cvs commit: fptools/libraries/base package.conf.in fptools/libraries/base/Data IntMap.hs IntSet.hs Map.hs FiniteMap.hs Set.hs

Ketil Malde ketil+haskell at ii.uib.no
Mon Jan 17 03:48:16 EST 2005


Tomasz Zielonka <tomasz.zielonka at gmail.com> writes:

>> Hmmm...I have perhaps a rather shallow reason.  When using a FiniteMap
>> to collect lists of values, I think it will be more efficient to use
>> (flip (++)), prepending instead of appending each new element. 

> If (flip (++)) is efficient for one interface, shouldn't (++) be as efficient
> for the other?

Yes.  Sorry if that wasn't clear; my point was that the new interface
saves you the 'flip' -- or alternatively, the naive user not
considering these issues will likely just use (++), which may be more
efficient for the new interface. 

> Why not use efficient catenable sequences or a ShowS trick, like here:

>     groupFM :: Ord a => [(a, b)] -> FiniteMap a [b]
>     groupFM l =
>         mapFM (\_ -> ($ [])) $
>             addListToFM_C (.) emptyFM [ (a, (b:)) | (a, b) <- l ]

Because it is more complex?  It's a neat trick, but it does takes me a
minute or two to see what's going on.  Perhaps groupFM should be part
of (Finite)Map?

-kzm
-- 
If I haven't seen further, it is by standing in the footprints of giants



More information about the Libraries mailing list