An Introduction to the Zen of Python

transcript

An Introduction tothe Zen of Python

Doug Hellmann February, 2011

Friday, February 18, 2011

The Zen of Pythonby Tim Peters

• Beautiful is better than ugly.• Explicit is better than implicit.• Simple is better than complex.• Complex is better than

complicated.• Flat is better than nested.• Sparse is better than dense.• Readability counts.• Special cases aren't special enough

to break the rules.• Although practicality beats purity.• Errors should never pass silently.• Unless explicitly silenced.• In the face of ambiguity, refuse the

temptation to guess.

• There should be one – and preferably only one – obvious way to do it.

• Although that way may not be obvious at first unless you're Dutch.

• Now is better than never.• Although never is often better than

*right* now.• If the implementation is hard to

explain, it's a bad idea.• If the implementation is easy to

explain, it may be a good idea.• Namespaces are one honking great

idea – let's do more of those!

Beautiful is better than ugly.

• Write programs for human readers

• Simple expression syntax

• Consistent syntax and behavior

1 i = 5 + 5 2 3 s = 'literal string' 4 5 m = """multi-line 6 string 7 """ 8 9 us = u'üñîçø∂é string'10 11 vals = [ 1, 3, 5, 7, ]12 vals[0]

13 14 t = ('a', 1, vals, None) 15 t[1] 16 17 d = { 'a':'a value', 18 'b':'b value', 19 } 20 d['a'] 21 22 if a or b: 23 do_something() 24

Simple Expressions

Explicit is better than implicit.

• Boolean type

• No “hidden” loop variables

• Use “self” to refer to object inside a method

Boolean Values

1 t = True 2 f = False 3 4 if t: 5 print 't is true' 6 7 if not f: 8 print 'f is false' 9 10 a = t or f 11 b = t and f 12 13 print a 14 print b

$ python booleans.pyt is truef is falseTrueFalse

for loops

1 print 'Numbers:' 2 3 for i in range(5): 4 print i 5 6 print 7 print 'Letters:' 8 9 for c in [ 'a', 'b', 'c' ]: 10 print c

$ python for_loop.py

Numbers:01234

Letters:abc

Defining a Class

1 class MyClass(object): 2 """Class documentation. 3 """ 4 5 def __init__(self, arg1, arg2): 6 self.attr1 = arg1 7 self.attr2 = arg2 8 9 def do_something(self): 10 """Does some work and returns a value. 11 """ 12 return self.attr1 * self.attr2 13 14 obj = MyClass(1, 2) 15 print obj.do_something()

Defining a Class

Simple is better than complex.

• Garbage collection

• print statement

• Interactive interpreter for experimentation

Printing

1 # -*- encoding: utf-8 -*- 2 3 print 'string:', 'string' 4 5 print 'unicode:', u'üñîçø∂é string' 6 7 print 'number:', 5 * 5 8 9 my_list = [ 1, 'a', None ] 10 print 'list:', my_list 11 12 my_dict = { 'a':'a value', 'b':'b value' } 13 print 'dict:', my_dict

$ python printing.py string: stringunicode: üñîçø∂é stringnumber: 25list: [1, 'a', None]dict: {'a': 'a value', 'b': 'b value'}

Printing, with printf-style Formatting

1 # -*- encoding: utf-8 -*- 2 3 my_list = [ 1, 'a', None ] 4 my_dict = { 'a':'a value', 'b':'b value' } 5 6 args = ('string', u'üñîçø∂é string', 5 * 5, 7 my_list, my_dict) 8 9 print ''' 10 string : %s 11 unicode: %s 12 number : %d 13 list : %s 14 dict : %s 15 ''' % args

$ python printing_with_format.py

string : stringunicode: üñîçø∂é stringnumber : 25list : [1, 'a', None]dict : {'a': 'a value', 'b': 'b value'}

Interactive Interpreter

Complex is better than complicated.

• Do complex things in a “Pythonic” way

• SciPy and NumPy for advanced science and math

Interactive Graphic Environment

SciPy, NumPy, matplotlib

NumPy Performance

1 a = range(10000000) 2 b = range(10000000) 3 c = [] 4 for i in range(len(a)): 5 c.append(a[i] + b[i])

Standard Python 1 import numpy as np 2 a = np.arange(10000000) 3 b = np.arange(10000000) 4 c = a + b

5-10 seconds “nearly instantaneous”

Flat is better than nested.

• Wide variety of modules in standard library

• Namespace is flat

• No need to use long names: com.company.java.blah.blah

Standard Library Modules: Networking

• socket

• select

• SocketServer

• BaseHTTPServer

• asyncore

• asynchat

• xmlrpclib

• SimpleXMLRPCServer

Standard Library Modules: Internet Protocols

• urllib

• httplib

• ftplib

• smtpd

• smtplib

• poplib

• imaplib

• json

Standard Library Modules: Command Line

• getopt

• optparse

• argparse

• fileinput

• cmd

• readline

Standard Library Modules: Parallel Processing

• subprocess

• threading

• multiprocessing

• Queue

Standard Library Modules: Databases

• anydbm

• pickle

• shelve

• sqlite3

Standard Library Modules: Miscellaneous

• datetime

• grp

• pwd

• profile

• timeit

• struct

• ctypes

• xml

• tarfile

• csv

Sparse is better than dense.

