[postgis-devel] performance of force_collection()

[Ron Mayer]

strk at refractions.net wrote:
> Just another note. Your patch is not forcing a copy on DETOAST.
> This would usually be the case, but if input is not a copy
> editing it is an unsafe operation.

Ah yes... so I guess I should have checked
    TYPE_GETTYPE(geom->type) == COLLECTIONTYPE
and just return without editing it if I wanted to
avoid copying potentially large geometries?


Hmm. At that point perhaps I should hack mapserver
to not make those calls instead. :)  Too bad I never
looked at it's source. :(