sandbox.py 文件源码

def load_ai(filename):
    # the sandbox is currently as clever as ... well ... it's not clever
    code = open(filename).read() # quick and dirty, yet production ready code ;)
    module = imp.new_module('ai')
    mdict = module.__dict__
    def getboard(board, x, y):
        if x in range(8) and y in range(8):
            return board[y][x]
        else:
            return None
    mdict['getboard'] = getboard

    # Pop the obvious exploits
    mdict.pop('exec',0)
    mdict.pop('eval',0)

    # Lexical analyzer
    # TODO: Fix the stupid
    forbidden = ['import subprocess', 'import sys', 'import os']
    for word in forbidden:
        if word in code:
            raise Exception('Found security issue in {} while looking for {}'.format(filename, word))

    exec(code, mdict) # 'exec' is the root of all eval, here we gonna sandbox
    return module