[Haskell-cafe] Trouble understanding records and existential
haskell at list.mightyreason.com
Fri Jan 26 05:32:41 EST 2007
John Meacham wrote:
> On Wed, Jan 24, 2007 at 05:03:18PM -0800, Stefan O'Rear wrote:
>> Haskell-98 style records are widely acknowledged as sucking, and there are
>> something like half a dozen proposals all of which are widely acknowledged
>> as vastly superior. Expect to be stuck with H98 records for the remainder
>> of time; see "bikeshed".
> actually, the problem is that we keep calling them records. Haskell 98
> records are actually labeled fields, not records, and as labeled fields
> they perform just fine.
> Not that records or named tuples or whatever you like to call them
> wouldn't be useful but they would likely be something in addition to labeled
> fields, not replacing it. (Not that the current labeled field mechanism
> couldn't be improved some.)
> personally, something based on Daan's scoped labels proposal is the
> clear leader of the bunch.
I also really liked Daan's "Extensible records with scoped labels", which is
available at http://www.cs.uu.nl/~daan/pubs.html#scopedlabels for those who
still have not read it. The system seems very simple, but also seems to have
required moving beyond an imperative viewpoint to come up with.
More information about the Haskell-Cafe