hike'] ) >>> for x in listWalker: print x, ... 23 47 hike In general, the calling sequence is: iter(s[, sentinel]) If the sentinel argument is omitted, the first argument must be either a sequence value that implements the .__getitem__() method or an instance of a class that has the .__iter__() method. If you provide a sentinel argument, the s argument must be callable. The iterator returned will call s() with no arguments and generate the values it returns until the return value equals sentinel, at which point it will raise StopIteration.

<code >len()</code >: Number of elements Given a sequence or dictionary, this function returns the number of elements. >>> len('') 0 >>> len ( [23, 47, 'hike'] ) 3 >>> len({1: 'foot', 2:'shoulder', 'feather': 'rare'}) 3

<code >list()</code >: Convert to a list This function creates a new list. The argument x may be any iterable (see ). >>> list(('Bentham', 'Locke', 'Hobbes')) ['Bentham', 'Locke', 'Hobbes'] >>> list("Bruce") ['B', 'r', 'u', 'c', 'e'] >>> list((42,)) [42] >>> list() []

<code >long()</code >: Convert to <code >long</code > type This function works exactly the same way as , except that it produces a result of type long. >>> long(43) 43L >>> long(43.889) 43L >>> long('12345678901234567890123457890') 12345678901234567890123457890L >>> long('potrzebie456', 36) 3381314581245790842L

<code >map()</code >: Apply a function to each element of an iterable The purpose of this function is to perform some operation on each element of an iterable. It returns a list containing the result of those operations. Here is the general form: map(f, S) f is a function that takes one argument and returns a value. S is any iterable. >>> def add100(x): ... return x+100 ... >>> map(add100, (44,22,66)) [144, 122, 166] To apply a function with multiple arguments to a set of sequences, just provide multiple iterables as arguments, like this. >>> def abc(a, b, c): ... return a*10000 + b*100 + c ... >>> map(abc, (1, 2, 3), (4, 5, 6), (7, 8, 9)) [10407, 20508, 30609] If you pass None as the first argument, Python uses the identity function to build the resulting list. This is useful if you want to build a list of tuples containing items from two or more iterables. >>> map(None, range(3)) [0, 1, 2] >>> map(None, range(3), 'abc', [44, 55, 66]) [(0, 'a', 44), (1, 'b', 55), (2, 'c', 66)]

<code >max()</code >: Largest element of an iterable Given an iterable S that contains at least one element, max(S) returns the largest element of the sequence. >>> max('blimey') 'y' >>> max ( [-505, -575, -144, -288] ) -144 >>> max([]) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: max() arg is an empty sequence You can also pass multiple arguments, and max() will return the largest. In the example below, 'cheddar' is the largest because lowercase letters have higher codes than uppercase letters. >>> max('Gumby', 'Lambert', 'Sartre', 'cheddar') 'cheddar' If you want to redefine the comparison function, you may provide a keyword argument key=f, where f is a function that takes one argument and returns a value suitable for comparisons. In this example, we use the .upper() method of the str class to compare the uppercased strings, then return the original string whose uppercased value is largest. >>> max('Gumby', 'Lambert', 'Sartre', 'cheddar', key=str.upper) 'Sartre'

<code >min()</code >: Smallest element of an iterable Given an iterable S containing at least one element, max(S) returns the largest element of S. >>> min('blimey') 'b' >>> min ( [-505, -575, -144, -288] ) -575 You may also pass multiple arguments, and the min() function will return the smallest. >>> min(-505, -575, -144, -288) -575 If you would like to use a different function to define the ordering, specify that function as a keyword argument key=f, where f is a function that takes one argument and returns a value suitable for comparisons. In this example, we want to order the values based on their inverse. >>> def rev(x): ... return -x ... >>> min(-505, -575, -144, -288, key=rev) -144

<code >next()</code >: Call an iterator This function attempts to get the next value from some iterator I (see ). (New in version 2.6.) next(I[, default) If the iterator produces another value, that value is returned by this function. If the iterator is exhausted and you provide a default value, that value is returned. If the iterator is exhausted and you do not provide a default value, the next() function raises a StopIteration exception. Here is an example. >>> it = iter(xrange(0,2)) >>> next(it, 'Done') 0 >>> next(it, 'Done') 1 >>> next(it, 'Done') 'Done' >>> next(it) Traceback (most recent call last): File "<stdin>", line 1, in <module> StopIteration

<code >oct()</code >: Convert to base 8 Given a number n, oct(n) returns a string containing an octal (base 8) representation of n. Consistent with Python's convention that any number starting with a zero is considered octal, the result of this function will always have a leading zero. >>> oct(0) '0' >>> oct(127) '0177' See also : octal conversion is supported by specifying type code 'o'.

<code >open()</code >: Open a file Open a file. For the calling sequence, see .

<code >ord()</code >: Find the numeric code for a character Given a string s containing a single character, ord(s) returns that character's numeric code. Compare . >>> ord('A') 65 >>> ord('\x00') 0 >>> ord(u'\u262e') 9774 >>> hex(9774) '0x262e'

<code >pow()</code >: Exponentiation There are two ways to compute xy in Python. You can write it as “x**y”. There is also a function that does the same thing: pow(x, y) For integer arithmetic, the function also has a three-argument form that computes xy%z, but more efficiently than if you used that expression: pow(x, y, z) Examples: >>> 2**4 16 >>> pow(2,4) 16 >>> pow(2.5, 4.5) 61.763235550163657 >>> (2**9)%3 2 >>> pow(2,9,3) 2

<code >range()</code >: Generate an arithmetic progression as a list This function generates a list containing the values of an arithmetic progression, that is, a sequence of numbers such that the difference between adjacent numbers is always the same. There are three forms: range(n) Returns the list [0, 1, 2, ..., n-1]. Note that the result never includes the value n.

range(start,
            stop)

Returns the list [start, start+1, start+2, ..., stop-1]. The result never includes the stop value.

range(start,
            stop, step)

Returns the list [start, start+step, start+2*step, ...], up to but not including the value of stop. The value of step may be negative. Examples: >>> range(4) [0, 1, 2, 3] >>> range(4,9) [4, 5, 6, 7, 8] >>> range(10,104,10) [10, 20, 30, 40, 50, 60, 70, 80, 90, 100] >>> range(5,-1,-1) [5, 4, 3, 2, 1, 0]

<code >raw_input()</code >: Prompt and read a string from the user To prompt for keyboard entry, use this function: raw_input(p) The argument p is a prompt string that is written to standard output. Then a line is read from standard input and returned as a string, without its trailing newline character. >>> party = raw_input('Party affiliation: ') Party affiliation: Slightly silly. >>> party 'Slightly silly.' If the user signals end of input (e.g., with Control-D under Unix), the function raises an EOFError exception.

<code >reduce()</code >: Sequence reduction The idea of reduction comes from the world of functional programming. There is a good introductory article on this concept in Wikipedia. In simple terms, a function of two arguments is applied repeatedly to the elements of an iterable to build up a final value. The idea of a “sum of elements of a sequence” is a reduction of those elements using “+” as the function. For example, the +-reduction of [2, 3, 5] is 2+3+5 or 10. Similarly, the product of a series of numbers is a reduction using the “*” operator: the multiply reduction of [2, 3, 5] is 2*3*5 or 30. There are two general forms: reduce(f, S) reduce(f, S, I) f is a function that takes two arguments and returns a value. S is an iterable. The result depends on the number of elements in S, and whether the initial value I is supplied. Let's look first at the case where argument I is not supplied. If S has only one element, the result is S[0]. If S has two elements, the result is

f(S[0],
            S[1])

. If S has three elements, the result is f(f(S[0], S[1]), S[2]). If S has four or more elements, f is applied first to S[0] and S[1], then to that result and S[2], and so on until all elements are reduced to a single value. If S is empty and no initial value was provided, the function raises a TypeError exception. If an initial value I is provided, the result is the same as reduce(f, [I]+list(S)). Some examples: >>> def x100y(x,y): ... return x*100+y ... >>> reduce(x100y, [15]) 15 >>> reduce(x100y, [1,2]) 102 >>> reduce(x100y, [1,2,3]) 10203 >>> reduce(x100y, (), 44) 44 >>> reduce(x100y, [1], 2) 201 >>> reduce(x100y, [1,2], 3) 30102 >>> reduce(x100y, [1,2,3], 4) 4010203 >>> reduce(x100y, []) Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: reduce() of empty sequence with no initial value

<code >reversed()</code >: Produce a reverse iterator This function, applied to a sequence S, returns an iterator that generates the elements of S in reverse order (see ). >>> L=[22,44,88] >>> backL = reversed(L) >>> for i in backL: ... print i, ... 88 44 22 The allowable values for S include all the types described in . It also works for any class provided one of these two conditions is met: The method described in . If the class has no .__reversed__() method, reversed() will still work provided that instances act like a sequence; that is, the class has a .__len__() method and a .__getitem__() method.

<code >round()</code >: Round to the nearest integral value To find the integral value nearest to some value x, use this function: round(x) The value is returned as a float. In the case that the fractional part of x is exactly one-half, the returned value is the integer farther from zero. Examples: >>> round ( 4.1 ) 4.0 >>> round(4.9) 5.0 >>> round(4.5) 5.0 >>> round(-4.1) -4.0 >>> round(-4.9) -5.0 >>> round(-4.5) -5.0 You can also provide an optional second argument that specifies how many digits to retain after the decimal. >>> from math import pi >>> round(pi) 3.0 >>> print round(pi,1) 3.1 >>> print round(pi,2) 3.14 >>> print round(pi, 4) 3.1416 >>> round(pi,30) 3.1415926535897931

<code >set()</code >: Create an algebraic set Use this function to create a set value. Here is the general form: set(S) The optional argument S is any iterable; the return value is a new set instance containing the unique values from S. When called with no arguments, this function returns a new, empty set. Examples: >>> empty = set() >>> empty set([]) >>> len(empty) 0 >>> swallows=set(['African', 'European']) >>> swallows set(['European', 'African']) >>> len(swallows) 2 >>> set ( (0, 0, 0, 58, 0, 0, 58, 17) ) set([0, 17, 58]) For more information about sets, see .

<code >sorted()</code >: Sort a sequence This function, applied to any iterable S, produces a new list containing the elements of S in ascending order (or some other order you specify). Here is the general form: sorted(S[, cmp[, key[, reverse]]]) The cmp, key, and reverse arguments are optional, and have the same meaning as in the .sort() method of the list type (see ). >>> L = ['geas', 'clue', 'Zoe', 'Ann'] >>> sorted(L) ['Ann', 'Zoe', 'clue', 'geas'] >>> def ignoreCase(x,y): ... return cmp(x.upper(), y.upper()) ... >>> sorted(L, ignoreCase) ['Ann', 'clue', 'geas', 'Zoe'] >>> sorted(L, None, str.upper) ['Ann', 'clue', 'geas', 'Zoe'] >>> L ['geas', 'clue', 'Zoe', 'Ann'] In the first example above, 'Zoe' precedes 'clue', because all uppercase letters are considered to be less than all lowercase letters. The second example shows the use of a cmp argument to sort strings as if they were all uppercase; the third example shows how to achieve the same result using the .upper() method of the str class as the key argument. Note in the last line that the original list L is unchanged.

<code >str()</code >: Convert to <code >str</code > type To convert any value x to a string, use this general form: str(x) For example: >>> str(17) '17' >>> str({'boy': 'Relmond', 'girl': 'Wirdley'}) "{'boy': 'Relmond', 'girl': 'Wirdley'}" For general information, see .

<code >sum()</code >: Total the elements of a sequence This function, applied to an iterable S, returns the sum of its elements. There are two general forms: sum(S) sum(S, I) In the second form, the summing process starts with the initial value I. Examples: >>> L=[1,2,3,4] >>> sum(L) 10 >>> sum(L,1000) 1010 >>> sum((), 1000) 1000

<code >tuple()</code >: Convert to a tuple To convert some iterable S to a tuple, use this general form: tuple(s) The result will be a new tuple with the same elements as S in the same order. For general information, see . To create an empty tuple, omit the argument. Examples: >>> tuple() () >>> tuple ( ['swallow', 'coconut'] ) ('swallow', 'coconut') >>> tuple ( 'shrubbery' ) ('s', 'h', 'r', 'u', 'b', 'b', 'e', 'r', 'y') >>> tuple ( ['singleton'] ) ('singleton',)

<code >type()</code >: Return a value's type This function can be applied to any value. It returns a type object corresponding to the type of that value. For built-in types, the type object is the same as the name of the type: int, str, list, and so on. To test whether a value x is some type T, you can use the predicate “

type(x)
        is T

”. If you display a type object in conversational mode, it will look like “<type 'T'>”. Examples: >>> type(i) <type 'int'> >>> type(i) is int True >>> type([2,4,8]) is list True

<code >unichr()</code >: Convert a numeric code to a Unicode character Given a number n, this function returns the Unicode character that has code point n. For more on Unicode, see . >>> unichr(0) u'\x00' >>> unichr(ord('A')) u'A' >>> unichr(0x3046) u'\u3046' >>> unichr(0xe0047) u'\U000e0047'

<code >unicode()</code >: Convert to a Unicode string Use this function to convert a string to type unicode. For more information about Unicode, see . >>> unicode('Pratt') u'Pratt' >>> unicode() u''

<code >xrange()</code >: Arithmetic progression generator The xrange() function has exactly the same arguments as the range() function (see ). The difference is that xrange() is a generator (see ), while range() actually builds a list for its result. This means you can use xrange() in situations where you want to generate a large series of the values from an arithmetic progression, but you don't have enough memory to build that series as a list. >>> for i in xrange(2000000000): ... print i, ... if i > 8: ... break ... 0 1 2 3 4 5 6 7 8 9 >>> for i in range(2000000000): ... print i, ... if i > 8: ... break ... Traceback (most recent call last): File "<stdin>", line 1, in <module> MemoryError

<code >zip()</code >: Combine multiple sequences The purpose of this function is to build a list of tuples from two or more iterables of the same length. Here is the general form: zip(S0, S1, S2, ...) Each Si must be in iterable. The result is a list [T0, T1, ...], where each Ti is the tuple

(S0[i],
        S1[i],
        S2[i],
        ...)

. Here are some examples. >>> L1=[1,2,3,4] >>> L2=['a', 'b', 'c', 'd'] >>> zip(L1, L2) [(1, 'a'), (2, 'b'), (3, 'c'), (4, 'd')] >>> L3=[10.0, 20.0, 30.0, 40.0] >>> zip(L1, L2, L3) [(1, 'a', 10.0), (2, 'b', 20.0), (3, 'c', 30.0), (4, 'd', 40.0)]

Advanced functions This section describes Python functions that most people will never need. However, the advanced Python programmer may find some of them quite useful.

<code >basestring</code >: The string base class The class basestring is the parent class for the two string types, str and unicode. It exists not because you can call it (you can't), but for type testing. To test whether some value s is an instance of either type of string, use this predicate: isinstance(s, basestring) See .

<code >callable()</code >: Is this thing callable? This predicate tests whether some value x can be called as a function. callable(x) Class names can be called to create an instance of the class. Instances can be called if they define a .__call__() special method; see . >>> def someFunction(): ... pass ... >>> callable(someFunction) True >>> callable(len) True >>> callable(int) True >>> callable(42) False

<code >classmethod()</code >: Create a class method The purpose of the classmethod() function is to convert a method into a class method. For a discussion of the purpose and usage of class methods, see . There are two ways to declare a class method within a class: You can use the function decorator syntax to declare that classmethod is a decorator for your method. Precede the method definition with a line reading: @classmethod def methodName(cls, ...): method body In some older versions of Python without the decorator syntax, you can still declare a class method by placing a line after the method definition, at the same indentation level as the method's def statement, having this form: methodName = classmethod(methodName)

<code >delattr()</code >: Delete a named attribute Use this function to delete an attribute named A of some instance I. It does not return a value. Here is the general form: delattr(I, A) For example, if an instance seabiscuit has a rider attribute, this statement would delete that attribute: delattr(seabiscuit, 'rider') If the instance has no such attribute, this function will raise an AttributeError exception.

