Performace tricks

  • use generators instead of lists were possible

  • run interpreter with -O

  • For big ranges use xrange() cause it creates a generator

  • Use xreadlines() for big files cause it also creates a generator

  • Setting environment variable PYTHONUNBUFFERED will force stdout / stderr to be unbuffered

  • os.fdopen("file", "r", 0) to open file without buffering

  • Use flush() on filehandle to flush I/O buffer

  • Use a list and join instead of string concatenation

  • Prefer while 1 instead of `` while True``

  • Write “x < y < z” instead of “x < y and y < z”

  • "test " + var is slower than "test %s" % (var,)

  • map(), generator and list comprehension are faster than a loop

  • put your global variables into a main function and call it by:

if __name__ == '__main__':
  main()

  • profile your code

import profile
profile.run(“any_python_expression”)