Python

Links

• The Python Tutorial
• Python 2.7.6 docs
• Python Quick Reference 2.7 — Extremelly complete

Other versions of Python are available [1]

• Nice example of generating / testing regex in Python (with nice / small test framework) [2]

PEP

• PEP 8 - Style Guide for Python Code
• PEP 257 - Docstring Conventions

Tools

• autopep8 — A tool that automatically formats Python code to conform to the PEP 8 style guide

sudo pip install --upgrade autopep8

Python variants

iPy

Use iPy (ipython) to get an interactive shell with auto-completion, instant help...

%magic                    # Get help on %magic commands (%run,...)
?run                      # Get help on %run magic
%run script.py            # Run given script
%run -i script.py         # ... with inspect mode on
%run -i -e script.py      # ... ... and ignore sys.exit() call
!cmd                      # Run shell command 'cmd', for instance ...
!ls                       # ... List file in current directory

Pypy

PyPy is a fast, compliant alternative implementation of the Python language, which usually runs python programs faster thanks to its Just-in-Time compiler.

Install

On Lucid 64-bit, the easiest is to download the dedicated tarball:

wget https://bitbucket.org/pypy/pypy/downloads/pypy-2.2.1-linux64.tar.bz2
tar -cvjf pypy-2.2.1-linux64.tar.bz2

Install virtualenv, then install pypy as virtual environment my-pypy-env

sudo apt-get install python-virtualenv
virtualenv -p pypy-2.2.1-linux64/bin/pypy my-pypy-env

Modules must be installed separatedly for this virtual environment. For instance

./my-pypy-env/bin/pip install libnum

Run

Run python programs using python or pypy

./my-pypy-env/bin/pypy

Reference

• range

Basic

for i in range(10):
    print i                      # carriage return

for i in range(10):
    print i,                     # no carriage return

for key in d:                       # Loop over keys in dictionary d
for key, value in d.iteritems():    # Loop over keys and values in dictionary d

a = 'global'
def afunction():
    global a
    a = 'still using global'
    b = 'local'

import os.path
os.path.isfile(fname)            # True if fname exists and is a file

if not os.path.exists(directory):
    os.makedirs(directory)       # Create directory if does not exists

try:                             # Avoid race condition if directory created by another process
    os.makedirs(path)            # But we could fix solution above as well
except OSError:                  # This one always trigger an exception in nominal case
    if not os.path.isdir(path):  
        raise
                                  

s.upper()                        # string s to uppercase
', '.join(set_3)                 # Join a sequence

sys.argv, len(sys.argv)          # Argument list, number of arguments ([0] -> exec name)

# Sort based on object attribute
ut.sort(key=lambda x: x.count, reverse=True)   # To sort the list in place...
newlist = sorted(ut, key=lambda x: x.count, reverse=True)  # To return a new list, use the sorted() built-in function...

From stackoverflow [3],

Modules

import datetime
print datetime.datetime.today()  
print datetime.datetime.now()    # similar, but possibly more accurate
print datetime.date.now()        # date only

Advanced

mymodule = __import__('mymodule')          # Import module from string - see http://effbot.org/zone/import-string.htm

Modular inverse [4]

# Using gmpy - FASTEST
import gmpy
gmpy.invert(1234567, p)                      # 1000000 loops, best of 3: 737 ns per loop (p 1024-bit)
gmpy.divm(1, 1234567, p)                     # 1000000 loops, best of 3: 933 ns per loop (p 1024-bit)

# Using egcd function - NO DEPS, BUT SLOWER
def egcd(a, b):
    if a == 0:
        return (b, 0, 1)
    else:
        g, y, x = egcd(b % a, a)
        return (g, x - (b // a) * y, y)

def modinv(a, m):
    g, x, y = egcd(a, m)
    if g != 1:
        raise Exception('modular inverse does not exist')
    else:
        return x % m
timeit modinv(1234567,p)                     # 100000 loops, best of 3: 13.6 us per loop (p 1024-bit)

# Using pow() - SIMPLEST BUT SLOWEST
timeit pow(1234567,p-2,p)                    # 100 loops, best of 3: 4.22 ms per loop

Tips

Simple HTTP Server

It's very easy to setup an ad-hoc HTTP server with Python. Just open a shell in a folder with some contents to share, and type:

python -m SimpleHTTPServer

More available at http://docs.python.org/2/library/internet.html (see BaseHTTPServer and CGIHTTPServer).