<code >dir()</code >: Display a namespace's names The purpose of the dir() function is to find out what names are defined in a given namespace, and return a list of those names. If called without arguments, it returns a list of the names defined in the local namespace. This function can be very handy for finding out what items are in a module or class. Certain special names are found in most or all namespaces: __doc__ is present in every namespace. Initially None, you can store documentation there. __name__ is the name of the current module (minus the .py extension). In the top-level script or in conversational mode, this name is set to the string '__main__'. __builtins__ is a list of the names of all built-in functions and variables. In this example sequence, we'll show you what is in the global namespace just after starting up Python. Then we'll import the math module and display its names. >>> dir() ['__builtins__', '__doc__', '__name__'] >>> x=5; forkTail='Tyrannus' >>> dir() ['__builtins__', '__doc__', '__name__', 'forkTail', 'x'] >>> print __doc__ None >>> print __name__ __main__ >>> import math >>> print math.__name__ math >>> dir(math) ['__doc__', '__file__', '__name__', 'acos', 'asin', 'atan', 'atan2', 'ceil' , 'cos', 'cosh', 'degrees', 'e', 'exp', 'fabs', 'floor', 'fmod', 'frexp', ' hypot', 'ldexp', 'log', 'log10', 'modf', 'pi', 'pow', 'radians', 'sin', 'si nh', 'sqrt', 'tan', 'tanh'] >>> print math.__doc__ This module is always available. It provides access to the mathematical functions defined by the C standard. >>> print math.log10.__doc__ log10(x) -> the base 10 logarithm of x. >>> print __builtins__ <module '__builtin__' (built-in)> >>> for k, name in enumerate(dir(__builtins__)): ... if (k%4)==3: print ... print name, ... ArithmeticError AssertionError AttributeError BaseException BufferError BytesWarning DeprecationWarning EOFError Ellipsis EnvironmentError Exception False FloatingPointError FutureWarning GeneratorExit IOError ImportError ImportWarning IndentationError IndexError KeyError KeyboardInterrupt LookupError MemoryError NameError None NotImplemented NotImplementedError OSError OverflowError PendingDeprecationWarning ReferenceError RuntimeError RuntimeWarning StandardError StopIteration SyntaxError SyntaxWarning SystemError SystemExit TabError True TypeError UnboundLocalError UnicodeDecodeError UnicodeEncodeError UnicodeError UnicodeTranslateError UnicodeWarning UserWarning ValueError Warning ZeroDivisionError _ __debug__ __doc__ __import__ __name__ __package__ abs all any apply basestring bin bool buffer bytearray bytes callable chr classmethod cmp coerce compile complex copyright credits delattr dict dir divmod enumerate eval execfile exit file filter float format frozenset getattr globals hasattr hash help hex id input int intern isinstance issubclass iter len license list locals long map max min next object oct open ord pow print property quit range raw_input reduce reload repr reversed round set setattr slice sorted staticmethod str sum super tuple type unichr unicode vars xrange zip

<code >eval()</code >: Evaluate an expression in source form This function evaluates a Python expression from a string. Example: >>> cent=100 >>> eval('cent**3') 1000000 If you want to evaluate the expression using different name environments, refer to the official documentation. For related features, see also and .

<code >execfile()</code >: Execute a Python source file To execute a sequence of Python statements in some file F, use this function: execfile(F) The function returns None. For additional features that allow you to control the environment of the executed statements, see the official documentation.

<code >getattr()</code >: Retrieve an attribute of a given name Use this function to retrieve an attribute of an instance I, where the attribute's name is a string s. getattr(I, s[, default]) If I has no attribute whose name matches s: If you supplied the optional default value, that value is returned. If you don't supply a default value and there is no such attribute in I, this function will raise an AttributeError exception. Example: >>> class C: ... def __init__(self, flavor): ... self.flavor = flavor ... >>> c=C('garlicky') >>> getattr(c, 'flavor') 'garlicky' >>> getattr(c, 'aroma', 'bland') 'bland' >>> getattr(c, 'aroma') Traceback (most recent call last): File "<stdin>", line 1, in <module> AttributeError: C instance has no attribute 'aroma'

<code >globals()</code >: Dictionary of global name bindings This function takes no arguments and returns a dictionary that represents the current global namespace. The keys of this dictionary are globally defined names, and each corresponding value is the value for that name. This example starts from a fresh execution of Python in conversational mode, so the global namespace has only the three special names discussed in . >>> globals() {'__builtins__': <module '__builtin__' (built-in)>, '__name__': '__main_ _', '__doc__': None} >>> finch = 'Fleep' >>> globals() {'__builtins__': <module '__builtin__' (built-in)>, '__name__': '__main_ _', '__doc__': None, 'finch': 'Fleep'} The special name __builtins__ is bound to a module; name __name__ is bound to the string '__main__'; and __doc__ is bound to None. Note that defining a new name adds an entry to the result of globals().

<code >hasattr()</code >: Does a value have an attribute of a given name? This predicate tests to see if some instance I has an attribute whose name is given by some string s: hasattr(I, s) If this function returns True, you can be sure that the instance has an attribute named s. However, if it returns False, attempts to access an attribute may still succeed, if the class provides dynamic attributes; see . Example: >>> class C: ... def __init__(self, disc): ... self.disk = disc ... >>> c=C('five') >>> hasattr(c, 'disk') True >>> hasattr(c, 'disc') False >>> hasattr(c, 'jukebox') False >>> c.jukebox = 'Nine' >>> hasattr(c, 'jukebox') True

<code >id()</code >: Unique identifier This function, given any Python value, returns an integer value that uniquely identifies it. In most implementations, it is the value's physical memory address. >>> i = 20 >>> id(i) 137727456

<code >isinstance()</code >: Is a value an instance of some class or type? Use this predicate to test whether some instance I is an instance of some class C (or an instance of any ancestor class from which C inherits). The general form: isinstance(I, C) The second argument may be a class name, a type object, or a tuple of class names and type objects. If a tuple, the function will test the instance against each of the class names or type objects. >>> class C1: ... pass ... >>> class C2(C1): ... pass ... >>> c2=C2() >>> isinstance(c2,C2) True >>> isinstance(c2,C1) True >>> isinstance(c2,int) False >>> isinstance(1,type(55)) True >>> isinstance(1, (int, float, long)) True >>> isinstance('Ni', (int, float, long)) False A most useful built-in Python class is basestring, which is the ancestor class of both str and unicode types. It is intended for cases where you want to test whether something is a string but you don't care whether it is str or unicode. >>> isinstance(42, str) False >>> isinstance('x', str) True >>> isinstance(u'x', str) False >>> isinstance('x', basestring) True >>> isinstance(u'x', basestring) True

<code >issubclass()</code >: Is a class a subclass of some other class? To test whether some class C1 is a subclass of another class C2, use this predicate: issubclass(C1, C2) Examples: >>> class Polygon: ... pass ... >>> class Square(Polygon): ... pass ... >>> issubclass(Square, Polygon) True >>> issubclass(Polygon, Square) False >>> issubclass(Square, Square) True >>> issubclass(unicode, basestring) True >>> issubclass(str, basestring) True For more information about the built-in basestring class, see .

<code >locals()</code >: Dictionary of local name bindings This function returns a dictionary containing the names and values of all variables in the local namespace. An example: >>> def f(a, b=1): ... c=2 ... print locals() ... >>> f(5) {'a': 5, 'c': 2, 'b': 1} For related functions, see and .

<code >property()</code >: Create an access-controlled attribute The purpose of this function is to create a property of a class. A property looks and acts like an ordinary attribute, except that you provide methods that control access to the attribute. There are three kinds of attribute access: read, write, and delete. When you create a property, you can provide any or all of three methods that handle requests to read, write, or delete that attribute. Here is the general method for adding a property named p to a new-style class C. class C(...): def R(self): ...read method... def W(self, value): ...write method... def D(self): ...delete method... p = property(R, W, D, doc) ... where: R is a getter method that takes no arguments and returns the effective attribute value. If omitted, any attempt to read that attribute will raise AttributeError. W is a setter method that takes one argument and sets the attribute to that argument's value. If omitted, any attempt to write that attribute will raise AttributeError. D is a deleter method that deletes the attribute. If omitted, any attempt to delete that attribute will raise AttributeError. doc is a documentation string that describes the attribute. If omitted, defaults to the documentation string of the R method if any, otherwise None. To retrieve a property's documentation, use this form: C.p.__doc__ where C is the class name and p is the property name. As an example, here is a small class that defines a property named x: class C(object): def __init__(self): self.__x=None def getx(self): print "+++ getx()" return self.__x def setx(self, v): print "+++ setx({0})".format(v) self.__x = v def delx(self): print "+++ delx()" del self.__x x=property(getx, setx, delx, "Me property 'x'.") Assuming that class is defined, here is a conversational example. >>> c=C() >>> print c.x +++ getx() None >>> print C.x.__doc__ Me property 'x'. >>> c.x=15 +++ setx(15) >>> c.x +++ getx() 15 >>> del c.x +++ delx() >>> c.x +++ getx() Traceback (most recent call last): File "<stdin>", line 1, in <module> File "<stdin>", line 6, in getx AttributeError: 'C' object has no attribute '_C__x' Starting with Python 2.6, this function can also be used as a decorator (see ). The decorated method is used as the getter method. Furthermore, the decorated method will itself have two decorators named setter and deleter; you can use these decorators to define setter and deleter methods. For example, suppose you want to provide your class with a property named state, and youra getter method returns a private attribute named ._state. You could define it like this: @property def state(self): '''The internal state property.''' return self._state In this example, not only will the .state() method be the getter for this property, but the documentation string

'''The internal state
        property.'''

will be stored as the documentation string for the property. Suppose further that you want to write a setter method that checks to make sure the argument is a positive number less than or equal to 2. To use the built-in setter method to write your setter, give the function the same name as the property, and decorate it with P.setter where P is the name of the previously defined getter: @state.setter def state(self, k): if not (0 <= k <= 2): raise ValueError("Must be 0 through 2 inclusive!") else: self._state = k Similarly, you can write a deleter method by decorating it with P.deleter: @state.deleter def state(self): del self._state

<code >reload()</code >: Reload a module This function reloads a previously loaded module (assuming you loaded it with the syntax “import moduleName”. It is intended for conversational use, where you have edited the source file for a module and want to test it without leaving Python and starting it again. General form: reload(moduleName) The moduleName is the actual name of the module, not a string containing its name. For example, if you have imported a module like this: import parrot you would say “reload(parrot)”, not “reload('parrot')”.

<code >repr()</code >: Representation The repr function attempts to convert any value to a string. Unlike the str() function, it attempts to display the value in Python source form, that is, in a form suitable for passing to eval() (see ). It works the same as the `...` operator. Examples: >>> s='Wensleydale' >>> print s Wensleydale >>> print str(s) Wensleydale >>> print repr(s) 'Wensleydale' >>> print `s` 'Wensleydale' To specify the behavior of the repr() when it is applied to an instance of a user-defined class, see .

<code >setattr()</code >: Set an attribute This function is the inverse of : it sets the value of some attribute whose name is A from an instance I to a new value V: setattr(I, A, V) Example: >>> class C5: ... def __init__(self, x): ... self.x = x ... >>> c=C5(14) >>> c.x 14 >>> setattr(c, 'x', 19) >>> c.x 19 >>> setattr(c, 'violateEncapsulation', True) >>> c.violateEncapsulation True As the last lines above show, you can use this function to create attributes that didn't even exist before. However, this is often considered bad style, as it violates the principle of encapsulation.

<code >slice()</code >: Create a slice instance This function creates an instance of type slice. A slice instance I can be used to index a sequence S in an expression of the form S[I]. Here is the general form: slice(start, limit, step) The result is a slice that is equivalent to start:limit:step. Use None to get the default value for any of the three arguments. Examples: >>> r = range(9) >>> r [0, 1, 2, 3, 4, 5, 6, 7, 8] >>> r[::2] [0, 2, 4, 6, 8] >>> r[slice(None, None, 2)] [0, 2, 4, 6, 8] >>> r[3:7] [3, 4, 5, 6] >>> r[slice(3,7)] [3, 4, 5, 6] >>> r[1::2] [1, 3, 5, 7] >>> odds = slice(1, None, 2) >>> r[odds] [1, 3, 5, 7] >>> 'roygbiv'[odds] 'ogi'

<code >staticmethod()</code >: Create a static method The purpose of the staticmethod function is to convert a method into a static method. See for definitions and usage. Typically you will declare a static method using the decorator syntax, like this: @staticmethod def methodName(...): method body An alternative is to place a line like this after the method's definition (at the same indentation level as its def): methodName = staticmethod(methodName)

<code >super()</code >: Superclass The purpose of this function is to retrieve the superclass of a given type or object. Superclasses are beyond the scope of this document; see the online documentation for built-in functions (scroll down).

<code >vars()</code >: Local variables This function returns a dictionary that represents a symbol table: its keys are variable names, and each key's corresponding value is its bound value. The official documentation warns you not to change anything in this dictionary, or Bad Things Will Happen. There are two forms of call: vars() Returns the local symbol table. vars(ns) Returns the symbol table of a namespace ns, where a namespace can be a module, an instance, or a class. Compare the similar functions , , and .

Simple statements Python statement types are divided into two groups. Simple statements, that are executed sequentially and do not affect the flow of control, are described first. Compound statements, which may affect the sequence of execution, are discussed in . Here, for your convenience, is a table of all the Python statement types, and the sections where they are described. The first one, the assignment statement, does not have an initial keyword: an assignment statement is a statement of the form “variable = expression”. Assignment . assert . break . continue . del . elif and . else . except . exec . finally . for . from . global . if . import . pass . print . raise . return . try . yield .

The assignment statement: <code ><replaceable >name</replaceable > = <replaceable >expression</replaceable ></code > The purpose of Python's assignment statement is to associate names with values in your program. It is the only statement that does not start with a keyword. An assignment statement is a line containing at least one single equal sign (=) that is not inside parentheses. Here is the general form of an assignment statement: target0 = target1 = ... = expression In most cases, there will be one target that is a name. Python will evaluate the expression, reducing it to a single value, and then bind that name to the that value. A binding is an association between a name and a value. It is important to note that in Python, unlike many other languages, names themselves are not associated with a specific type. A name is just a label, and it can be bound to any value of any type at any time. In this example, name x is bound first to an int value 5, then to a str value 'Some string'. >>> x = 5 >>> x 5 >>> x = 'Some string' >>> print x Some string If a target name was already bound to a value, the name is unbound from that value before it is rebound to the new value. For each value in a running program, Python keeps track of how many names are bound to that value. When the value has no more names bound to it, the value's memory is automatically recycled. If the value is an instance of a class, its destructor may be called; see . There are several other forms of assignment statement. n0 = n1 = ... = expression If you supply multiple target names, each target will be assigned the value of the expression. Example: >>> i = j = errorCount = 0 >>> i 0 >>> j 0 >>> errorCount 0 n0, n1, ... = expression If the target is a comma-separated list of names, the expression must evaluate to an iterable with the same number of values as there are names. Each value in the expression is then bound to the corresponding name. Example: >>> L = ["Bell's Blue", "male", 6] >>> name, sex, age = L >>> name "Bell's Blue" >>> sex 'male' >>> age 6 This feature, called “unpacking,” generalizes to arbritrarily nested sequences within sequences. You may group targets inside parentheses (...) or brackets [...] to show the levels of nesting. Here is an example: >>> s = [1, [2, 3, [4, 5], 6], 7] >>> a, (b, c, [d, e], f), g = s >>> print a,b,c,d,e,f,g 1 2 3 4 5 6 7 All the assignments are effectively simultaneous. Therefore, you can safely exchange the values of two variables using a statement like this: v1, v2 = v2, v1 Examples: >>> a=5; b=9998 >>> print a,b 5 9998 >>> a,b=b,a >>> print a,b 9998 5 >>> c=432 >>> a,b,c = b,c,a >>> print a,b,c 5 432 9998 name[i] = expression If name is an iterable, the expression i must evaluate to an integer. The element after position i is replaced by the value of the expression. >>> L = range(6) >>> L [0, 1, 2, 3, 4, 5] >>> L[2] 2 >>> L[2] = 888 >>> L [0, 1, 888, 3, 4, 5] If name is a dictionary (or other mapping), and name does not have a key-value pair whose key equals index, a new key-value pair is added to name with key i and value expression. >>> d={'pudding': 'figgy'} >>> d {'pudding': 'figgy'} >>> d['tart'] = 'strawberry' >>> d {'pudding': 'figgy', 'tart': 'strawberry'} >>> d["tart"] = "rat" >>> d {'pudding': 'figgy', 'tart': 'rat'} As the last two lines show, if the dictionary already has a key-value pair for key i, the old value of that pair is replaced by the expression value.

name[start:end] =
            S

If name is a list or other mutable sequence, you can replace the elements of a slice of that sequence with the elements from some sequence S. (For an explanation of slicing, see .) This may result in addition, deletion, or replacement of the elements of name. Some examples will give the flavor of this kind of assignment. >>> L=range(6) >>> L [0, 1, 2, 3, 4, 5] >>> L[2:4] [2, 3] >>> L[2:4] = [111, 222, 333, 444, 555] >>> L [0, 1, 111, 222, 333, 444, 555, 4, 5] >>> L[3] 222 >>> L[3:3] [] >>> L[3:3] = [41.0, 42.0, 43.0] >>> L [0, 1, 111, 41.0, 42.0, 43.0, 222, 333, 444, 555, 4, 5] >>> L[4:7] [42.0, 43.0, 222] >>> L[4:7] = () >>> L [0, 1, 111, 41.0, 333, 444, 555, 4, 5] The “=” signs in an assignment is not an operator, as it is in some other languages. You cannot assign a value to a name inside an expression; an assignment statement must stand alone. >>> a = 5 + (a=7) File "<stdin>", line 1 a = 5 + (a=7) ^ SyntaxError: invalid syntax Python also supports augmented assignment. In this form, you may place certain operators before the “=”. Here is the general form: name operator= expression An assignment of this general form has the same semantics as this form: name = name operator expression Supported operator symbols include: + - * / % ** >> << & ^ | Examples: >>> i = 1 >>> i += 3 >>> i 4 >>> i *= 5 >>> i 20

The <code >assert</code > statement: Verify preconditions To check for “shouldn't happen” errors, you can use an assert statement: assert e1 assert e1, e2 where e1 is some condition that should be true. If the condition is false, Python raises an AssertionError exception (see ). If a second expression e2 is provided, the value of that expression is passed with the exception. Assertion checking can be disabled by running Python with the -O (optimize) option.

