Потрошим змею: dis module

Почему некоторые операции быстрее других - ответы на очевидные вопросы с использованием timeit и dis.

Transcript

1. Потрошим змею: dis module Юлия Селютина MoscowDjango 2015
2. life cycle • C++: drivers • Python: scripts for Internet service provider • Python: banks data • Python: sites, statistic
3. dis – Python Bytecode Disassembler • Convert code objects to a human-readable representation of the bytecodes for analysis.
4. dis – Python Bytecode Disassembler • Understand inside
5. dis – Python Bytecode Disassembler • Debugging
6. literal syntax [] or separate object list()
7. literal syntax {} or separate object dict()
8. avoid the global lookup cost
9. avoid the loop cost
10. avoid the loop cost: loop hoisting
11. avoid the loop cost: timeit
12. avoid the loop cost: LIST_APPEND
13. LOAD_FAST and LOAD_GLOBAL • O(1), but const for LOAD_FAST is better • Cache in local variable
14. str vs % vs format
16. fast faster fastest list(), dict(), tuple() [], {}, () for i in xrange(100): l.append(i) la =l.append for i in xrange(100): la(i) [x for x in xrange(10)] “{}”.format(90) str(90) “%s” % 90 results
17. Questions