substituting list comprehensions for map()

Steven D'Aprano steve at REMOVE-THIS-cybersource.com.au
Tue Nov 3 05:04:47 CET 2009


On Mon, 02 Nov 2009 19:19:41 +1100, Ben Finney wrote:

> "Jon P." <jbperez at gmail.com> writes:
> 
>> I'd like to do:
>>
>> resultlist = operandlist1 + operandlist2
> 
> That's an unfortunate way of expressing it; it's valid Python syntax
> that doesn't do what you're describing (in this case, it will bind
> ‘resultlist’ to a new list that is the *concatenation* of the two
> original lists).

True, but it is valid mathematical syntax if you interpret lists as 
vectors. I'm sure there are languages where [1,2]+[3,4] will return 
[4,6]. Possibly R or Mathematica?


> Yes, just about any ‘map()’ operation has a corresponding list
> comprehension. (Does anyone know of a counter-example, a ‘map()’
> operation that doesn't have a correspondingly simple list
> comprehension?)

Everyone forgets the multiple argument form of map.

map(func, s1, s2, s3, ...)

would need to be written as:

[func(t) for f in itertools.izip_longest(s1, s2, s3, ...)]

which I guess is relatively simple, but only because izip_longest() does 
the hard work.


On the other hand, list comps using an if clause can't be written as pure 
maps. You can do this:

[func(x) for x in seq if cond(x)]

filter(cond, map(func, seq))


but the second version may use much more temporary memory if seq is huge 
and cond very rarely true.

And I don't think you could write this as a single map:

[func(a, b) for a in seq1 for b in seq2]


-- 
Steven



More information about the Python-list mailing list