The <code >del</code > statement: Delete a name or part of a value The purpose of the del statement is to delete things. The general form is: del L0, L1, ... where each Li is an item to be deleted. You can delete: A name. For example, the statement del i, j causes names i and j to become unbound, that is, undefined. An element or slice from a list. For example: del L[5], M[-2:] would delete the sixth element of list L and the last two elements of list M. One entry in a dictionary. For example, if D is a dictionary, del D['color'] would delete from D the entry for key 'color'.

The <code >exec</code > statement: Execute Python source code To dynamically execute Python code, use a statement of this form: exec E0 [in E1 [, E2]] Expression E0 specifies what to execute, and may be a string containing Python source code, an open file, or a code object. If E1 is omitted, the code is executed in the local scope. If E1 is given but E2 is not, E1 is a dictionary used to define the names in the global and local scopes. If E2 is given, E1 is a dictionary defining the global scope, and E2 is a dictionary defining the local scope.

The <code >global</code > statement: Declare access to a global name The purpose of the global statement is to declare that a function or method intends to change the value of a name from the global scope, that is, a name from outside the function. When Python reads the definition of a function, it checks each name to see if that name's value may possibly be changed anywhere in the function—that is, if the name shows up on the left side of an assignment statement, or as the induction variable in a for loop, or in any other context where the name's value can be changed. Such names are assumed to be local to the function unless you override this behavior by declaring that name in a global statement. Here is the general form: global name1, name2, ... Some conversational examples may help make this clear. Suppose you define a global variable x; you can use that name inside a function. >>> x = 5 >>> def show1(): ... print x ... >>> show1() 5 However, if you assign a value to x inside the function, the name x is now local to the function. It is said to shadow the global variable with the same name, and any changes to the value associated with that name inside the function will operate on a local copy, and will not affect the value of the global variable x. >>> x = 5 >>> def show2(): ... x = 42 ... print x ... >>> show2() 42 >>> x 5 But if you actually do want to change the value of the global variable inside the function, just declare it global like this: >>> x = 5 >>> def show3(): ... global x ... x = 42 ... print x ... >>> show3() 42 >>> x 42 Notice what happens in this case: >>> x = 5 >>> def show4(): ... print x, "Before" ... x = 42 ... print x, "After" ... >>> show4() Traceback (most recent call last): File "<stdin>", line 1, in <module> File "<stdin>", line 2, in show4 UnboundLocalError: local variable 'x' referenced before assignment Because the line “x = 42” changes the value of x, and because it is not declared as a global, execution fails because the value of the local variable x is used before that variable has had a value assigned to it.

The <code >import</code > statement: Use a module One of the cornerstones of Python's design philosophy is to keep the language relatively small and well-defined, and move all non-essential functionality to library modules. The import and from statements allow your programs to use items from these library modules. Your Python installation will come with a large collection of released modules. You can also create your own modules. Just place Python statements defining variables, functions, and classes into a file whose name ends in “.py”. There are two different statements you can use to import items from a module: The from statement copies items from a module into your namespace. After importing an item in this way, you can refer to the item simply by its name. General forms: from moduleName import * from moduleName import name1, name2, ... The first form imports all the items from the module named moduleName. If you want to import only specific items, use the second form, and enumerate the names you want from that module. The import statement makes an entire module's content available to you as a separate namespace. To refer to some item named N in a module named M, use the dot notation, M.N. Here is the general form: import moduleName, ... If you want to use some module M in this way, but you want to change the name to some different name A, use this form: import M as A Here are some examples that use the standard math module that is always available in a proper Python install. This module has functions such as sqrt() (square root), as well as variables such as pi. (Although π is a constant in the mathematical sense, the name pi is a variable in the Python sense.) >>> from math import * >>> sqrt(16) 4.0 >>> pi 3.1415926535897931 If you wanted only the sqrt function and the variable pi, this statement would do the job: from math import sqrt, pi Now some examples of the second form. >>> import math >>> math.sqrt(25) 5.0 >>> math.pi 3.1415926535897931 Suppose your program already used the name math for something else, but you still want to use functions from the math module. You can import it under a different name like this: >>> import math as crunch >>> crunch.sqrt(25) 5.0 >>> crunch.pi 3.1415926535897931

The <code >pass</code > statement: Do nothing Python's pass statement is a placeholder. It does nothing. Here's an example. Suppose you have a function named arr() that does something and then returns a True or False value. You want to keep calling this function until it returns a false value. This code would suffice: while arr(): pass

The <code >print</code > statement: Display output values Use this statement to display values on the standard output stream. General form: print thing1, thing2, ...[,] Each thing must be a string, or a value that can be converted into a string by the str() function (see ). These strings are written to the standard output stream, with one space between each value. A print statement by itself prints an empty line. >>> print 4.3, 'Sir Robin', 1./7 4.3 Sir Robin 0.142857142857 >>> for i in range(4): ... print i**4, ... 0 1 16 81 >>> Normally, a newline is printed after the last value. However, you can suppress this behavior by appending a comma to the end of the list. For example, this statement: print 'State your name:', would print the string followed by one space and leave the cursor at the end of that line.

The <code >print()</code > function In Python 3, print is a function, not a statement. To make it easier to convert your programs to the new syntax, first use this import statement (introduced in Python 2.6): from __future__ import print_function Here is the interface to this function: print(*args, sep=' ', end='\n', file=None) args One or more positional arguments whose values are to be printed. sep By default, consecutive values are separated by one space. You may specify a different separator string using this keyword argument. end By default, a newline ("\n") is written after the last value in args. You may use this keywoard argument to specify a different line terminator, or no terminator at all. file Output normally goes to the standard output stream (sys.stdout). To divert the output to another writeable file, use this keyword argument. Here's an example. Suppose you are writing three strings named clan, moiety, and distro to a writeable file named spreader, and you want to separate the fields with tab ("\t") characters, and use ASCII CR, Carriage Return ("\r"), as the line terminator. Your call to the print() function would go something like this: print(clan, moiety, distro, file=spreader, end='\r', sep='\t')

Compound statements The statements in this section alter the normal sequential execution of a program. They can cause a statement to be executed only under certain circumstances, or execute it repeatedly.

Python's block structure One unusual feature of Python is the way that the indentation of your source program organizes it into blocks within blocks within blocks. This is contrary to the way languages like C and Perl organize code blocks by enclosing them in delimiters such as braces

{
        ... }

. Various Python branching statements like if and for control the execution of blocks of lines. At the very top level of your program, all statements must be unindented—they must start in column one. Various Python branching statements like if and for control the execution of one or more subsidiary blocks of lines. A block is defined as a group of adjacent lines that are indented the same amount, but indented further than the controlling line. The amount of indentation of a block is not critical. You can use either spaces or tab characters for indentation. However, mixing the two is perverse and can make your program hard to maintain. Tab stops are assumed to be every eight columns. Blocks within blocks are simply indented further. Here is an example of some nested blocks: if i < 0: print "i is negative" else: print "i is nonnegative" if i < 10: print "i has one digit" else: print "i has multiple digits" If you prefer a more horizontal style, you can always place statements after the colon (:) of a compound statement, and you can place multiple statements on a line by separating them with semicolons (;). Example: >>> if 2 > 1: print "Math still works"; print "Yay!" ... else: print "Huh?" ... Math still works Yay! You can't mix the block style with the horizontal style: the consequence of an if or else must either be on the same line or in a block, never both. >>> if 1: print "True" ... print "Unexpected indent error here." File "<stdin>", line 2 print "Unexpected indent error here." ^ IndentationError: unexpected indent >>>

The <code >break</code > statement: Exit a <code >for</code > or <code >while</code > loop The purpose of this statement is to jump out of a for or while loop before the loop would terminate otherwise. Control is transferred to the statement after the last line of the loop. The statement looks like this: break Here's an example. >>> for i in [1, 71, 13, 2, 81, 15]: ... print i, ... if (i%2) == 0: ... break ... 1 71 13 2 Normally this loop would be executed six times, once for each value in the list, but the break statement gets executed when i is set to an even value.

The <code >continue</code > statement: Jump to the next cycle of a <code >for</code > or <code >while</code > Use a continue statement inside a for or while loop when you want to jump directly back to the top of the loop and go around again. If used inside a while loop, the loop's condition expression is evaluated again. If the condition is False, the loop is terminated; if the condition is True, the loop is executed again. Inside a for loop, a continue statement goes back to the top of the loop. If there are any values remaining in the iterable that controls the loop, the loop variable is set to the next value in the iterable, and the loop body is entered. If the continue is executed during the last pass through the loop, control goes to the statement after the end of the loop. Examples: >>> i = 0 >>> while i < 10: ... print i, ... i += 1 ... if (i%3) != 0: ... continue ... print "num", ... 0 1 2 num 3 4 5 num 6 7 8 num 9 >>> for i in range(10): ... print i, ... if (i%4) != 0: ... continue ... print "whee", ... 0 whee 1 2 3 4 whee 5 6 7 8 whee 9

The <code >for</code > statement: Iteration over a sequence Use a for statement to execute a block of statements repeatedly. Here is the general form. (For the definition of a block, see .) for V in S: B V is a variable called the induction variable. S is any iterable; see . This iterable is called the controlling iterable of the loop. B is a block of statements. The block is executed once for each value in S. During each execution of the block, V is set to the corresponding value of S in turn. Example: >>> for color in ['black', 'blue', 'transparent']: ... print color ... black blue transparent In general, you can use any number of induction variables. In this case, the members of the controlling iterable must themselves be iterables, which are unpacked into the induction variables in the same way as sequence unpacking as described in . Here is an example. >>> fourDays = ( ('First', 1, 'orangutan librarian'), ... ('Second', 5, 'loaves of dwarf bread'), ... ('Third', 3, 'dried frog pills'), ... ('Fourth', 2, 'sentient luggages') ) >>> for day, number, item in fourDays: ... print ( "On the {1} day of Hogswatch, my true love gave " ... "to me".format(day) ) ... print "{0} {1}".format(number, item) ... On the First day of Hogswatch, my true love gave to me 1 orangutan librarian On the Second day of Hogswatch, my true love gave to me 5 loaves of dwarf bread On the Third day of Hogswatch, my true love gave to me 3 dried frog pills On the Fourth day of Hogswatch, my true love gave to me 2 sentient luggages You can change the induction variable inside the loop, but during the next pass through the loop, it will be set to the next element of the controlling iterable normally. Modifying the controlling iterable itself won't change anything; Python makes a copy of it before starting the loop. >>> for i in range(4): ... print "Before:", i, ... i += 1000 ... print "After:", i ... Before: 0 After: 1000 Before: 1 After: 1001 Before: 2 After: 1002 Before: 3 After: 1003 >>> L = [7, 6, 1912] >>> for n in L: ... L = [44, 55] ... print n ... 7 6 1912

The <code >if</code > statement: Conditional execution The purpose of the if construct is to execute some statements only when certain conditions are true. Here is the most general form of an if construct: if E0: B0 elif E1: B1 elif ...: ... else: Bf In words, this construct means: If expression E0 is true, execute block B0. If expression E0 is false but E1 is true, execute block B1. If there are more elif clauses, evaluate each one's expression until that expression has a true value, and then execute the corresponding block. If all the expressions in if and elif clauses are false, execute block Bf. An if construct may have zero or more elif clauses. The else clause is also optional. Examples: >>> for i in range(5): ... if i == 0: ... print "i is zero", ... elif i == 1: ... print "it's one", ... elif i == 2: ... print "it's two", ... else: ... print "many", ... print i ... i is zero 0 it's one 1 it's two 2 many 3 many 4 >>> if 2 > 3: print "Huh?" ... >>> if 2 > 3: print "Huh?" ... else: print "Oh, good." ... Oh, good. >>> if 2 > 3: print "Huh?" ... elif 2 == 2: print "Oh." ... Oh.

The <code >raise</code > statement: Cause an exception Python's exception mechanism is the universal framework for dealing with errors—situations where your program can't really proceed normally. For an overview, see . There are three forms of the raise statement: raise raise E1 raise E1, E2 The first form is equivalent to “

raise
        None,None

” and the second form is equivalent to “raise E1, None”. Each form raises an exception of a given type and with a given value. The type and value depend on how many expressions you provide: E1 E2 Exception type Exception value None None Re-raise the current exception, if any. This might be done, for example, inside an except, else, or finally block; see . class None E1 E1() class instance of E1 E1 E2 class tuple E1 E1(*E2) class none of the above E1 E1(E2) instance None type(E1) E1 The current recommended practice is to use a raise statement of this form: raise E(...) where E is some class derived from the built-in Exception class: you can use one of the built-in exceptions, or you can create your own exception classes. For classes derived from Exception, the constructor takes one argument, an error message—that is, a string explaining why the exception was raised. The resulting instance makes that message available as an attribute named .message. Example: >>> try: ... raise ValueError('The day is too frabjous.') ... except ValueError as x: ... pass ... >>> type(x) <type 'exceptions.ValueError'> >>> x.message 'The day is too frabjous.' To create your own exceptions, write a class that inherits from Exception and passes its argument to the parent constructor, as in this example. >>> class VocationError(Exception): ... def __init__(self, mismatch): ... Exception.__init__(self, mismatch) ... >>> try: ... print "And now, the Vocational Guidance Counsellor Sketch." ... raise VocationError("Does not have proper hat") ... print "This print statement will not be reached." ... except VocationError as problem: ... print "Vocation problem: {0}".format(problem) ... And now, the Vocational Guidance Counsellor Sketch. Vocation problem: Does not have proper hat

The <code >return</code > statement: Exit a function or method Within any function or method, you can exit the function with a return statement. There are two forms: return expression return In the first form, execution resumes at the point where the function or method was called, and the value of the expression is substituted into the calling statement. The second form is the equivalent of “return None”. (See .)

The <code >try</code > statement: Anticipate exceptions The purpose of a “try” construct is to specify what actions to take in the event of errors. For an introduction to Python's exception mechanism, see . When an exception is raised, two items are associated with it: An exception type, and an exception value. Here is the most general form of a try construct. Symbols like B0 and B1 represent indented blocks of statements. Each except clause names some exception class Ei (or a tuple of exception classes), and optionally a variable vi that will be set to the exception value. try: B0 except E1 [as v1]: B1 except E2 [as v2]: B2 except ...: ... else: Be finally: Bf The else: and finally: blocks are optional. There must be at least one except block, but there may be any number. Here is a simplified description of the execution of a try block in general: If B0 executes without raising any exceptions, the else block Be is executed, then the finally block Bf. If the execution of block B0 raises some exception with type E0, that type is compared against each except clause until one of them matches the raised exception or there are no more except clauses. The matching condition is slightly complicated: for some clause “except Ei as vi:”, expression Ei is either an exception class or a tuple of exception classes. If Ei is a single class, it is considered a match if E0 is either the same class as Ei or a subclass of Ei. If Ei is a tuple of exception classes, the raised exception is compared to each to see if it is the same class or a subclass, as in the single-class case. If multiple except clauses match, the first matching clause is said to handle the exception. The corresponding variable vi (if present) is bound to the raised exception instance, and control passes to the corresponding block Bi. If there is a finally clause, it is executed, whether the exception was caught or not. If the exception was not caught, it is re-raised after the end of the finally clause. Examples: >>> try: ... raise ValueError("Strike three!") ... except IOError as x: ... print "I/O error caught:", x ... except ValueError as x: ... print "Value error caught:", x ... except SyntaxError as x: ... print "Syntax error caught:", x ... else: ... print "This is the else clause" ... finally: ... print "This is the finally clause" ... Value error caught: Strike three! This is the finally clause >>> try: ... raise ValueError("Uncaught!") ... except IOError as x: ... print "I/O error:", x ... else: ... print "This is the else clause" ... finally: ... print "This is the finally clause" ... This is the finally clause Traceback (most recent call last): File "<stdin>", line 2, in <module> ValueError: Uncaught! >>> try: ... print "No exceptions raised" ... except ValueError as x: ... print "ValueError:", x ... else: ... print "This is the else clause" ... finally: ... print "This is the finally clause" ... No exceptions raised This is the else clause This is the finally clause For those of you who are interested in the gory details, the fun begins when a second or even a third exception is raised inside an except, else, or finally clause. The results are well-defined, and here is a pseudocode description of the edge cases. In this procedure, we'll use two internal variables named pending and detail. Set pending to None. Attempt to execute block B0. If this block raises an exception E0 with detail d0, set pending to E0 and set detail to d0. If pending is None, go to . Find the first block

except Ei,
            vi:

such that issubclass(E0, Ei ). If there is no such match, go to . Set vi to detail. Attempt to execute block Bi. If this block raises some new exception En with detail dn, set pending to En and set detail to dn. However, if block Bi executes without exception, set pending to None. In this case, the original exception is said to have been caught or handled. Go to . If there is no else: clause, go to . Attempt to execute the else: block Be. If this block raises some new exception En with detail dn, set pending to En and set detail to dn. If there is no finally: clause, proceed to . Attempt to execute the finally: block Ef. If this block raises some new exception En with detail dn, set pending to En and set detail to dn. If pending is not None, re-raise the exception as in this statement: raise pending, detail If pending is None, fall through to the statement following the try: block.

