On Fri, Nov 12, 2010 at 03:05, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "David E. Wheeler" <david@kineticode.com> writes:
>> So are you planning to implement multisets? It's a feature I'd love to see
>
> What actual functionality does it buy? AFAICT from Itagaki-san's
> description, it's an array only you ignore the specific element order.
> So what? You can write functions that work that way now.
I think there are almost no difference between a multiset and an array
in terms of functions I described in the first mail.
However, if we have separated multiset data type, we could have special
comparison operators for them; "array = array" returns true only if they
have the same elements in the same order, but "multiset = multiset" only
checks elements in them. Also, we could optimize on-disk structure of
multiset for fast UNION operations or for dataset that has many duplicates.
For example, we could use a sorted array of {value, count} pairs.
If we decide to have data type IDs for multiset, I'll go for it (ex. int4,
_int4, and an additional $int4), but it consumes +50% of typoids. If it
is not preferable, only function support might be better at the first try.
--
Itagaki Takahiro