• Standard library stays “shallow”

• http://pypi.python.org/

• Allows separate release schedules

• pip (“pip installs Python packages”)

Readability counts.

• Whitespace for block structures

• Minimal punctuation

• Built-in documentation

• Extensive online documentation at http://docs.python.org/

Code Blocks

Minimal Punctuation

Inline Documentation

Special cases aren’t special enough to break the rules.

• Everything is an object

• Methods and functions differ by scope

• Dynamic typing and runtime attribute lookup

• Most “features” in external modules

String Methods

1 print ' string value '.split() 2 print ' string value '.strip() 3 print 'str' in ' string value ' 4 print 'CHUGALUG'.startswith('C') 5 print 'CHUGALUG'.endswith('end') 6 print ':'.join(['a', 'b', 'c'])

$ python string_methods.py['string', 'value']string valueTrueTrueFalsea:b:c

String Methods

Function and Method Scope

1 class MyClass(object): 2 3 def do_something(self, arg1, arg2): 4 return arg1 * arg2 5 6 def my_function(arg1, arg2): 7 return arg1 * arg2

Dynamic Attribute Lookup

1 class MyClass(object): 2 3 CLASS_ATTR = 'class value' 4 5 def __init__(self, arg): 6 self.attr = arg 7 8 o = MyClass('instance value') 9 print o.attr 10 print o.CLASS_ATTR

$ python dynamic_attribute_lookup.pyinstance valueclass value

Dynamic Typing

1 class A(object): 2 def do_something(self): 3 print 'in A' 4 5 class B(object): 6 def do_something(self): 7 print 'in B' 8 9 def work_on_obj(c): 10 c.do_something() 11 12 a = A() 13 b = B() 14 15 work_on_obj(a) 16 work_on_obj(b)

$ python dynamic_typing.pyin Ain B

Regular Expressions

1 import re 2 3 pattern = 'this' 4 text = 'Does this text match the pattern?' 5 6 match = re.search(pattern, text) 7 8 s = match.start() 9 e = match.end() 10 11 print 'Found :', match.re.pattern 12 print 'In :', match.string 13 print 'Range :', s, '-', e 14 print 'Substring:', text[s:e]

$ python using_regex.pyFound : thisIn : Does this text match the pattern?Range : 5 - 9Substring: this

Although practicality beats purity.

• Multiple Programming Models

• OOP

• Procedural

• Functional

Procedural

1 def my_function(arg1, arg2): 2 """This function multiplies arg1 by arg2. 3 """ 4 return arg1 * arg2 5 6 results = [] 7 for i in range(5): 8 results.append(my_function(i, i)) 9 10 print results

$ python procedural.py[0, 1, 4, 9, 16]

Functional: List Comprehension

1 results = [ i*i for i in range(5) ] 2 3 print results

$ python functional.py[0, 1, 4, 9, 16]

Functional: Generator Expression

1 results = ( i*i for i in range(5) ) 2 3 print results 4 5 for i in results: 6 print i

1 $ python generator_expr.py 2 <generator object <genexpr> at 0x1004d20a0> 3 0 4 1 5 4 6 9 7 16

Functional: Generators

1 def gen_values(n): 2 for i in range(n): 3 yield i*i 4 5 print gen_values(5) 6 7 for i in gen_values(5): 8 print i

1 $ python generator.py 2 <generator object gen_values at 0x1004d2050> 3 0 4 1 5 4 6 9 7 16

Errors should never pass silently.

• Exception-based error handling

• Tracebacks aid debugging

Exceptions and Tracebacks

1 def f(n): 2 if n > 0: 3 return n * f(n-1) 4 return 1 5 6 print f('5')

$ python traceback.pyTraceback (most recent call last): File ".../traceback.py", line 6, in <module> print f('5') File ".../traceback.py", line 3, in f return n * f(n-1)TypeError: unsupported operand type(s) for -: 'str' and 'int'

Raising Exceptions

1 def f(n): 2 if not isinstance(n, int): 3 raise TypeError('n should be an int') 4 if n > 0: 5 return n * f(n-1) 6 return 1 7 8 print f('5')

$ python raising.pyTraceback (most recent call last): File ".../raising.py", line 8, in <module> print f('5') File ".../raising.py", line 3, in f raise TypeError('n should be an int')TypeError: n should be an int

Raising Exceptions

1 def f(n): 2 if not isinstance(n, int): 3 raise TypeError('n should be an int') 4 if n > 0: 5 return n * f(n-1) 6 return 1 7 8 print f('5')

$ python raising.pyTraceback (most recent call last): File ".../raising.py", line 8, in <module> print f('5') File ".../raising.py", line 3, in f raise TypeError('n should be an int')TypeError: n should be an int

Unless explicitly silenced.