The <code >with</code > statement and context managers The purpose of this statement is to protect a block of code with a context manager that insures that certain initialization and cleanup steps get performed, regardless of whether that block raises an exception. A context manager is a class that has .__enter__() and .__exit__() methods. The .__enter__() method performs any necessary initialization, and returns a value. The .__exit__() method is always executed to perform necessary cleanup actions. Starting with version 2.6, Python now was a with statement that executes a block using a context manager. Here is the general form, where B is the block to be executed. with E[ as V]: B E is an expression that evaluates to a context manager. If you provide a variable V in the optional part “as V”, that variable will be set to the value returned by the context manager's .__enter__() method. Python's file class is a context manager; its .__enter__() method returns the opened file, and its .__exit__() method closes the file. For example, suppose you want to call a function beat() and pass it an opened file named "goat", but you want to be sure the file is closed even if the beat() function raises an exception. This code would do that: with open("goat") as inFile: beat(inFile) When the beat() function terminates (either normally or because it raised an exception), the file will be closed (in the file.__exit__() method). If the function raised an exception, that exception will then be re-raised. Starting in Python 2.7, you can enclose a block in multiple context managers by repeating the “E[ as V” parts of the statement. For example: with open('in') as inFile, open('out', 'w') as outFile: for rawLine in inFile: outFile.write(rawLine)

The <code >yield</code > statement: Generate one result from a generator A generator is any function or method that contains at least one yield statement. Generators are a special type of iterator; see . Here is the general form: yield expression Unlike ordinary functions or methods that use the return statement to return a single value, a generator is a mechanism that produces a sequence of zero or more values. Each execution of a yield statement produces an additional value. To signal the caller that there are no more values, use this raise statement: raise StopIteration As an example, here is a function that generates the sequence 0, 1, 2, ..., n-1, n, n-1, n-2, ..., 2, 1, 0. >>> def updown(n): ... '''Generate the values 0, 1, 2, ..., n-1, n, n-1, n-2, ...0. ... ''' ... for answer in range(0,n): ... yield answer ... for answer in range(n, -1, -1): ... yield answer ... raise StopIteration ... >>> for x in updown(4): ... print x, ... 0 1 2 3 4 3 2 1 0

<code >def()</code >: Defining your own functions The def construct is used to define functions and methods. Here is the general form: def n(p0[=e0][,p1[=e1]]...[,*pv][,**pd]): B The name n of the function is followed by a pair of parentheses containing descriptions of the arguments to the function. The block B is called the body of the function, and is executed when the function is called. A function may have no arguments at all. If there are arguments to be passed to the function when it is called, they must be declared in this order: A positional argument is a name that is not followed by an equal sign (=) and default value. A keyword argument is followed by an equal sign and an expression that gives its default value. If a function has both positional arguments and keyword arguments, all positional arguments must precede all keyword arguments. If there is a *pv parameter, when the function is called that name is bound to a (possibly empty) tuple of all positional arguments passed to the function that do not correspond to other positional or keyword arguments in the def. If there is a **pd parameter, when the function is called that name is bound to a dictionary of all keyword arguments passed to the function that do not appear in the function's def. When you call a function, the argument values you pass to it must obey these rules: There are two kinds of arguments: positional (also called non-default arguments) and keyword (also called default arguments). A positional argument is simply an expression, whose value is passed to the argument. A keyword argument has this form: name=expression All positional arguments in the function call (if any) must precede all keyword arguments (if any). >>> def wrong(f=1, g): ... print f, g ... File "<stdin>", line 1 SyntaxError: non-default argument follows default argument You must supply at least as many positional arguments as the function expects. >>> def wantThree(a, b, c): ... print a,b,c ... >>> wantThree('nudge', 'nudge', 'nudge') nudge nudge nudge >>> wantThree('nudge') Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: wantThree() takes exactly 3 arguments (1 given) If you supply more positional arguments than the function expects, the extra arguments are matched against keyword arguments in the order of their declaration in the def. Any additional keyword arguments are set to their default values. >>> def f(a, b, c=1, d='elk'): ... print a,b,c,d ... >>> f(99, 111) 99 111 1 elk >>> f(99, 111, 222, 333) 99 111 222 333 >>> f(8, 9, 10, 11, 12, 13) Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: f() takes at most 4 arguments (6 given) You may supply arguments for keyword parameters in any order by using the form k=v, where k is the keyword used in the declaration of that parameter and v is your desired argument. >>> def blackKeys(fish='Eric', dawn='Stafford', attila='Abdul'): ... print fish, dawn, attila ... >>> blackKeys() Eric Stafford Abdul >>> blackKeys(attila='Gamera', fish='Abdul') Abdul Stafford Gamera If you declare a parameter of the form “*name”, the caller can provide any number of additional keyword arguments, and the name will be bound to a tuple containing those additional arguments. >>> def posish(i, j, k, *extras): ... print i,j,k,extras ... >>> posish(38, 40, 42) 38 40 42 () >>> posish(44, 46, 48, 51, 57, 88) 44 46 48 (51, 57, 88) Similarly, you may declare a final parameter of the form “**name”. If the caller provides any keyword arguments whose names do not match declared keyword arguments, that name will be bound to a dictionary containing the additional keyword arguments as key-value pairs. >>> def extraKeys(a, b=1, *c, **d): ... print a, b, c, d ... >>> extraKeys(1,2) 1 2 () {} >>> extraKeys(3,4,6,12, hovercraft='eels', record='scratched') 3 4 (6, 12) {'record': 'scratched', 'hovercraft': 'eels'}

A function's local namespace Any name that appears in a function's argument list, or any name that is set to a value anywhere in the function, is said to be local to the function. If a local name is the same as a name from outside the function (a so-called global name), references to that name inside the function will refer to the local name, and the global name will be unaffected. Here is an example: >>> x = 'lobster' >>> y = 'Thermidor' >>> def f(x): ... y = 'crevettes' ... print x, y ... >>> f('spam') spam crevettes >>> print x, y lobster Thermidor Keyword parameters have a special characteristic: their names are local to the function, but they are also used to match keyword arguments when the function is called.

Iterators: Values that can produce a sequence of values Closely related to Python's concept of sequences is the concept of an iterator:

For a given sequence S, an iterator I is essentially a set of instructions for producing the elements of S as a sequence of zero or more values.

To produce an iterator over some sequence S, use this function: iter(S) The result of this function is an “iterator object” that can be used in a for statement. >>> continents = ('AF', 'AS', 'EU', 'AU', 'AN', 'SA', 'NA') >>> worldWalker = iter(continents) >>> type(worldWalker) <type 'tupleiterator'> >>> for landMass in worldWalker: ... print "Visit {0}.".format(landMass,) ... Visit AF. Visit AS. Visit EU. Visit AU. Visit AN. Visit SA. Visit NA. All iterators have a .next() method that you can call to get the next element in the sequence. This method takes no arguments. It returns the next element in the sequence, if any. When there are no more elements, it raises a StopIteration exception. >>> trafficSignal = [ 'green', 'yellow', 'red' ] >>> signalCycle = iter(trafficSignal) >>> type(signalCycle) <type 'listiterator'> >>> signalCycle.next() 'green' >>> signalCycle.next() 'yellow' >>> signalCycle.next() 'red' >>> signalCycle.next() Traceback (most recent call last): File "<stdin>", line 1, in <module> StopIteration Once an iterator is exhausted, it will continue to raise StopIteration indefinitely. You can also use an iterator as the right-hand operand of the “in” operator. >>> signalCycle = iter(trafficSignal) >>> 'red' in signalCycle True

Generators: Functions that can produce a sequence of values Unlike conventional functions that return only a single result, a generator is a function that produces a sequence of zero or more results. Generators are a special case of iterators (see ), so they can be used as the controlling iterable in for statements and the other places where iterators are allowed. In a conventional function, the body of the function is executed until it either executes a return statement, or until it runs out of body statements (which is the equivalent of a “return None” statement). By contrast, when a generator function is called, its body is executed until it either has another value to produce, or until there are no more values. When a function wishes to return the next generated value, it executes a statement of this form: yield e where the e is any Python expression. The difference between yield and return is that when a return is executed, the function is considered finished with its execution, and all its current state diasppears. By contrast, when a function executes a yield statement, execution of the function is expected to resume just after that statement, at the point when the caller of the function needs the next generated value. A generator signals that there are no more values by executing this statement: raise StopIteration For an example of a generator, see . If you are writing a container class (that is, a class whose instances are containers for a set of values), and you want to define an iterator (see ), that method can be a generator. Here is a small example. The constructor for class Bunch takes a sequence of values and stores them in instance attribute .__stuffList. The iterator method .__iter__() generates the elements of the sequence in order, except it wraps each of them in parentheses: >>> class Bunch(object): ... def __init__(self, stuffList): ... self.__stuffList = stuffList ... def __iter__(self): ... for thing in self.__stuffList: ... yield "({0})".format(thing) ... raise StopIteration ... >>> mess = Bunch(('lobster Thermidor', 'crevettes', 'Mornay')) >>> for item in mess: ... print item, ... (lobster Thermidor) (crevettes) (Mornay) >>> messWalker = iter(mess) >>> for thing in messWalker: print thing, ... (lobster Thermidor) (crevettes) (Mornay)

Decorators The purpose of a Python decorator is to replace a function or method with a modified version at the time it is defined. For example, the original way to declare a static method was like this: def someMethod(x, y): ... someMethod = staticmethod(someMethod) Using Python's decorator syntax, you can get the same effect like this: @staticmethod def someMethod(x, y): ... In general, a function or method may be preceded by any number of decorator expressions, and you may also provide arguments to the decorators. If a function f is preceded by a decorator expression of the form “@d”, it is the equivalent of this code: def f(...): ... f = d(f) You may provide a parenthesized argument list after the name of your decorator. A decorator expression d(...) is the equivalent of this code: def f(...): ... f = d(...)(f) First, the decorator is called with the argument list you provided. It must return a callable object. That callable is then called with one argument, the decorated function. The name of the decorated function is then bound to the returned value. If you provide multiple decorators, they are applied inside out, in sequence from the last to the first. Here is an example of a function wrapped with two decorators, of which the second has additional arguments: @f1 @f2('Pewty') def f0(...): ... This is the equivalent code without using decorators: def f0(...): ... f0 = f1 ( f2('Pewty') ( f0 ) ) First function f2 is called with one argument, the string 'Pewty'. The return value, which must be callable, is then called with f0 as its argument. The return value from that call is then passed to f1. Name f0 is then bound to the return value from the call to f1.

Exceptions: Error signaling and handling Python's exception system provides a way to signal error conditions and other disruptions in normal processing, and also a way for programs to recover from these conditions. . . .

Definitions of exception terms Some definitions: To raise an exception means to signal that the program cannot proceed normally due to an error or other condition. (In other programming languages, such as Java, the equivalent term is to throw an exception.) Two values accompany the raising of an exception: the type and the value. For example, if a program attempts to open an existing disk file but there is no such file, the type is IOError, and the value is an instance of the IOError class that contains additional information about this error. For more information about raising exceptions, see . A program may choose to handle an exception. That is, a program may say that if a certain exception or category of exceptions occurs in a specific block of code, Python must execute another code block called a handler. A traceback is a message from Python showing where an exception occurred. If you type a statement in conversational mode that causes an exception, you will see a short traceback like this: >>> x = 59 / 0 Traceback (most recent call last): File "<stdin>", line 1, in <module> ZeroDivisionError: integer division or modulo by zero The above example showed that the offending statement was read from the standard input stream (<stdin>). When looking at a traceback, always look at the last line first. It tells you the general type of exception (in the example, a ZeroDivisionError), followed by additional details (“integer division or modulo by zero”). If an exception occurs inside one or more function calls, the traceback will give a complete list of the functions involved, from outermost to innermost. Again, the last line shows the exception type and details. >>> def f(): g() ... >>> def g(): h() ... >>> def h(): return 1/0 ... >>> f() Traceback (most recent call last): File "<stdin>", line 1, in <module> File "<stdin>", line 1, in f File "<stdin>", line 1, in g File "<stdin>", line 1, in h ZeroDivisionError: integer division or modulo by zero

Life cycle of an exception If you anticipate that executing a particular statement may cause an exception and you don't want your program to terminate and display a traceback, you can use a try construct to specify handlers to be executed if an exception occurs. For details, see . If an exception occurs inside a function and it is not handled at that level by a try construct, Python will work back through the pending function calls until it either finds a handler for that exception or runs out of pending function calls. If there are multiple handlers for the exception in calling functions, the innermost will be used. If there are no handlers for the exception in calling functions, you will get a stack traceback and the program will terminate. >>> def f(): ... try: ... g() ... except ValueError, detail: ... print "Caught a ValueError:", detail.message ... >>> def g(): h() ... >>> def h(): ... raise ValueError('This is a test.') ... >>> f() Caught a ValueError: This is a test. In the example above, function f() calls function g(), which in turn calls function h(). Function h() raises a ValueError exception, but there is no try: block around it. Python looks to see if there is a ValueError handler in g(), but there is not. Finally a handler for ValueError is found inside function f(), so control resumes inside that handler. Note that no stack traceback is displayed, because the ValueError exception was handled successfully.

Built-in exceptions Python defines a complete hierarchy of built-in exception classes. When you write a handler, you can specify any class in this hierarchy, and that handler will apply to that class and any derived classes. This allows you to write generic handlers that catch whole groups of exception types. In describing this hierarchy, we will use indentation to show the subclass/parent class relationships. Generally a raise statement will name one of the “leaf” classes, that is, a class that does not have any subclasses. Ancestor classes that are not usually raised are marked with an asterisk (*) in the section below. BaseException*: This is the ancestor of all exception classes. The constructor for this class takes one argument, a string containing the error message. That argument is available as the .message attribute. SystemExit: Raise this exception to terminate execution of your program; it is a special case in that it does not produce a stack traceback. In module sys, the .exit() method raises this exception. It is possible to write a handler for this exception. To defeat such a handler and force immediate termination, import module os and use method os._exit(). KeyboardInterrupt: Raised when the user signals an interruption with the keyboard (del under Windows, or Control-C in Linux or Mac environments). This class inherits from BaseException rather than from Exception so that catch-all handlers for class Exception will not prevent program termination. Exception*: This is the preferred base class for all built-in and user-defined exceptions. If you want to write a handler for all these exception types, use Exception as the handler's type. >>> try: ... x = 1 / 0 ... except Exception, detail: ... print "Fail:", detail.message ... Fail: integer division or modulo by zero >>> try: ... x = noSuchVariable ... except Exception, detail: ... print "Fail:", detail.message ... Fail: name 'noSuchVariable' is not defined A catch-all handler like this can mask any number of errors. Do not use such a handler unless your program must absolutely stay running. StopIteration: This is the exception that a generator must raise in order to signal that no more generated values are available. See . StandardError*: This is the base class for all built-in exceptions that are considered errors. ArithmeticError*: This is the base class for errors involving arithmetic computations. FloatingPointError: This is raised for arithmetic errors involving the float type. OverflowError: This is raised when the result of an operation cannot be represented. ZeroDivisionError: An attempt to divide by zero. AssertionError: An assert statement has failed. See . AttributeError: Failure to access an attribute. EnvironmentError*: Errors caused by functions outside of Python, such as the operating system or peripheral devices. IOError: Errors related to file input or output. OSError: Errors signaled from the operating system. ImportError: Failure to import a module or to import items from a module. LookupError*: Superclass for errors caused by attempts to retrieve values from inside a container class. IndexError: Attempt to retrieve a sequence member S[I], where I is not a valid index in sequence S. KeyError: Attempt to retrieve a dictionary member D[K], where K is not a valid key in D. MemoryError: No more processor memory is available. NameError: Attempt to retrieve a name that is not defined. UnboundLocalError: Attempt to retrieve the value of a local name when no value has yet been assigned to it. RuntimeError: An error that doesn't fit the other categories. NotImplementedError: This is the preferred way for the virtual methods of a base class to signal that they have not been replaced by a concrete method in a derived class. SyntaxError: Attempt to execute invalid Python source code. TypeError: Attempt to perform an operation on a value that does not support that operation, such as trying to use exponentiation (**) on a string. ValueError: Caused by an operation that is performed on values of the correct type, but the actual values are not valid. Example: taking a negative number to a fractional power.

