Блог пользователя MokaMoka

For Python Coders
Автор MokaMoka, история, 8 лет назад, По-английски

Hello Codeforces,

I started learning Python a week ago, and I've just solved my first problem using it. However; I was surprised by the huge execution time and memory for a very trivial problem, which makes me doubt if there's any effective tricks/advice to follow when using Python for competitive programming.

Wouldn't it be nice if we just collect some of those ideas here or somewhere else and make it available for everyone?

Please don't comment such comments like "The only advice I can tell is Don't use Python for competitive programming" :P, this is serious :)

  • Проголосовать: нравится
  • +16
  • Проголосовать: не нравится

»
8 лет назад, # |
← Rev. 2   Проголосовать: нравится +5 Проголосовать: не нравится

submit solutions using pypy.

even though pypy 2 passed, it barely passed O(nm) where 0 ≤ n, m ≤ 1000 in one second. i would keep a back-up language in case the problem requires a lot of computation.

→ Ответить
  • »
    »
    8 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    That sounds cool but, I'm a little bit confused. I've just submit the same solution again but this time using PyPy 2 instead of Python 2, and I was surprised that the execution time got slower :| as well as my code now uses much more memory.

    Can you provide a reason for that, please?

    → Ответить
    • »
      »
      »
      8 лет назад, # ^ |
      ← Rev. 2   Проголосовать: нравится 0 Проголосовать: не нравится

      And now it's getting even more weird :(

      → Ответить
      • »
        »
        »
        »
        8 лет назад, # ^ |
          Проголосовать: нравится +3 Проголосовать: не нравится

        i know this is too late.. but i need to say something..

        When there's lot of calculation, loops pypy will be faster than python..

        and if solution relies heavily on data structure(set, dict) python will be faster than pypy..

        → Ответить
»
8 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

array[j + i * m] instead array[i][j] will speed up python code

p.s. m — row size

→ Ответить
»
8 лет назад, # |
  Проголосовать: нравится +8 Проголосовать: не нравится

I like using Python a lot. Read the 3 links below, you may find them helpful:

  1. Python PerformanceTips
  2. Dict vs Lists for Graphs
  3. Only Python solution accepted so far. Had to optmize quite a bit

In short:

  1. Write your solution inside a function. Lookup variables in the global scope is more expensive, than using variables in the local scope.
  2. Avoid for loops like the plague. Favor Python builtins. Always use map, filter, concatenate strings with join, etc.
  3. Writing output is really expensive. Writting a gigantic string once, may be better than writting several small strings.
  4. Type casting is really expensive. Avoid if possible (string to int).
  5. Avoid dots. In other words: Cache function lookups. Example: Avoid using stdout.write directly, put it in a variable w = stdout.write.
  6. Careful with Dict vs Lists for graphs
→ Ответить