• Catch exceptions with try:except

• Process

• Convert

• Ignore

Catching Exceptions

1 def f(n): 2 if not isinstance(n, int): 3 raise TypeError('n should be an int') 4 if n > 0: 5 return n * f(n-1) 6 return 1 7 8 try: 9 print f('5') 10 except TypeError, err: 11 print 'Had an error:', err 12 else: 13 print 'No error' 14 finally: 15 print 'Always run'

$ python catching.pyHad an error: n should be an intAlways run

In the face of ambiguity, refuse the temptation to guess.

• Coerce types only when not surprising

• Cannot add strings and numbers

• Can multiply them!

Type Coercion

1 print 5 * 1.0 2 print 2 ** 100 3 print 2 * 'abc' 4 print 2 + 'abc'

1 $ python type_coercion.py 2 5.0 3 1267650600228229401496703205376 4 abcabc 5 Traceback (most recent call last): 6 File ".../type_coercion.py", line 4, in <module> 7 print 2 + 'abc' 8 TypeError: unsupported operand type(s) for +: 'int' and 'str'

Type Coercion

There should be one – and preferably only one – way to do it.

• Eliminate redundancy

• Easier to learn

• Easier to remember

Index API

1 a = [ 1, 2, 3 ] 2 print a[0] 3 4 a = (1, 2, 3) 5 print a[0] 6 7 a = 'abc' 8 print a[0] 9 10 a = {0:'zero value', 11 1:'one value', 12 } 13 print a[0]

$ python index_api.py11azero value

Although that way may not be obvious at first unless you’re Dutch.

• GUI

• GTK

• KDE

• Cocoa

• MS Windows

• Tk

• wxPython

• Web

• Django

• AppEngine

• Pyramid

• Flask

• TurboGears

• Pecan

Now is better than never.

• ctypes

• Jython

• IronPython

• Python 2 vs. Python 3

Although never is often better than right now.

• Language moratorium

• Not everything goes into stdlib any more

If the implementation is hard to explain, it’s a bad idea. Python Enhancement Proposals

If the implementation is easy to explain, it may be a good idea.

Namespaces are one honking great idea – let’s do more of those!

Closures

1 def outer(arg): 2 print 'outer(%s)' % arg 3 4 def inner(val): 5 print 'inner(%s) (arg=%s)' % (val, arg) 6 return val * arg 7 8 return inner 9 10 print 'Creating thrice:' 11 thrice = outer(3) 12 13 print '\nCalling thrice:' 14 print thrice(3) 15 print thrice('a')

Closures

1 def outer(arg): 2 print 'outer(%s)' % arg 3 4 def inner(val): 5 print 'inner(%s) (arg=%s)' % (val, arg) 6 return val * arg 7 8 return inner 9 10 print 'Creating thrice:' 11 thrice = outer(3) 12 13 print '\nCalling thrice:' 14 print thrice(3) 15 print thrice('a')

$ python closure.pyCreating thrice:outer(3)

Calling thrice:inner(3) (arg=3)9inner(a) (arg=3)aaa

References

• http://docs.python.org/

• http://www.doughellmann.com/PyMOTW/

• Learning Python by Mark Lutz

• Python Essential Reference by David Beazley

• The Python Standard Library by Example by Doug Hellmann

• http://us.pycon.org/

Questions?

Image Credits

1. http://www.gaialandscapedesignbackgrounds.net/landscape_design_architecture_background_Zen_Garden_chan3.htm

3. http://www.flickr.com/photos/grabadonut/539151245/sizes/l/in/photostream/

13. http://www.flickr.com/photos/tobascodagama/3106398602/

17. http://www.flickr.com/photos/benandliz/11065336/

21 & 22. http://scipy.org

24. http://www.otherlandtoys.co.uk/russian-matryoshka-nesting-dolls-pieces-p-2126.html

32. http://www.dandelionsummers.com/html/building_block.html

36. http://www.flickr.com/photos/photoinference/2494576240/in/faves-40068198@N04/

47. http://www.flickr.com/photos/smithsonian/3112472619/in/faves-40068198@N04/

54. http://www.flickr.com/photos/mherring/3722272868/sizes/l/in/photostream/

57. http://www.webstaurantstore.com/9-3-4-x-4-1-2-unbleached-natural-coffee-filter-12-cup-1000-cs/121CF12U.html

59. http://brainstorm-services.com/wcu-lit/ambiguity.html

66. http://commons.wikimedia.org/wiki/File:Dutch_windmill.jpg

67. http://www.epicpm.org/epic-blogging

68. http://www.skyscanner.net/news/articles/2009/09/002990-travel-tip-10-get-on-last-and-learn-the-art-of-patience.html

69. http://gizmodo.com/5015735/the-top-10-rube-goldberg-machines-featured-on-film

71. http://www.nutricaplabs.com/capsule-manufacturer.aspx

75. http://wmltech.wordpress.com/2006/12/09/got-questions/

An Introduction to the Zen of Python

Documents