Classes: Defining your own types This section assumes you already understand the basics of object-oriented programming in Python, and that you know the meaning of concepts such as class, instance, method, and attribute. For a general tutorial on these concepts, see the introduction to object-oriented Python programming in the Tech Computer Center's Python tutorial. Here is the general form of the class declaration for some class C with one or more parent classes P1, P2, …: class C(P1, P2, ...): attribute definitions ... To declare a class that does not inherit from any parent classes: class C: attribute definitions ... The attribute definitions may include any number of def blocks that declare methods of the class, and any number of class variable declarations. Functionally, a class is really just a namespace. This namespace is just a place to store the pieces of the class mechanisms: its methods and class variables. When Python reads a “class” declaration, it creates a new, empty namespace. When Python reads a “def” within a class, the name of that method is added to the class's namespace. If you define a class variable (that is, if you assign a value to a name inside a class but outside of any methods of the class), the class variable's name and value are added to the namespace. A brief conversational session may serve to illustrate these concepts. We'll make use of the built-in function dir() to show the contents of the class's namespace; see . >>> class Taunter: ... tauntCount = 0 ... def taunt(self): ... print "Go away, or I shall taunt you a second time!" ... >>> dir(Taunter) ['__doc__', '__module__', 'taunt', 'tauntCount'] >>> type(Taunter.__doc__) <type 'NoneType'> >>> Taunter.__module__ '__main__' >>> Taunter.tauntCount 0 >>> Taunter.taunt <unbound method Taunter.taunt> When Python reads this line, it adds the name Taunter to the current local namespace, bound to a new, empty namespace of type class. Because this assignment takes place inside class Taunter but not inside a def, name tauntCount becomes a class variable, bound to an int value of zero. The next two lines define a method named taunt() within the class. After we've finished entering the class definition, we use dir(Taunter) to see what names are in the class's namespace. Variables __doc__ and __module__ are added automatically. Because there was no documentation string in the class, __doc__ is bound to None. The __module__ variable has the value '__main__' because the class was entered in conversational mode. To retrieve the value of a class variable V in class C, use the syntax “C.V”. Name taunt in the class namespace is bound to an object of type “unbound method.” An unbound method is a method (function) that is inside a class, but it is not associated with an instance of the class. An instance of a class is also a namespace. When the instance is created, all the names from the class's namespace are copied into the instance namespace. From that point on, any changes made to the instance's namespace do not affect the class namespace: >>> frenchy = Taunter() >>> dir(frenchy) ['__doc__', '__module__', 'taunt', 'tauntCount'] >>> frenchy.where = 'crenelations' >>> dir(frenchy) ['__doc__', '__module__', 'where', 'taunt', 'tauntCount'] >>> frenchy.where 'crenelations' >>> dir(Taunter) ['__doc__', '__module__', 'taunt', 'tauntCount'] >>> frenchy.tauntCount 0 >>> frenchy.tauntCount += 1 >>> frenchy.tauntCount 1 >>> Taunter.tauntCount 0 >>> type(frenchy.taunt) <type 'instancemethod'> >>> frenchy.taunt() Go away, or I shall taunt you a second time! >>> Taunter.taunt(frenchy) Go away, or I shall taunt you a second time! This class does not have a constructor (__init__) method, so when an instance is created, the instance is a namespace with the same names as the class, and the same values. This line adds a new name where to the instance's namespace. It is bound to the string value 'crenelations'. Note that the instance namespace now contains the name where, but the class's namespace is unchanged. To retrieve an attribute A of an instance I, use the syntax “I.A”. Initially, the instance variable has the same value as the class variable of the same name. Here, we add one to the instance variable tauntCount. The instance variable has the new value, but the class variable tauntCount is unchanged. Within the instance namespace, name taunt is now a bound method: it is associated with the instance frenchy. The next two lines show two equivalent methods of calling the taunt method. Most method calls are bound method calls. To call a bound method B of an instance I, use the syntax “I.B(...)”. When a method B is bound to an instance I, the instance namespace I becomes the “self” argument passed in to the method. This line has the same effect as the previous line, but it is an unbound method call. The expression “Taunter.taunt” retrieves the unbound method from the class definition. When you call an unbound method, you must supply the “self” argument explicitly as the first argument. Unbound method calls are not terribly common, but you will need to know about them when you write the constructor for a derived class: you must call the parent class constructor as an unbound call. Generally, if class D has parent class C, the derived class might look something like this: class D(C): def __init__(self, ...): C.__init__(self, ...) ... Namespaces are very much like dictionaries. Where a dictionary has unique keys, a namespace has unique names. As a matter of fact, classes and instances have a special built-in attribute called “__dict__” which, for most purposes, is the namespace as a dictionary. Continuing the examples above: >>> Taunter.__dict__ {'taunt': <function taunt at 0xb7ed002c>, '__module__': '__main__', 'tau ntCount': 0, '__doc__': None} >>> newFrenchy=Taunter() >>> newFrenchy.__dict__ {} >>> frenchy.__dict__ {'tauntCount': 1, 'where': 'crenelations'} The class's dictionary has the four names we expect: the built-ins __module__ and __doc__, the class variable tauntCount, and the method taunt. But notice that the __dict__ attribute of the newly created instance newFrenchy does not have the four names copied from the class. In fact, it is empty. And the __dict__ of instance frenchy contains only the names that have changed since its instantation. What actually happens when you refer to an attribute is that Python looks first in the instance's __dict__; if the name is not found there, it looks in the __dict__ of the class. For derived classes, Python will also search the __dict__ attributes of all the ancestor classes. So, in our example, a reference to frenchy.tauntCount would find the value of 1 in the instance. A reference to newFrenchy.tauntCount would fail to find that name in newFrench.__dict__, but would succeed in finding the class variable value 0 in Taunter.__dict__['tauntCount']. Let's now look at the life cycles of classes in more detail. Due to improvements made in the language since it was first introduced, Python has two kinds of classes, old-style and new-style. We encourage you to use new-style classes; old-style classes will no longer be supported in the next major release, Python 3000. All new-style classes must declare at least one parent class that is either the top-level class object or some other class that derives ultimately from object. Such a class is said to be derived from, or inherits from, the object class. To declare a new-style class C that inherits from object: class C(object): ...class methods and variables... An old-style class is one that doesn't declare a parent class at all, or a class that inherits from an existing old-style class. The life cycle of an old-style class is described in . In most respects, the two classes perform identically. We'll start by explaining old-style classes in . To benefit from the many functional improvements of new-style classes, and especially if you expect to migrate your code to the major changes of Python 3.0, see .

Old-style classes Old-style classes are those declared without a parent class, or classes that inherit from an existing old-style class. Here is an outline of the birth, life, and death of an old-style class and its instances. . . . . .

Defining an old-style class To define an old-style class C with no parent class, use this general form: class C: ...class methods and variables... To create a class that inherits from one or more parent classes P1, P2, …: class C(P1, P2, ...): ...class methods and variables... As Python reads the definition of your class, it first creates a new, empty namespace called the class namespace. You can access the class namespace directly as an attribute named __dict__, a dictionary whose keys are the names in that namespace. As Python reads each method or class variable in your class declaration, it adds the name of that method or variable to the class namespace.

Instantiation of an old-style class: The constructor, <code >.__init__()</code > The creation of a new instance of a class happens when a running Python program encounters a call to that class, that is, the class name with a pair of parentheses after it, with zero or more arguments inside the parentheses. Here is the general form: C(a1, a2, ...) The instance creation (also called instantiation) is handled by the __init__() or constructor method. First Python creates the instance with an empty .__dict__ attribute that will contain the instance's values. Python then calls the constructor. The argument list for this call always has the special first argument self (the instance), followed by whatever arguments were used in the initial call. The constructor call is equivalent to this: C.__init__(self, a1, a2, ...) The constructor method then executes. Typically the constructor will set up new instance attributes by assignments of this form: self.name = expression When the constructor finishes executing, the instance is returned to the constructor's caller.

Attribute references in old-style classes The names inside the instance are called attributes. (The methods are technically attributes—attributes that are of type function.) There are three operations on attributes: get, set, and delete. To get an attribute means to retrieve its value. Python searches the instance's .__dict__ for the attribute's name; if that name is found, its value is returned. If the instance's .__dict__ does not have a binding for the given name, Python searches the class's .__dict__. If no value is found there, Python searches the namespaces of the ancestor classes (if any). >>> class C: ... def __init__(self, x): ... self.thingy = x ... >>> c=C(42) >>> c.thingy 42 >>> c.__dict__['thingy'] 42 When you call a method M of an instance I in an expression of the form “I.M(...)”, this is considered just another attribute “get” operation: the get operation I.M retrieves the method, and then that method is called using the arguments inside the “(...)”. To set an attribute means to give it a value. If there is an existing attribute with the same name, its old value is discarded, and the attribute name is bound to the new value. The new value is stored in the instance's .__dict__. >>> c.thingy 42 >>> c.thingy = 58 >>> c.thingy 58 >>> c.__dict__['thingy'] 58 You can delete an attribute from an instance using a del statement (see ). >>> c.thingy 58 >>> del c.thingy >>> c.thingy Traceback (most recent call last): File "<stdin>", line 1, in <module> AttributeError: C instance has no attribute 'thingy' In addition to ordinary attributes and methods, your class can accept references to names that do not exist in the instance or class namespace. You can define special methods that will be called when some statement tries to get, set, or delete an attribute that isn't found in the instance's .__dict__. See , , and . If all else fails—if an attribute is not found in the instance's namespace and the class does not provide a special method that handles the attribute reference—Python will raise an AttributeError exception.

Method calls in an old-style class There are two different ways to call a method M of some class C: Most calls are bound method calls of this form, where I is an instance of some class C: I.methodName(a1, a2, ...) The instance I replaces self as the first argument when setting up the arguments to be passed to the method. The following form, called an unbound method call, is exactly equivalent to the above: C.methodName(i, a1, a2, ...) Here is a demonstration of the equivalence of bound and unbound method calls. >>> class C: ... def __init__(self, x): ... self.x = x ... def show(self, y): ... print "*** ({0},{1}) ***".format(self.x, y) ... >>> c=C(42) >>> c.show(58) *** (42,58) *** >>> C.show(c,58) *** (42,58) ***

Instance deletion: the destructor, <code >.__del__()</code > When there are no values that refer to an instance any more, the storage occupied by the instance is recycled. However, if there are certain cleanup operations that must be done (such as closing external files), these operations can be placed into a destructor method that will be called before recycling the instance. Here is the general form of a destructor: def __del__(self): ...cleanup statements...

Life cycle of a new-style class Most of the features of new-style classes are the same as for old-syle classes. This section will discuss only the differences. We won't cover a few of the more obscure advanced features here; for information on such topics as descriptors and metaclasses, see the “Data model” section of the Python Reference Manual. The declaration of a new-style class looks the same as for an old-style class, with one constraint: the class must inherit from the universal base class named object, or from one or more other new-style classes. . . . .

<code >__new__()</code >: New instance creation New-style classes have a new special method name, __new__(), that is called on instantiation before the constructor. It handles the creation of a new instance. The .__new__() method is called when an instance is created. Method .__new__() is always a static method (see ), even if you do not specifically make it a static method. A constructor call for some class C has this general form: C(*p, **k) That is, it can have any number of positional arguments and any number of keyword arguments. The equivalent call to the .__new__() method will look like this: def __new__(cls, *p, **k): ... The first argument cls must be the class being created. The .__new__() method must call the parent class's .__new__() method to create the instance. For example, if your class inherits directly from object, you must call: object.__new__(cls) The value returned by that call is the new instance. In most cases, the .__new__() method will return a new instance of cls, and that class's .__init__() will then be called with that instance as its self argument, and the positional and keyword arguments p and k will be passed to that constructor as well. >>> class Test(object): ... def __new__(cls, *p, **k): ... inst = object.__new__(cls) ... return inst ... def __init__(self, *p, **k): ... print "p={0} k={1}".format(p, k) ... >>> t=Test('egg', 'kale', sauce='Bearnaise') p=('egg', 'kale') k={'sauce': 'Bearnaise'} However, if the .__new__() method does not return an instance of class cls, the constructor method .__init__() will not be called. This allows the class more control over how new instances are created and initialized. You can return an instance of an entirely different class if you like.

Attribute access control in new-style classes New-style classes give you more ways to control what happens when an instance's attribute is accessed. Here is the general procedure for access to attribute a of instance i, where C is the class of i. If the instance has a __getattribute__() special method (see ), execute that method, which must either return the attribute value or raise AttributeError. If the instance has a __slots__ attribute (see ), return the value of the slot with name a. If a does not match any of the slot names, or if the named slot has never been set to a value, raise AttributeError. If a is a key in i.__dict__, return the corresponding value. Search for attribute a in class C. If that fails, search all the parent classes of C all the way back to object. If all searches in the preceding step failed and the instance has a .__getattr__() special method, call that method. See ; please note the differences from . If all the above steps fail to produce a value, raise AttributeError.

Properties in new-style classes: Fine-grained attribute access control The outline of attribute access in is slightly oversimplified in one respect. Any of the attribute search steps in this procedure may produce a property rather than the actual attribute value. A property is a special object that is produced by the property() function. For a discussion of the three types of attribute access (get, set, and delete), the protocols for the accessor functions, and examples, see .

Conserving memory with <code >__slots__</code > Normally, you can add new attributes to an instance's namespace with any name you want. The instance's .__dict__ attribute is effectively a dictionary, and you can add any number of names to it. However, in a new-style class, you may specify a given, limited set of attribute names that are allowed in instances of the class. There are two reasons why you might want to do this: If your program is going to create large numbers of instances of a class, to the point where you may run out of memory, you can save some storage within each instance by sacrificing the ability to add arbitrary attribute names. If you limit the set of permissible attribute names, Python will detect any reference to a name not in the permissible set, and raise an AttributeError exception. This may help you catch certain programming errors. To limit the set of attribute names in a new-style class, assign to a class variable named __slots__ a tuple containing the allowable names, like this: __slots__ = (n1, n2, ...) Here's a small example. Suppose you want instances of class Point to contain nothing more than two attributes named .x and .y: >>> class Point(object): ... __slots__ = ('x', 'y') ... def __init__(self, abscissa, ordinate): ... self.x, self.y = abscissa, ordinate ... >>> x2=Point(3, 7) >>> x2.x 3 >>> x2.y 7 >>> x2.temperature = 98.6 Traceback (most recent call last): File "<stdin>", line 1, in <module> AttributeError: 'Point' object has no attribute 'temperature' When you declare a __slots__ attribute in a new-style class, instances will not have a .__dict__ attribute.

Special method names Within a class, a number of reserved method names have special meaning. Here is a list of the ones covered in this document. __abs__ __add__ __and__ __call__ __cmp__ __complex__ __contains__ __del__ __delattr__ __delitem__ __divmod__ __div__ __divmod__ __enter__ __exit__ __eq__ __floordiv__ __float__ __ge__ __getattr__ __getattribute__ __getitem__ __gt__ __hex__ __iadd__ __iand__ __idiv__ __ifloordiv__ __ilshift__ __imod__ __imul__ __init__ __int__ __invert__ __ior__ __ipow__ __irshift__ __isub__ __iter__ __ixor__ __le__ __len__ __long__ __lshift__ __lt__ __mod__ __mul__ __ne__ __neg__ __new__ __nonzero__ __oct__ __or__ __pos__ __pow__ __radd__ __rand__ __rdiv__ __repr__ __reversed__ __rfloordiv__ __rlshift__ __rmod__ __rmul__ __ror__ __rpow__ __rrshift__ __rshift__ __rsub__ __rxor__ __setattr__ __setitem__ __str__ __sub__ __unicode__ __xor__ The most important special method name is the class constructor, .__init__(). For a general introduction to class constructors, see . For old-style class constructors, see . For new-style class constructors, see . Many special methods fall into broad categories: : for methods that implement comparisons such as “<=” and “==”. : for operators that operate on two operands, such as “%”. : for operators that operate on a single operand, such as negation, “-”. : for classes that handle calls to built-in functions such as “str()”.

