Add lwcollection_homogenize and ST_Homogenize

[colivier]

/*
** Given a generic collection, return the "simplest" form.
**
** eg: GEOMETRYCOLLECTION(MULTILINESTRING()) => MULTELINESTRING()
**     GEOMETRYCOLLECTION(MULTILINESTRING(), MULTILINESTRING(), POINT())
**      => GEOMETRYCOLLECTION(MULTILINESTRING(), MULTIPOINT())
**
** In general, if the subcomponents are homogeneous, return a properly
** typed collection.
** Otherwise, return a generic collection, with the subtypes in minimal
** typed collections.

I need this stuff for ST_GeomFromKML, to be able to return simplest geometry forms if KML input is mixed.

[colivier]

Synopsys become:

/*
** Given a generic geometry, return the "simplest" form.
**
** eg: 
**     LINESTRING() => LINESTRING()
**
**     MULTILINESTRING(with a single line) => LINESTRING()
**
**     GEOMETRYCOLLECTION(MULTILINESTRING()) => MULTILINESTRING()
**
**     GEOMETRYCOLLECTION(MULTILINESTRING(), MULTILINESTRING(), POINT())
**      => GEOMETRYCOLLECTION(MULTILINESTRING(), POINT())
*/

Commited in lwgeom as r5265 and r5266

Do you think is worth to be exposed as SQL function also ? And if yes ST_Homogenize is a good name choice ?

[colivier]

If no one react, i guess that it doesn't worth that much to expose to SQL this function.

I close the ticket.

[strk]

Reacting, as I like this function Olivier: still interested in bringing it in ?

[colivier]

Sandro,

Core function is already in. just need to expose it trought PostGIS API.

Can took it, with an help on the related documentation.

[robe]

Olivier,

If you provide PostGIS API I'll help you document it.

[pramsey]

OK, I've updated the internal function to handle more types, now to do the SQL binding.

[pramsey]

And the function is committed to trunk at r9009

[pramsey]

Docs committed at r9010

[strk]

Why did this one end up being called ST_CollectionHomogenize rather than just ST_Homogenize ? It also homogenizes MULTI*, not just collection. Are we in time to rename before going final or should we live with the current name forever ?

[strk]

Beside, documentation is also wrong: "Collections that are homogeneous will be returned as the appropriate multi-type", while instead a GEOMETRYCOLLECTION composed by a single element is returned as a point, not a multi-point. Or maybe that's what "singleton" was about..

[pramsey]

Replying to strk:

Why did this one end up being called ST_CollectionHomogenize rather than just ST_Homogenize ? It also homogenizes MULTI*, not just collection.

It puts it next to ST_CollectionExtract it's similar-but-different functional cousin.