FP in Python [was: Re: why python annoys me]

[Alex Martelli]

"Oliver Korpilla" <korpo at 01019freenet.de> wrote in message
news:3AE19990.B41D5C9D at 01019freenet.de...
> Good to see I always do learn something new from this newsgroup!
>
> Actually I was relying on book knowledge about Python, a book that
> especially mentions FP tools in Python but NOT list comprehension at all
> (never saw it in my books).
>
> Is it a new feature ? (version introduced ?)

It's relatively new, I believe it was introduced in 1.6 in the
summer of the year 2000 (or was it in 2.0? hard to recall,
since 1.6 final and the first 2.0 alpha came the same _day_).


> Since I didn't see it on list manipulation either!
>
> Where do I get more info about it ? (part of the tutorial)

List comprehension is documented in the standard Python
docs, but the language manual isn't meant for easy reading.
http://www.python.org/doc/current/ref/lists.html, anyway.


> Besides:
> If inlining functions does the trick for sharing local variables into
> local function definitions (without being parameters), then filter f xs
> could do the trick as well as the list comprehension does.. is this
> correct ?

Sorry, I'm not sure I understand your sentence.  Nested scoping,
if that's what you mean, is in Python 2.1 (experimental, yet, you
have to explicitly enable it; will be standard in 2.2).  Anyway, in
my opinion the examples I give DO show that 'filter does the
trick as well as list comprehension', and that the only extra that
nested scoping will give you for this qsort program is more choice
in what you choose to name or not name.  In Haskell, too, you
could use filters instead of the two list comprehensions, and
again you get choices regarding naming or not naming things.

Not quite the same choices as Python -- referring to head and
tail of list is vastly easier WITH names in Haskell, while in
Python the [0] indexing and [1:] slicing mean that naming
them is strictly a stylistic choice; on the other hand, a local
function in Python can be unnamed [lambda] only _with_
serious constraints on what it can do, you need to name it
if you want full generality.  Six of one, half a dozen of the
other.  Plus, I wouldn't obsess about what needs to be given
a name or may be given a name and what may or must be
left unnamed -- that doesn't seem the crux of the matter.


Alex