Rich comparison methods These special methods allow your class to specify what happens when comparison operators such as “<=” are used, and the left-hand operator is an instance of your class. (In most cases the right-hand operator is also an instance of the same class, but this is not actually required.) In each case, the calling sequence for the method must look like this: def __method__(self, other): ... The self argument is the left-hand operand and the other argument is the operand on the right hand of the operator. Each method must return a numeric value: A negative number indicates that self precedes other. Zero indicates that self and other are considered equal. A positive number indicates that other precedes self. If the method does not implement the operation, it may return the special value NotImplemented. Operator Method name == __eq__ >= __ge__ > __gt__ <= __le__ < __lt__ != __ne__

Special methods for binary operators Your class can define special methods with these names to tell Python how to handle binary operators such as “*” or “%”. In each case, the calling sequence will look something like this: def __method__(self, other): ... The self argument is the left-hand operand, and the other argument is the right-hand operand. Your method will return the result of the operation. For each operator, you may supply up to three methods: The method in the first column performs the normal operation. The method in the second column is used when the left-hand operand does not support the given operation and the operands have different types. In these methods, self is the right-hand operand and other is the left-hand operand. The third column implements the “augmented assignment” operators such as “+=”. For example, for method __iadd__(self, other), the method must perform the equivalent of “

self +=
              other

”. Operator Normal Reversed Augmented + __add__ __radd__ __iadd__ & __and__ __rand__ __iand__ / __div__ __rdiv__ __idiv__ // __floordiv__ __rfloordiv__ __ifloordiv__ << __lshift__ __rlshift__ __ilshift__ % __mod__ __rmod__ __imod__ * __mul__ __rmul__ __imul__ | __or__ __ror__ __ior__ ** __pow__ __rpow__ __ipow__ >> __rshift__ __rrshift__ __irshift__ - __sub__ __rsub__ __isub__ ^ __xor__ __rxor__ __ixor__

Unary operator special methods You can define these special methods in your class to specify what happens when a unary operator such as “-” (negate) is applied to an instance of the class. In each case, the definition will look like this: def __method__(self): ... The method returns the result of the operation. Operator Method ~ __invert__ - __neg__ + __pos__

Special methods to emulate built-in functions You can define special methods that will handle calls to some of Python's built-in functions. The number of arguments will be the same as for the built-in functions, except that self is always the first argument. For example, a special method to handle calls to function

divmod(x,
          y)

will look like this: def __divmod__(self, other): ... In this method, the value of the first argument will be passed to self and the second argument to other. Function Method abs __abs__ complex __complex__ divmod __divmod__ hex __hex__ int __int__ len __len__ long __long__ mod __mod__ oct __oct__ str __str__ unicode __unicode__

<code >__call__()</code >: What to do when someone calls an instance If a class has a .__call__() method, its instances can be called as if they were functions. Any arguments passed in that function call become arguments to the .__call__() method. Example: >>> class CallMe(object): ... def __call__(self, *p, **k): ... print "CallMe instance called with:" ... print "Positional arguments", p ... print "Keyword arguments", k ... >>> c=CallMe() >>> c(1, 'rabbit', fetchez='la vache', hamster='elderberries') CallMe instance called with: Positional arguments (1, 'rabbit') Keyword arguments {'fetchez': 'la vache', 'hamster': 'elderberries'}

<code >__cmp__()</code >: Generalized comparison The purpose of this special method is to implement comparison operations between instances. It will be called in these situations: If the built-in cmp() function is called to compare an instance of a class to some other value, and the class has a .__cmp__() method, that method is called to perform the comparison. When an instance appears on the left-hand side of a comparison (relational) operator, and that instance's class has a the corresponding rich-comparison method (such as .__eq__() for the “==” operator; see ), the rich-comparison method will be called to perform the comparison. and, if so, that method is called. The comparison operators are “<”, “<=”, “==”, “!=”, “>”, and “>=”. However, if the class does not have the correct rich-comparison method, but it does have a .__cmp__() method, that method will be called to evaluate the comparison. The calling sequence is: def __cmp__(self, other): ... The convention for return values is the same one described in : negative if self precedes other, positive if other precedes self, zero if they are considered equal.

<code >__contains__()</code >: The “<code >in</code >” and “<code >not in</code >” operators This special method defines how instances of a class behave when they appear as the right-hand operand of Python's “in” and “not in” operators. Here is the calling sequence: def __contains__(self, x): ... The method returns True if x is considered to be in self, False otherwise.

<code >__del__()</code >: Destructor If a class has a .__del__() method, that method is called when an instance is deleted. For details, see .

<code >__delattr__()</code >: Delete an attribute If a class has a .__delattr__() method, that method is called when an attribute is deleted. Here is the calling sequence: def __delattr__(self, name): ... The name argument is the name of the attribute to be deleted.

<code >__delitem__()</code >: Delete one item of a sequence This method defines the behavior of a del statement of this form: del s[i] Such a statement can be used either on objects that act like sequences, where i specifies the position of the element to be deleted, or mapping objects (that is, dictionary-like objects), where i is the key of the key-value pair to be deleted. The calling sequence is: def __delitem__(self, i): ...

<code >__enter__</code >: Context manager initialization For a general explanation of context managers, see . A class that acts a content manager must provide this special method as well as the one described in . The expression that follows the word with in the with statement must evaluate to a context manager, whose .__enter__() method is called with no arguments (other than self). The value it returns will be bound to the variable named in the with statement's as clause, if one was provided.

<code >__exit__</code >: Context manager cleanup For a general explanation of context managers, see . A class that acts a content manager must provide this special method as well as the one described in . When the body of a with statement completes its execution, the .__exit__() method of the related context manager M is called with three arguments. If the block terminated without raising an exception, all three arguments will be None; otherwise see below. M.__exit__(self, eType, eValue, eTrace) eType The type of the exception. eValue The exception instance raised. eTrace A traceback instance. For more information about stack traces, see the documentation for the traceback module. Your .__exit__() method's return value determines what happens next if the block raised an exception. If it returns True, Python ignores the exception and proceeds with execution at a point just after the with block. If you don't want your context manager to suppress the exception, don't re-raise it explicitly, just return False and Python will then re-raise the exception.

<code >__format__</code >: Implement the <code >format()</code > function Use this special method to determine how an instance of your class acts when passed to the function described in . The calling sequence is: def __format__(self, fmt): ... The interpretation of the fmt argument is entirely up to you. The return value should be a string representation of the instance. If the call to the format() function does not provide a second argument, the fmt value passed to this method will be an empty string.

<code >__getattr__()</code >: Handle a reference to an unknown attribute This method, if present, handles statements that get an attribute value of an instance, but there is no such entry in the instance's namespace. The calling sequence is: def __getattr__(self, name): ... The argument is the name of the desired attribute. The method must either return the attribute's value or raise an AttributeError exception. Compare , which is called even if the instance namespace does have an attribute with the desired name.

<code >__getattribute__()</code >: Intercept all attribute references This method is called whenever an attribute is referenced, whether the instance or class namespace has an attribute with the given name or not. It works only with new-style classes. For an overview and examples, see .

<code >__getitem__()</code >: Get one item from a sequence or mapping If a class defines it, this special method is called whenever a value is retrieved from a sequence or mapping (dictionary-like object) using the syntax “v[i]”, where v is the sequence or mapping and i is a position in a sequence, or a key in a mapping. Here is the calling sequence: def __getitem__(self, i): ... The method either returns the corresponding item or raises an appropriate exception: IndexError for sequences or KeyError for mappings.

<code >__iter__()</code >: Create an iterator If a class defines an .__iter__() method, that method is called: Whenever the built-in iter() function is applied to an instance of the class. In any situation where the instance is iterated over, such as when it appears as the controlling iterable of a for statement. The calling sequence is: def __iter__(self): ... The return value must be an iterator. An iterator is any object that has a .next() method that returns the next value in the sequence, or raises StopIteration when the sequence is exhausted. For more information, see .

<code >__nonzero__()</code >: True/false evaluation If a class defines it, this special method is called whenever an instance is converted to a Boolean value, either implicitly (for example, when it is the test in an “if” statement) or explicitly via the built-in bool() function. Here is the calling sequence: def __nonzero__(self): ... Return a Boolean value, either True or False.

<code >__repr__()</code >: String representation If a class defines it, this special method is called to find the “representation” of an object. There are two ways to get a representation: Via the function described in . Via the “back-quote operator”, enclosing a Python expression in open single quotes. The calling sequence is: def __repr__(self): ... The method returns the representation of self as a string.

<code >__reversed__()</code >: Implement the <code >reversed()</code > function If provided, this method allows the reversed() function to be applied to an instance of the class. It returns an iterator that iterates over the contained elements in reverse order.

<code >__setattr__()</code >: Intercept all attribute changes If a class defines it, this method is called whenever a new value is stored into an attribute. Calling sequence: def __setattr__(self, name, value): ... The method sets the attribute given by the name argument to the value argument, or raises AttributeError if that operation is not permitted.

<code >__setitem__()</code >: Assign a value to one item of a sequence If a class defines it, this method is called whenever a new value is stored into a sequence or mapping (dictionary-like object), such as in statements of this form: V[i] = expr Here is the calling sequence: def __setitem__(self, i, value): ... For sequence-type objects, the i argument specifies the position in the sequence to be modified. For mappings, the i argument is the key under which the value is to be stored.

Static methods A static method of a Python class is a method that does not obey the usual convention in which self, an instance of the class, is the first argument to the method. To declare a static method, declare the method normally, and wrap it with the built-in staticmethod function, using either of the techniques described in . Once you have declared a method to be static, the arguments you pass it are exactly the arguments it receives. Example: >>> class Hobbs: ... @staticmethod ... def represent(): ... print "Hobbs represent!" ... >>> Hobbs.represent() Hobbs represent!

Class methods A class method is similar to a static method (see ), except that its first argument is always the class that contains the method. To declare a class method, use a declaration of this general form: @classmethod def methodName(cls, ...): ... When this method is called, the first argument (cls) will be the class containing methodName. There are two ways to call a class method: using its class C, or an instance i. These two general forms are: C.methodName(...) i.methodName(...) In the first case, the class C is passed as the cls argument of the method. In the second case, the class of instance i is passed as the cls argument. >>> class Jal(object): ... @classmethod ... def whatClass(cls, n): ... print "cls={0} n={1}".format(cls, n) ... def __init__(self, color): ... self.color = color ... >>> Jal.whatClass(5) cls=<class '__main__.Jal'> n=5 >>> eunice=Jal('green') >>> eunice.whatClass(17) cls=<class '__main__.Jal'> n=17

<code >pdb</code >: The Python interactive debugger The Python debugger allows you to monitor and control execution of a Python program, to examine the values of variables during execution, and to examine the state of a program after abnormal termination (post-mortem analysis). . . .

Starting up <code >pdb</code > To execute any Python statement under the control of pdb: >>> import pdb >>> import yourModule >>> pdb.run('yourModule.test()') # Or any other statement where yourModule.py contains the source code you want to debug. To debug a Python script named myscript.py: python /usr/lib/python2.5/pdb.py myscript.py To perform post-mortem analysis: >>> import pdb >>> import yourModule >>> yourModule.test() [crash traceback appears here] >>> pdb.pm() (pdb) Then you can type debugger commands at the (pdb) prompt.

Functions exported by <code >pdb</code > The pdb module exports these functions: pdb.run(stmt[,globals[,locals]]) Executes any Python statement. You must provide the statement as a string. The debugger prompts you before beginning execution. You can provide your own global name space by providing a globals argument; this must be a dictionary mapping global names to values. Similarly, you can provide a local namespace by passing a dictionary locals. pdb.runeval(expr[,globals[,locals]]) This is similar to the pdb run command, but it evaluates an expression, rather than a statement. The expr is any Python expression in string form. pdb.runcall(func[,arg]...) Calls a function under pdb control. The func must be either a function or a method. Arguments after the first argument are passed to your function.

Commands available in <code >pdb</code > The debugger prompts with a line like this: (pdb) At this prompt, you can type any of the pdb commands discussed below. You can abbreviate any command by omitting the characters in square brackets. For example, the where command can be abbreviated as simply w. expr Evaluate an expression expr and print its value. !stmt Execute a Python statement stmt. The “!” may be omitted if the statement does not resemble a pdb command. (empty line) If you press Enter at the (pdb) prompt, the previous command is repeated. The list command is an exception: an empty line entered after a list command shows you the next 11 lines after the ones previously listed. a[rgs] Display the argument names and values to the currently executing function. b[reak] [[filename:]lineno[,condition]] The break command sets a breakpoint at some location in your program. If execution reaches a breakpoint, execution will be suspended and you will get back to the (pdb) prompt. This form of the command sets a breakpoint at a specific line in a source file. Specify the line number within your source file as lineno; add the filename: if you are working with multiple source files, or if your source file hasn't been loaded yet. You can also specify a conditional breakpoint, that is, one that interrupts execution only if a given condition evaluates as true. For example, the command break 92,i>5 would break at line 92 only when i is greater than 5. When you set a breakpoint, pdb prints a “breakpoint number.” You will need to know this number to clear the breakpoint. b[reak] [function[,condition]] This form of the break command sets a breakpoint on the first executable statement of the given function. c[ont[inue]] Resume execution until the next breakpoint (if any). cl[ear] [lineno] If used without an argument, clears all breakpoints. To clear one breakpoint, give its breakpoint number (see break above). h[elp] [cmd] Without an argument, prints a list of valid commands. Use the cmd argument to get help on command cmd. l[ist] [begin[,end]] Displays your Python source code. With no arguments, it shows 11 lines centered around the current point of execution. The line about to be executed is marked with an arrow (->), and the letter B appears at the beginning of lines with breakpoints set. To look at a given range of source lines, use the begin argument to list 11 lines around that line number, or provide the ending line number as an end argument. For example, the command list 50,65 would list lines 50-65. n[ext] Like step, but does not stop upon entry to a called function. q[uit] Exit pdb. r[eturn] Resume execution until the current function returns. s[tep] Single step: execute the current line. If any functions are called in the current line, pdb will break upon entering the function. tbreak Same options and behavior as break, but the breakpoint is temporary, that is, it is removed after the first time it is hit. w[here] Shows your current location in the program as a stack traceback, with an arrow (->) pointing to the current stack frame.

Commonly used modules The sections below discuss only a tiny fraction of the official and unofficial module library of Python. For a full set, consult the Python Library Reference. If you want to use any of these modules, you must import them. See .

<code >math</code >: Common mathematical operations This module provides the usual basic transcendental mathematical functions. All trig functions use angles in radians. (For a similar set of functions using the complex number system, see the Python library documentation for the cmath module.) The math module has two constant attributes: pi The constant 3.14159... e The base of natural logarithms, 2.71828... Functions in this module include: acos(x) Angle (in radians) whose cosine is x, that is, arccosine of x. acosh(x) Inverse hyperbolic cosine of x asin(x) Arcsine of x. asinh(x) Inverse hyperbolic sine of x atan(x) Arctangent of x. atanh(x) Inverse hyperbolic tangent of x atan2(y,x) Angle whose slope is y/x, even if x is zero. ceil(x) True ceiling function, defined as the smallest integer that is greater than or equal to x. For example, ceil(3.9) yields 4.0, while ceil(-3.9) yields -3.0. cos(x) Cosine of x, where x is expressed in radians. cosh(x) Hyperbolic cosine of x. degrees(x For x in radians, returns that angle in degrees. erf(x) Error function. erfc(x) Error function complement. exp(x) e to the x power. fabs(x Returns the absolute value of x as a float value. factorial(n) Returns the factorial of n, which must be a nonnegative integer. floor(x) True floor function, defined as the largest integer that is less than or equal to x. For example, floor(3.9) is 3.0, and floor(-3.9) is -4.0. fmod(x,y) Returns (x-int(x/y)*y). frexp(x) For a float value x, returns a tuple (m, e) where m is the mantissa and e is the exponent. For x=0.0, it returns (0.0, 0); otherwise, abs(m) is a float in the half-open interval [0.5, 1) and e is an integer, such that x == m*2**e. gamma(x) Gamma function. hypot(x,y) The square root of (x2+y2).

ldexp(x,
                i)

Returns

x *
                (2**i)

. This is the inverse of frexp(). lgamma(x) Natural log of abs(gamma(x)).

log(x[,
                b)

With one argument, returns the natural log of x. With the second argument, returns the log of x to the base b. log10(x) Common log (base 10) of x. modf(x) Returns a tuple (f, i) where f is the fractional part of x, i is the integral part (as a float), and both have the same sign as x. radians(x) For x in degrees, returns that angle in radians. sin(x) Sine of x. sinh(x) Hyperbolic sine of x. sqrt(x) Square root of x. tan(x) Tangent of x. tanh(x) Hyperbolic tangent of x.

<code >string</code >: Utility functions for strings Variables and functions for working with character strings. ascii_letters A string containing all the letters from ascii_uppercase and ascii_lowercase. ascii_lowercase The characters that are considered lowercase letters in the ASCII character set, namely: "abcdefghijklmnopqrstuvwxyz" ascii_uppercase "ABCDEFGHIJKLMNOPQRSTUVWXYZ" The value of this constant does not depend on the locale setting; see . digits The decimal digits: "0123456789". hexdigits The hexadecimal digits: "0123456789abcdefABCDEF". letters A string containing all the characters that are considered letters in the current locale. lowercase A string containing all the characters that are considered lowercase letters in the current locale. maketrans(s, t) Builds a translation table to be used as the first argument to the .translate() string method. The arguments s and t are two strings of the same length; the result is a translation table that will convert each character of s to the corresponding character of t. >>> import string >>> bingo=string.maketrans("lLrR", "rRlL") >>> "Cornwall Llanfair".translate(bingo) 'Colnwarr Rranfail' >>> "Cornwall Llanfair".translate(bingo, 'ai') 'Colnwrr Rrnfl' octdigits The octal digits: "01234567". printable A string containing all the printable characters. punctuation All printable characters that are not letters or digits in the current locale. If ASCII is the locale's current encoding, these characters are included: !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~ uppercase A string containing all the characters that are considered uppercase letters in the current locale. May not be the same as ascii_uppercase whitespace A string containing all the characters considered to be white space in the current locale. For ASCII this will be: "\t\n\x0b\x0c\r " See .

<code >random</code >: Random number generation This module provides for the generation of pseudorandom numbers.