Detect interactive mode

References: [5], [6]

Started with	First method	Second method	Third method	Fourth method
	import __main__ as main print hasattr(main, '__file__')	def in_ipython(): try: __IPYTHON__ except NameError: return False return True	import sys print hasattr(sys, 'ps1'):	import sys print bool(sys.flags.interactive)
python mymod.py	True	-	-	-
python -i mymod.py	True	-	-	True
python then import mymod	-	-	True	-
ipython mymod.py	True	True	-	-
ipython -i mymod.py	True	True	-	-
ipython then run mymod.py	True	True	-	-
ipython then run -i mymod.py	True	True	-	-
ipython then import mymod	-	True	-	-
ipython -i then import mymod	-	True	-	-

Find duplicates in list

From stackoverflow [7]

import collections

def fastest():                         # 134 us - Fastest
    seen = set()
    seen_add = seen.add                                            # To avoid lookup 'add' ever time an item is inserted
    seen_twice = set( x for x in l if x in seen or seen_add(x) )   # adds all elements it doesn't know yet to seen and all other to seen_twice
    return list( seen_twice )                                      # turn the set into a list (as requested)

def compact():                         # 415 us
    return [x for x, y in collections.Counter(l).items() if y > 1]

def slowest():                         # 19.2 ms
    return list(set([x for x in l if l.count(x) > 1]))

Start post-mortem debugger on exception

From stackoverflow [8]

>>> import pdb
>>> pdb.pm()

Miscellaneous

Detect whether a variable is defined

Note it is bad practice to define a variable conditionally [9]. An interesting use case is to run code and define variable conditionally based on interactive status.

# Using try ... except
try: myvar
except NameError: print "variable 'myvar' IS defined"

# Using vars() / globals()
'myvar' in vars() or 'myvar' in globals()
# ...pedantic...
'myvar' in vars(__builtins__)

Analyse memory usage

Dowser

• See [10] — seems better suited to find memory leaks, not to analyse usage for memory hungry applications

memory_profiler

• See [11]
• Install

sudo pip install -U memory_profiler
sudo pip install psutil

• Add @profile decorator

@profile
def primes(n): 
    ...

• Run the profiler

 python -m memory_profiler primes.py

Do's and don't's

foo = 'abcdef' l = list(foo) # DO	foo = 'abcdef' l = [c for c in foo] # don't
foo = list(...) g = map(blah,foo] # DO	foo = list(...) g = [blah(i) for i in foo] # don't

Traps

Frequent mistakes. Beware the snake can bite you!

Confuse a method and a property in a test

SOLUTION: Stick to a convention. Like always define methods like isxyyz() or hasabc() as methods. Note that defining them as property would raise an exception if used as a function, and hence might be safer.

if A.isdummy():            # This will fail isdummy is a property
if A.isdummy:              # Always True if isdummy is a method

Mix 0 with None in a sequence

Testing whether an element is defined is more difficult.

a = [0,None,None,None]
bool(a[0])           # --> False
bool(a[1])           # --> False !!! How can we tell them apart?
a[1] == None         # --> True      This works, but is unusual and likely bad practice

Mixing property and normal getter

SOLUTION: prefix all getter method with get, like getvalue()

b = a.prop           # Using a property, OR
b = a.getprop()      # Using a getter

Docstrings

Specifications: pep-0257

• To write good module docstrings, "think about somebody doing help(yourmodule) at the interactive interpreter's prompt — what do they want to know?" [12].
• See pep-0257 for more recommendations

Using doctest

You can include tests, in the form of examples, in your Python modules' docstrings. Properly written, these tests can be executed and verified by the doctest module. [13]

Libraries

Big numbers

• gmpy based on GMP
• [https://github.com/hellman/libnum libnum a lighter bignum library, but compatible with pypy.

Unicode

Set source file encoding

Add any of these lines [14]:

# -*- coding: utf-8 -*-
# vim: set fileencoding=utf-8 :

Write the BOM

See [15]

import codecs

file = codecs.open("lol", "w", "utf-8")
file.write(u'\ufeff')                          # or use unicode name: u'\N{ZERO WIDTH NO-BREAK SPACE}'
file.close()

# Using https://docs.python.org/2/library/codecs.html#module-encodings.utf_8_sig
with codecs.open("test_output", "w", "utf-8-sig") as temp:
    temp.write("hi mom\n")

Handling unicode

Some recommends to always process unicode internally, and decode on input and encode on output [16]:

line = line.decode('utf-8')
# ...treat line as unicode...
print line.encode('utf-8')

But this is error prone. So another solution proposed is to redefine sys.stdout:

import sys
import codecs
sys.stdout = codecs.getwriter('utf8')(sys.stdout)

An hackish way (not recommended):

# -*- coding: utf-8 -*-
import sys
reload(sys)
sys.setdefaultencoding('utf-8')
print u"åäö"