Most programming languages use data sharing to improve performance, but this can bite you in the a** when you didn't expect sharing. Here's a simple example (python) and simple solution that I ran into recently.
I had this code:
respdict={}
for val in ['foo','bar','baz']:
respdict[val] = lambda arg: (arg == val+"")
myfunc = respdict['foo']
print myfunc('foo')
print myfunc('bar')
False
False
hunh???? I expected True and False.
What's happening is that 'val' is being shared across all the lambda's (anonymous functions) that are being "created" in the for-loop. While this sounds crazy, it's a actually fairly common in languages which use sharing to improve performance.
There's complex per-language solutions, but here's a much simpler one: just "force" the language to make a copy of the inner value by performing some no-op operation on the data. Numbers aren't shared so it's a non-issue, and other data types you need to call an explicit copy function. In languages with smart optimizing compilers, watch out for the optimizer which can out-smart you and know that the operation is a no-op, and remove it-- if this happens to you, just be extra devious e.g. hit the string with a regexp.
respdict={}
def mkfunc(val):
return lambda arg: (arg == val)
for val in ['foo','bar','baz']:
respdict[val+""] = mkfunc(val+"")
myfunc = respdict['foo']
print myfunc('foo')
print myfunc('bar')
True
False