John von Neumann Any one who considers arithmetical methods of producing random digits is, of course, in a state of sin. For, as has been pointed out several times, there is no such thing as a random number—there are only methods to produce random numbers, and a strict arithmetic procedure of course is not such a method.

choice(S) Returns a randomly selected element from an iterable S. >>> for i in range(9): print choice ( ['s', 'h', 'd', 'c'] ), ... s c c h c s s h d normalvariate(m,s) Generate a normally distributed pseudorandom number with mean m and standard deviation s. randint(x,y) Returns a random integer in the closed interval [x,y]; that is, any result r will satisfy x <= r <= y. >>> for i in range(20): print randint(1,6), ... 3 4 6 3 2 1 1 2 1 2 1 2 3 3 3 5 6 4 4 2 random() Returns a random float in the half-open interval [0.0, 1.0); that is, for any result r, 0.0 <= r < 1.0. >>> for count in range(48): ... print "{0:.3f}".format(random()), ... if (count % 12) == 11: print ... 0.012 0.750 0.899 0.339 0.371 0.561 0.358 0.931 0.822 0.990 0.682 0.847 0.245 0.541 0.992 0.151 0.394 0.335 0.702 0.885 0.986 0.350 0.417 0.748 0.918 0.103 0.109 0.328 0.423 0.180 0.203 0.689 0.600 0.794 0.201 0.008 0.564 0.920 0.906 0.469 0.510 0.818 0.142 0.589 0.590 0.290 0.650 0.889 randrange([start,]stop[,step]) Return a random element from the sequence range(start,stop,step) . >>> from random import * >>> for i in range(35): print randrange(4), ... 0 2 2 2 1 1 2 3 1 3 3 2 2 2 3 0 2 0 0 1 2 0 2 1 1 1 2 2 2 1 1 3 1 1 2 >>> for i in range(35): print randrange(1,5), ... 3 3 2 1 1 1 4 4 3 2 1 1 3 2 1 2 4 4 1 4 2 4 4 1 1 1 1 1 4 4 1 1 2 2 1 >>> range(2,18,5) [2, 7, 12, 17] >>> for i in range(28): print randrange(2,18,5), ... 12 2 7 2 17 17 7 7 12 17 17 2 7 17 12 7 7 12 17 17 7 12 7 7 7 7 7 7 shuffle(L) Randomly permute the elements of a sequence L. Here's an example. First we build a (small) deck of cards, using a list comprehension to build a list of all possible combinations of three ranks (ace, king, queen) and four suits (spades, hearts, diamonds, and clubs). Then we shuffle the deck twice and inspect the results. >>> ranks = 'AKQ' >>> suits = 'shdc' >>> deck = [ r+s ... for s in suits ... for r in ranks ] >>> deck ['As', 'Ks', 'Qs', 'Ah', 'Kh', 'Qh', 'Ad', 'Kd', 'Qd', 'Ac', 'Kc', 'Qc'] >>> shuffle(deck) >>> deck ['Qh', 'Ks', 'Kh', 'As', 'Kc', 'Kd', 'Qd', 'Qc', 'Ah', 'Ad', 'Qs', 'Ac'] >>> shuffle(deck) >>> deck ['As', 'Qs', 'Ks', 'Kc', 'Ad', 'Kh', 'Qh', 'Ac', 'Ah', 'Qc', 'Qd', 'Kd'] uniform(x,y) Returns a random float in the half-open interval [x,y); that is, each result r will satisfy

x <=
              r < y

. An assortment of other pseudorandom distributions is available. See the Python Library Reference for details.

<code >time</code >: Clock and calendar functions This module provides minimal time and date functions. For a newer and more complete module, see the datetime module. Epoch time is the time in seconds since some arbitrary starting point. For example, Unix measures time in seconds since January 1, 1970. UTC is Coordinated Universal Time, the time on the zero meridian (which goes through London). DST refers to Daylight Savings Time. A time tuple is a 9-tuple T with these elements, all integers: T[0] Four-digit year. T[5] Second, in [0,59]. T[1] Month, 1 for January, 12 for December. T[6] Day of week, 0 for Monday, 6 for Sunday. T[2] Day of month, in [1,31]. T[7] Ordinal day of the year, in [1,366]. T[3] Hour, in [0,23]. T[8] DST flag: 1 if the time is DST, 0 if it is not DST, and -1 if unknown. T[4] Minute, in [0,59]. Contents of the time module: altzone The local DST offset, in seconds west of UTC (negative for east of UTC). asctime([T]) For a time-tuple T, returns a string of exactly 24 characters with this format: "Thu Jun 12 15:25:31 1997" The default time is now. clock() The accumulated CPU time of the current process in seconds, as a float. ctime([E]) Converts an epoch time E to a string with the same format as asctime(). The default time is now. daylight Nonzero if there is a DST value defined locally. gmtime([E]) Returns the time-tuple corresponding to UTC at epoch time E; the DST flag will be zero. The default time is now. localtime([E]) Returns the time-tuple corresponding to local time at epoch time E. The default time is now. mktime(T) Converts a time-tuple T to epoch time as a float, where T is the local time. sleep(s) Suspend execution of the current process for s seconds, where s can be a float or integer. strftime(f[,t]) Time formatting function; formats a time-tuple t according to format string f. The default time t is now. As with the old string format operator, format codes start with %, and other text appears unchanged in the result. See the table of codes below. time() The current epoch time, as a float. timezone The local non-DST time zone as an offset in seconds west of UTC (negative for east of UTC). This value applies when daylight savings time is not in effect. tzname A 2-tuple (s, d) where s is the name of the non-DST time zone locally and d is the name of the local DST time zone. For example, in Socorro, NM, you get ('MST', 'MDT'). Format codes for the strftime function include: %a Abbreviated weekday name, e.g., 'Tue'. %A Full weekday name, e.g., 'Tuesday'. %b Abbreviated month name, e.g., 'Jul'. %B Full month name, e.g., 'July'. %d Day of the month, two digits with left zero fill; e.g. '03'. %H Hour on the 24-hour clock, two digits with zero fill. %I Hour on the 12-hour clock, two digits with zero fill. %j Day of the year as a decimal number, three digits with zero fill, e.g. '366'. %m Decimal month, two digits with zero fill. %M Minute, two digits with zero fill. %p Either 'AM' or 'PM'. Midnight is considered AM and noon PM. %S Second, two digits with zero fill. %w Numeric weekday: 0 for Sunday, 6 for Saturday. %y Two-digit year. Not recommended! %Y Four-digit year. %Z If there is a time zone, a string representing that zone; e.g., 'PST'. %% Outputs the character %.

<code >re</code >: Regular expression pattern-matching The re module provides functions for matching strings against regular expressions. See the O'Reilly book Mastering Regular Expressions by Friedl and Oram for the whys and hows of regular expressions. We discuss only the most common functions here. Refer to the Python Library Reference for the full feature set. . . . .

Characters in regular expressions Note: The raw string notation r'...' is most useful for regular expressions; see raw strings, above. These characters have special meanings in regular expressions: . Matches any character except a newline. ^ Matches the start of the string. $ Matches the end of the string. r* Matches zero or more repetitions of regular expression r. r+ Matches one or more repetitions of r. r? Matches zero or one r. r*? Non-greedy form of r*; matches as few characters as possible. The normal * operator is greedy: it matches as much text as possible. r+? Non-greedy form of r+. r?? Non-greedy form of r?. r{m,n} Matches from m to n repetitions of r. For example, r'x{3,5}' matches between three and five copies of letter 'x'; r'(bl){4}' matches the string 'blblblbl'. r{m,n}? Non-greedy version of the previous form. [...] Matches one character from a set of characters. You can put all the allowable characters inside the brackets, or use a-b to mean all characters from a to b inclusive. For example, regular expression r'[abc]' will match either 'a', 'b', or 'c'. Pattern r'[0-9a-zA-Z]' will match any single letter or digit. [^...] Matches any character not in the given set. rs Matches expression r followed by expression s. r|s Matches either r or s. (r) Matches r and forms it into a group that can be retrieved separately after a match; see MatchObject, below. Groups are numbered starting from 1. (?:r) Matches r but does not form a group for later retrieval. (?P<n>r) Matches r and forms it into a named group, with name n, for later retrieval. (?P=n) Matches whatever string matched an earlier (?P<n>r) group. (?#...) Comment: the “...” portion is ignored and may contain a comment. (?=...) The “...” portion must be matched, but is not consumed by the match. This is sometimes called a lookahead match. For example, r'a(?=bcd)' matches the string 'abcd' but not the string 'abcxyz'. Compared to using r'abcd' as the regular expression, the difference is that in this case the matched portion would be 'a' and not 'abcd'. (?!...) This is similar to the (?=...): it specifies a regular expression that must not match, but does not consume any characters. For example, r'a(?!bcd)' would match 'axyz', and return 'a' as the matched portion; but it would not match 'abcdef'. You could call it a negative lookahead match. The special sequences in the table below are recognized. However, many of them function in ways that depend on the locale; see . For example, the r'\s' sequence matches characters that are considered whitespace in the current locale. \n Matches the same text as a group that matched earlier, where n is the number of that group. For example, r'([a-zA-Z]+):\1' matches the string "foo:foo". \A Matches only at the start of the string. \b Matches the empty string but only at the start or end of a word (where a word is set off by whitespace or a non-alphanumeric character). For example, r'foo\b' would match "foo" but not "foot". \B Matches the empty string when not at the start or end of a word. \d Matches any digit. \D Matches any non-digit. \s Matches any whitespace character. \S Matches any non-whitespace character. \w Matches any alphanumeric character plus the underbar '_'. \W Matches any non-alphanumeric character. \Z Matches only at the end of the string. \\ Matches a backslash (\) character.

Functions in the <code >re</code > module There are two ways to match regular expressions with the re module. Assuming you import the module with import re, you can test whether a regular expression r matches a string s with the construct: re.match(r,s) However, if you will be matching the same regular expression many times, the performance will be better if you compile the regular expression like this: re.compile(r) The re.compile() function returns a compiled regular expression object. You can then check a string s for matching by using the .match(s) method on that object. Here are the functions in module re: compile(r[,f]) Compile regular expression r. This function returns a compiled regular expression object; see . To get case-insensitive matching, use re.I as the f argument. There are other flags that may be passed to the f argument; see the Python Library Reference. match(r,s[,f]) If r matches the start of string s, return a MatchObject (see below), otherwise return None. search(r,s[,f]) Like the match() method, but matches r anywhere in s, not just at the beginning. split(r,s[,maxsplit=m]) Splits string s into pieces where pattern r occurs. If r does not contain groups, returns a list of the parts of s that match r, in order. If r contains groups, returns a list containing all the characters from s, with parts matching r in separate elements from the non-matching parts. If the m argument is given, it specifies the maximum number of pieces that will be split, and the leftovers will be returned as an extra string at the end of the list. sub(r,R,s[,count=c]) Replace the leftmost non-overlapping parts of s that match r using R; returns s if there is no match. The R argument can be a string or a function that takes one MatchObject argument and returns the string to be substituted. If the c argument is supplied (defaulting to 0), no more than c replacements are done, where a value of 0 means do them all.

Compiled regular expression objects Compiled regular expression objects returned by re.compile() have these methods: .match(s[,[ps][,pe]]) If the start of string s matches, return a MatchObject; if there is no match, return None. If ps is given, it specifies the index within s where matching is to start; this defaults to 0. If pe is given, it specifies the maximum length of s that can be used in matching. .pattern The string from which this object was compiled. .search(s[,[ps][,pe]]) Like match(), but matches anywhere in s. .split(s[,maxsplit=m]) Like re.split(). .sub(R,s[,count=c]) Like re.sub().

Methods on a <code >MatchObject</code > A MatchObject is the object returned by .match() or other methods. Such an object has these methods and attributes: .end([n]) Returns the location where a match ended. If no argument is given, returns the index of the first character past the match. If n is given, returns the index of the first character past where the nth group matched. .endpos The effective pe value passed to .match() or .search(). .group([n]) Retrieves the text that matched. If there are no arguments, or if n is zero, it returns the entire string that matched. To retrieve just the text that matched the nth group, pass in an integer n, where the groups are numbered starting at 1. For example, for a MatchObject m, m.group(2) would return the text that matched the second group, or None if there were no second group. If you have named the groups in your regular expression using a construct of the form (?P<name>...), the n argument can be the name as a string. For example, if you have a group (?P<year>[\d]{4}) (which matches four digits), you can retrieve that field using m.group('year'). .groups([default]) Return a tuple (s1,s2,...) containing all the matched strings, where si is the string that matched the ith group. For groups that did not match, the corresponding value in the tuple will be None, or an optional default value that you specify in the call to this method. .groupdict([default]) Return a dictionary whose keys are the named groups in the regular expression. Each corresponding value will be the text that matched the group. If a group did not match, the corresponding value will be None, or an alternate default value that you supply when you call the method. .lastgroup Holds the name of the last named group (using the (?P<n>r) construct) that matched. It will be None if no named groups matched, or if the last group that matched was a numbered group and not a named group. .lastindex Holds the index of the last group that matched, or None if no groups matched. .pos The effective ps value passed to .match() or .search(). .re The regular expression object used to produce this MatchObject. .span([n]) Returns a 2-tuple (m.start(n),m.end(n)). .start([n]) Returns the location where a match started. If no argument is given, returns the index within the string where the entire match started. If an argument n is given, returns the index of the start of the match for the nth group. .string The s argument passed to .match() or .search().

<code >sys</code >: Universal system interface The services in this module give you access to command line arguments, standard input and output streams, and other system-related facilities. argv sys.argv[0] is the name of your Python script, or '-c' if in interactive mode. The remaining elements, sys.argv[1:], are the command line arguments, if any. builtin_module_names A list of the names of the modules compiled into your installation of Python. exit(n) Terminate execution with status n. modules A dictionary of the modules already loaded. path The search path for modules, a list of strings in search order. Note: You can modify this list. For example, if you want Python to search directory /u/dora/python/lib for modules to import before searching any other directory, these two lines will do it: import sys sys.path.insert(0, "/u/dora/python/lib") platform A string identifying the software architecture. stderr The standard error stream as a file object. stdin The standard input stream as a file object. stdout The standard output stream as a file object.

<code >os</code >: The operating system interface The variables and methods in the os module allow you to interact with files and directories. In most cases the names and functionalities are the same as the equivalent C language functions, so refer to Kernighan and Ritchie, The C Programming Language, second edition, or the equivalent for more details. chdir(p) Change the current working directory to that given by string p chmod(p,m) Change the permissions for pathname p to m. See module stat, below, for symbolic constants to be used in making up m values. chown(p,u,g) Change the owner of pathname p to user id u and group id g. environ A dictionary whose keys are the names of all currently defined environmental variables, and whose values are the values of those variables. error The exception raised for errors in this module. execv(p,A) Replace the current process with a new process executing the file at pathname p, where A is a list of the strings to be passed to the new process as command line arguments. execve(p,A,E) Like execv(), but you supply a dictionary E that defines the environmental variables for the new process. _exit(n) Exit the current process and return status code n. This method should be used only by the child process after a fork(); normally you should use sys.exit(). fork() Fork a child process. In the child process, this function returns 0; in the parent, it returns the child's process ID. getcwd() Returns the current working directory name as a string. getegid() Returns the effective group ID. geteuid() Returns the effective user ID. getgid() Returns the current process's group ID. To decode user IDs, see the grp standard module. getpid() Returns the current process's process ID. getppid() Returns the parent process's PID (process ID). getuid() Returns the current process's user ID. To decode user IDs, see the pwd standard module. kill(p,s) Send signal s to the process whose process ID is p. link(s,d) Create a hard link to s and call the link d. listdir(p) Return a list of the names of the files in the directory whose pathname is p. This list will never contain the special entries '.' and '..' for the current and parent directories. The entries may not be in any particular order. lstat(p) Like stat(), but if p is a link, you will get the status tuple for the link itself, rather than the file it points at.

makedirs(p[,
              mode])

Works like mkdir(), but will also create any intermediate directories between existing directories and the desired new directory. mkdir(p[,m]) Create a directory at pathname p. You may optionally specify permissions m; see module stat below for the interpretation of permission values. mkfifo(p,m) Create a named pipe with name p and open mode m. The server side of the pipe will open it for reading, and the client side for writing. This function does not actually open the fifo, it just creates the rendezvous point. nice(i) Renice (change the priority) of the current process by adding i to its current priority. readlink(p) If p is the pathname to a soft (symbolic) link, this function returns the pathname to which that link points. remove(p) Removes the file with pathname p, as in the Unix rm command. Raises OSError if it fails. removedirs(p) Similar to remove(), but also removes any other parent directory in the path that has no other children. rename(po, pn) Rename path po to pn. rmdir(p) Remove the directory at path p. stat(p) Return a status tuple describing the file or directory at pathname p. See module stat, below, for the interpretation of a status tuple. If p is a link, you will get the status tuple of the file to which p is linked. symlink(s,d) Create a symbolic link to path s, and call the link d. system(c) Execute the command in string c as a sub-shell. Returns the exit status of the process. times() Returns a tuple of statistics about the current process's elapsed time. This tuple has the form (u,s,u',s',r) where u is user time, s is system time, u' and s' are user and system time including all child processes, and r is elapsed real time. All values are in seconds as floats. tmpfile() Returns a new, open temporary file, with update mode "w+b". This file will not appear in any directory, and will disappear when it is no longer in use. umask(m) Sets the “umask” that determines the default permissions for newly created files. Returns the previous value. Each bit set in the umask corresponds to a permission that is not granted by default. uname() Returns a tuple of strings descriping the operating system's version: (s,n,r,v,m) where s is the name of the operating system, n is the name of the processor (node) where you are running, r is the operating system's version number, v is the major version, and m describes the type of processor. urandom(n) Return a string of n random bytes. These bytes should be sufficiently random for use in cryptographic applications. utime(p,t) The t argument must be a tuple (a, m) where a and m are epoch times. For pathname p, set the last access time to a and the last modification to m. wait() Wait for the termination of a child process. Returns a tuple (p,e) where p is the child's process ID and e is its exit status. waitpid(p,o) Like wait(), but it waits for the process whose ID is p. The option value o specifies what to do if the child is still running. If o is 0, you wait for the child to terminate. Use a value of os.WNOHANG if you don't want to wait. WNOHANG See waitpid() above.

