lambda loop variable closure fix

[asishm]

Closes #2918

Issue arises because of this issue - https://stackoverflow.com/questions/2295290/what-do-lambda-function-closures-capture

the function generator could use a better name.

[stevenbird]

Thanks for the fix @asishm. How about using map() with a lambda argument to avoid the need to name the function, roughly like this:

map(lambda ..., conds)

[asishm]

So something like map(lambda cond: lambda sent_index, word_indices: cond, conds) ?

[tomaarsen]

That seems to work!

conds = "iterable"

print("Current approach")
functions = [lambda sent_index, word_indices: cond for cond in conds]
for function in functions:
    print(function(None, None))

print("Proposed approach")
def func_gen(x):
    return lambda sent_index, word_indices: x

functions = [func_gen(cond) for cond in conds]
for function in functions:
    print(function(None, None))

print("Proposed approach 2")
# For all conditions, create a 2-argument functions that returns the condition regardless of input
functions = map(lambda cond: lambda sent_index, word_indices: cond, conds)
for function in functions:
    print(function(None, None))

outputs:

Current approach
e
e
e
e
e
e
e
e
Proposed approach
i
t
e
r
a
b
l
e
Proposed approach 2
i
t
e
r
a
b
l
e

However, I can't properly verify this in Boxer, as I can't manage to download/install it correctly. The original source of C&C and Boxer seems to be down nowadays: svn.ask.it.usyd.edu.au/trac/candc

It seems a bit risky to merge this without verifying that it works as intended now.

[stevenbird]

@asishm Are you able to confirm that the fix works please?