Reconsider recommending map over list comprehensions

[gbrova]

Currently the examples in the structure section imply a list comprehension is "Better"

# create a concatenated string from 0 to 19 (e.g. "012..1819")
nums = [str(n) for n in range(20)]
print "".join(nums)

but using map is "Best"

# create a concatenated string from 0 to 19 (e.g. "012..1819")
nums = map(str, range(20))
print "".join(nums)

However, the guide doesn't recommend one approach over the other in the style section.

My impression is most people prefer list comprehensions because they are cleaner even if sometimes microscopically slower. See e.g. a discussion on StackOverflow and Google's style guide.

Is it intentional to recommend map over list comprehensions? If yes, we should add this recommendation in the style section for consistency; if not, we should remove the recommendation in the structure section. FWIW, my preference is list comprehensions.

[dbader]

Good point—also related to this discussion:

Guido seems to prefer list comps over map/filter

[mpoulin]

@dbader @gbrova
It seems to me that Python is evolving in the direction of iterators and lazy evaluation. Many of the built-in functions that returned lists in 2.x now return iterators in 3.x.

I think the recommendation should be

• Use generator expressions when you don't need to capture the entire output sequence.
• Use list comprehensions in the few cases when you really need to create a second list.
• Use generator functions when the logic is too complicated for a generator expression or list comprehension.

Lastly, never use a list comprehension just for its side effects.

BAD:

[print(x) for x in seqeunce]

BEST:

for x in sequence:
    print(x)

[gbrova]

That's a good point about lazy evaluation and generator expressions.

Should I throw together a PR?

[mpoulin]

Yes, I would like to see what you come up with 👍