<code >stat</code >: Interpretation of file status The stat module contains a number of variables used in encoding and decoding various items returned by certain methods in the os module, such as stat() and chmod(). First, there are constants for indexing the components of a “status tuple” such as that returned by os.stat(): ST_ATIME The epoch time of last access (see the time module for interpretation of times). ST_CTIME The epoch time of the file's last status change. ST_DEV The device number. ST_GID The group ID. ST_INO The i-node number. ST_MODE The file's permissions. ST_MTIME The epoch time of last modification. ST_NLINK The number of hard links. ST_SIZE The current size in bytes. ST_UID The user ID. The following functions are defined in the stat module for testing a mode value m, where m is the ST_MODE element of the status tuple. Each function is a predicate: S_ISBLK(m) Is this a block device? S_ISCHR(m) Is this a character device? S_ISDIR(m) Is this a directory? S_ISFIFO(m) Is this a FIFO? S_ISLNK(m) Is this a soft (symbolic) link? S_ISREG(m) Is this an ordinary file? S_ISSOCK(m) Is this a socket? These constants are defined for use as mask values in testing and assembling permission values such as those returned by os.stat() in . S_IRGRP Group read permission. S_IROTH World read permission. S_IRUSR Owner read permission. S_ISGID SGID (set group ID) bit. S_ISUID SUID (set user ID) bit. S_IWGRP Group write permission. S_IWOTH World write permission. S_IWUSR Owner write permission. S_IXGRP Group execute permission. S_IXOTH World execute permission. S_IXUSR Owner execute permission.

<code >os.path</code >: File and directory interface These functions allow you to deal with path names and directory trees. To use a given method in this module, import the os module and then use os.path.method(). For example, to get the base name of a path p, use os.path.basename(p). abspath(p) Return the absolute path name that is equivalent to path p. basename(p) Return the base name portion of a path name string p. See split(), below. commonprefix(L) For a list L containing pathname strings, return the longest string that is a prefix of each element in L. exists(p) Predicate for testing whether pathname p exists. expanduser(p) If p is a pathname starting with a tilde character (~), return the equivalent full pathname; otherwise return p. isabs(p) Predicate for testing whether p is an absolute pathname (e.g., starts with a slash on Unix systems). isfile(p) Predicate for testing whether p refers to a regular file, as opposed to a directory, link, or device. islink(p) Predicate for testing whether p is a soft (symbolic) link. ismount(p) Predicate for testing whether p is a mount point, that is, whether p is on a different device than its parent directory. join(p,q) If q is an absolute path, returns q. Otherwise, if p is empty or ends in a slash, returns p+q, but otherwise it returns p+'/'+q. normcase(p) Return pathname p with its case normalized. On Unix systems, this does nothing, but on Macs it lowercases p. samefile(p,q) Predicate for testing whether p and q are the same file (that is, the same inode on the same device). This method may raise an exception if os.stat() fails for either argument. split(p) Return a 2-tuple (H,T) where T is the tail end of the pathname (not containing a slash) and H is everything up to the tail. If p ends with a slash, returns (p,''). If p contains no slashes, returns ('',p). The returned H string will have its trailing slash removed unless H is the root directory. splitext(p) Returns a 2-tuple (R,E) where E is the “extension” part of the pathname and R is the “root” part. If p contains at least one period, E will contain the last period and everything after that, and R will be everything up to but not including the last period. If p contains no periods, returns (p,''). walk(p,V,a) Walks an entire directory structure starting at pathname p. See below for more information. The os.path.walk(p,V,a) function does the following for every directory at or below p (including p if p is a directory), this method calls the “visitor function” V with arguments V(a,d,N) a The same a passed to os.path.walk(). You can use a to provide information to the V() function, or to accumulate information throughout the traversal of the directory structure. d A string containing the name of the directory being visited. N A list of all the names within directory d. You can remove elements from this list in place if there are some elements of d that you don't want walk() to visit.

<code >argparse</code >: Processing command line arguments This module is the current recommended standard module for processing arguments from the command line. The module has many more features than those described here: refer to the full documentation.

Types of command line arguments For the purposes of this module, arguments are divided into two types: Positional arguments do not begin with a hyphen. When there are multiple positional arguments, the meaning of each one is inferred from their position relative to the other positional arguments. Optional arguments begin with a hyphen. In general each optional argument may have two forms, a short form consisting of a hyphen and a letter (e.g., “-h” for help), and a long form consisting of two hyphens and a full name (e.g., “--help”). Generally optional arguments are not required, but your program can make them required if you wish. Some optional arguments may allow or require a value. For example, let's suppose your script has an option that specifies the name of a river, and that the short form is “-r” and the long form is “--river”. On the command line, the user can specify the Nile in any of these four ways: -r Nile -rNile --river Nile --river=Nile Users may group multiple short-form options together. For example, if a script named sss has options “-a”, “-m”, “-p”, and “-s”, these two examples are valid and equivalent: sss -a -m -p -s sss -spma A short-form option that takes an argument may occur as part of such a group, but only if it is the last option in the group. Contrary to Unix practice, optional arguments may occur anywhere relative to the positional arguments. Also, the names of long-form optional arguments may be abbreviated if the abbreviation is unambiguous. For example, if a script has two long-form options --pratt and --polonius, option --pr will be accepted as an abbreviation for --pratt; but --p would not be acceptable because it is ambiguous.

Overall flow of argument processing Here is the general procedure for using the argparse module to check and process your command line arguments, which in most cases will come from sys.argv (see ). Create an argparse.ArgumentParser instance. See . For each command line argument, call the .add_argument() method of that instance once to define what that argument may be. See . Call the .parse_args() method of the instance. If there are any errors in the command line, this method will print a message summarizing the arguments, and then terminate execution. If the arguments are correct according to the rules that you have specified, all the values will be packaged into an object containing all your arguments as attributes, with default values filled in where necessary. See .

The <code >ArgumentParser()</code > constructor Here is the calling sequence to create an ArgumentParser instance P: import argparse P = argparse.ArgumentParser(**kw) Valid keyword arguments include: description A string describing the overall purpose of your script. This string will be displayed in the generated help message. add_help=True By default, the returned parser will implement options -h and --help to display the help message showing the valid option syntax. Use add_help=False if you don't want this behavior. prog By default, when the name of your script is displayed, it will be the name from sys.argv[0]. If you would like to display a different program name, pass that name as the value of this keyword argument. Here's an example. parser = argparse.ArgumentParser(prog="nile-source", description="Find the source of the Nile")

The <code >ArgumentParser.add_argument()</code > method To define each command line argument, call the .add_argument() method of an ArgumentParser instance P. There are two forms, depending on whether you are defining a positional command line argument or an optional argument. In either case, there are a number of keyword arguments denoted as “**kw”; we will describe these arguments below. To define a positional command line argument, use this form: P.add_argument(posName, **kw) The posName is a string that specifies the name of the argument (which cannot begin with “-”). For example, if your ArgumentParser instance is p, to define a positional argument called “inFile”, your call might begin like this: p.add_argument("inFile", ...) To define an optional command line argument, use this form: P.add_argument(s0, s1, ..., , **kw) Each si is a string defining the option name, starting with either "-" for short-form options or "--" for options with the long form. For example, if you have an option whose short form is “-x” and whose long form is “--exec”, your method call would begin: p.add_argument("-x", "--exec", ...) You can specify any number of short and long form options in this way. Here are the principle keyword arguments to the .add_argument() method. Some of these require information to be passed through other keyword arguments. dest Name of the attribute where the value of this argument will be stored in the result returned by the ArgumentParser.parse_args() method. If you don't specify this, the attribute name will be: For positional arguments, the attribute name will be the name passed as the first argument to .add_argument(). For optional arguments, the attribute name is the first long-form option name given if there is at least one; otherwise the attribute name is the first short-form option name given. For example, if the method call looks like

.add_argument('-x', '--exec', '--run',
                    ...)

, the value will be stored in the .exec attribute of the result returned by .arg_parse(). action Specifies what happens when this command line argument is processed. The value must be one of the following: action='store' Store the argument value as a string in the result returned by .arg_parse(). The name of the attribute where it is stored is given by the dest keyword argument, or the default name as explained above. action='store_const' You must provide a keyword argument const=V, where V is the value to be stored in the result returned from .arg_parse(). action='store_true' Store a True value in the returned result. If the user does not supply this option, .parse_args() stores a False value in the returned result. action='store_false' Store a False value in the returned result. If the user doesn't supply this option, .parse_args() stores a True value in the returned result. action='append' For arguments that may be repeated, this action causes each repeated argument to be appended to the list of values in the result returned by .arg_parse() . action='append_const' This works like action='append', but the value V, specified elsewhere by const=V is appended to the list of arguments. action='version' This option instructs the ArgumentParser instance to implement a --version option that reports the current version of your script. You must provide a version=V argument that defines the version string as V. If you don't supply an action argument, the default is action=None. In this case: For positional command line arguments, the value of the argument is stored. For optional command line arguments, the stored value is None, unless you provide a default=S argument to .add_argument(). nargs Specifies the number of this kind of argument. This feature works for both positional and optional arguments. In the value returned by .parse_args(), the attribute associated with this argument will be a list, not a single value. If the value of the nargs option is an integer, exactly that many arguments must be provided. nargs='*' means zero or more. For positional arguments, this means all the remaining arguments supplied will be included in the returned list of values. nargs='+' means one or more: there must be at least one such argument, but there may be any number. nargs='?' means that this argument is optional. For positional command line arguments, the returned value will be the value from the command line if there is one; otherwise you will supply the default returned value D by providing keyword argument default=D to the .add_argument() call. For optional command line arguments, nargs='?' signifies that the option may be given a value. If the user does not provide this option, the value returned by .arg_parse() will be the default value from the const=C argument to .add_argument(). If the user provides this option but does not follow it with a value, the value returned by .arg_parse() will be the default value from the default=D argument to .add_argument(). If the user provides this option with a value V, the attribute of the value returned by .arg_parse() will be V. The default value is nargs=None. In this case: For a positional command line argument, this means exactly one is expected. For an optional command line argument, the stored value is None unless you provide a default value D with default=D. const See above under action='store_const' and action='append_const'. default Provides a default value; see above under nargs. type Convert the value to a given type. For example, type=int would attempt to convert the associated argument to a Python int; if the argument is not a valid integer, .arg_parse() will print the usage message and terminate. For arguments that are file names, the argparse module will even open the file for you. Here is the general form: type=argparse.FileType(mode=M) where M is the mode string as in the second argument to open(). For example, this form would attempt to open a new file for writing, using the value of the command line option as the file name. p.add_argument(..., type=argparse.FileType(mode='w'), ...) Note that .arg_parse() may raise an IOError or OSError exception if the file can't be opened. You can specify any converter function as the value of the type keyword argument. This function takes one argument, a string, and returns a value of whatever type you like. Your converter function may also raise an argparse.ArgumentTypeError exception to signify an invalid value. choices An iterable that specifies the only valid choices. For example, choices=('red', 'grn', 'blu') would allow only those three specific strings as values of the associated command line argument. required If an argument that starts with a hyphen is not actually optional, use required=True. help A string describing what this option does. Strongly recommended, and it will be displayed in the help message. metavar Specifies the name of this optional for external display. For example, suppose your .add_argument() call starts like this: P.add_argument('inFile', metavar='INFILE', ...) Then the argument value will be stored in attribute .inFile of the result returned by .arg_parse(), but the help message will refer to this argument as “INFILE”.

The <code >ArgumentParser.parse_args()</code > method Once you have added all your command line arguments, the call to the .parse_args() method looks like this, where P is your ArgumentParser instance. P.parse_args(args=None, namespace=None) The args parameter specifies a set of command line arguments as a list of strings. If you omit this, the command line arguments will be taken from sys.argv. By default, the returned value will be an instance of class argparse.Namespace. The values returned by parsing the command line will be stored as attributes in this instance. However, you may instead use namespace to specify some instance to which the attributes will be added. Here's an extended example. This script sets up four command line arguments and then tests it against various simulated argument lists. #!/usr/bin/env python from __future__ import print_function import sys import argparse def test(p, argList): print("\n=== Test with", argList) r = p.parse_args(args=argList) print(vars(r)) p = argparse.ArgumentParser(prog='larch', description="Number 1: The Larch") p.add_argument('-n', '--name', default='Dinsdale', help='Name your amoeba') p.add_argument('-x', '--exec', action='store_true', help='Shoot amoeba afterwards') p.add_argument('in', help='Input file', metavar='INFILE') p.add_argument('outs', nargs='*', help='Output file(s)', metavar='OUTFILE') print("=== Usage message:") p.print_usage() print("\n=== Help message:") p.print_help() test(p, ['ingoat']) test(p, ['-x', 'Brian']) test(p, ['--exec', 'Brian', 'Reg', 'Dirk']) test(p, ['-n', 'Brian', 'Reg', 'Dirk']) test(p, ['--name=Pinnet', 'notlob', 'bolton']) test(p, ['--nosuch', 'Centurion']) Output of this script: === Usage message: usage: larch [-h] [-n NAME] [-x] INFILE [OUTFILE [OUTFILE ...]] === Help message: usage: larch [-h] [-n NAME] [-x] INFILE [OUTFILE [OUTFILE ...]] Number 1: The Larch positional arguments: INFILE Input file OUTFILE Output file(s) optional arguments: -h, --help show this help message and exit -n NAME, --name NAME Name your amoeba -x, --exec Shoot amoeba afterwards === Test with ['ingoat'] {'in': 'ingoat', 'name': 'Dinsdale', 'outs': [], 'exec': False} === Test with ['-x', 'Brian'] {'in': 'Brian', 'name': 'Dinsdale', 'outs': [], 'exec': True} === Test with ['--exec', 'Brian', 'Reg', 'Dirk'] {'in': 'Brian', 'name': 'Dinsdale', 'outs': ['Reg', 'Dirk'], 'exec': True} === Test with ['-n', 'Brian', 'Reg', 'Dirk'] {'in': 'Reg', 'name': 'Brian', 'outs': ['Dirk'], 'exec': False} === Test with ['--name=Pinnet', 'notlob', 'bolton'] {'in': 'notlob', 'name': 'Pinnet', 'outs': ['bolton'], 'exec': False} === Test with ['--nosuch', 'Centurion'] usage: larch [-h] [-n NAME] [-x] INFILE [OUTFILE [OUTFILE ...]] larch: error: unrecognized arguments: --nosuch Notes on this example: The print_function import uses the Python 3.x print() function; see . The vars() function is used to convert the argparse.Namespace instance returned by the .parse_args() method to a dictionary for display; see . The last time the test() function is called, the function does not return, because the specified set of options is not valid. The last two lines shown in the output below were sent to sys.stderr, not to sys.stdout like all the preceding lines.

Other useful <code >ArgumentParser</code > methods These methods are available on your ArgumentParser instance: .print_usage(file=None) Prints the short summary of argument usage. The default file is sys.stdout. .print_help(file=None) Prints the full summary of argument usage, including the help text for each argument. The default file is sys.stdout. .format_usage() Formats the short summary of argument usage and returns it as a string. .format_help() Formats the full help text and returns it as a string. .exit(status=0, message=None) Terminates execution with status 0 (or the status value you provide). If you provide a message string, that message will be printed before termination. .error(message) Prints the usage message, plus the message string you provide, then terminates execution with status 2. .add_mutually_exclusive_group(required=False) If you have two or more options that cannot be specified on the same command line, use this method to create an option group. Then call the .add_argument() on the group instance to add these options. If you specify required=True, the user is required supply one of the options in the group. Suppose for example that you have two mutually exclusive options --english and --metric. This code would prohibit the user from specifying both at once: p = argparse.ArgumentParser() g = p.add_mutually_exclusive_group() g.add_argument("-e", "--english", dest="isMetric", action="store_false") g.add_argument("-m", "--metric", dest="isMetric", action="store_true") .set_defaults(**kw) Use this method to specify the default values of any variable. For each keyword argument n=v, the value of n in the result returned by .parse_args() will have value v in case the user does not specify a value explicitly. For example, if you have two mutually exclusive options, but you don't require one or the other, the .set_defaults() method is a good way to specify the value of the option when neither is given. Here is an interactive example showing this technique. >>> parser = argparse.ArgumentParser() >>> group = parser.add_mutually_exclusive_group() >>> group.add_argument("-y", "--yes", dest="which", action="store_true") >>> group.add_argument("-n", "--no", dest="which", action="store_false") >>> parser.set_defaults(which=True) >>> print parser.parse_args(["--no"]) Namespace(which=False) >>> print parser.parse_args(["-y"]) Namespace(which=True) >>> print parser.parse_args([]) Namespace(which=True)