Add Geometry Validation functions from PostGIS

[blikij]

Would it be possible to add the geometry validation functions. Such as ST_IsValid, ST_IsValidDetail, ST_IsValidReason and ST_MakeValid? The European Environment Agency could use this excellent work in our data reporting tool making spatial validation as well possible.

[sheinbergon]

Hi @blikij

I can add this PostGIS like functionality, sure.
I'll make it a part of the upcoming release.

Bear in mind that much like PostGIS, validation will work on sane geometries only (for example, you can't create a polygon comprised of a non closed ring). Such shapes will cause queries to fail (which is also what happens in PostGIS)

It will of course, work for softer errors (like self Intersecrting polygons)

[sheinbergon]

@bliki Bothj ST_IsValid and ST_IsValidReason (including their ESRI flag variants) will be part of the next release (27.2).

ST_IsValidDetail would not be implemented for the time being, as I've yet to unrevel the whole expand object thingy in Dremio ( there are no composite tuple types like Postgres).

I'll also try and push ST_MakeValid into that relese, but it'll require more work

[blikij]

Thank you so much. ST_IsValidDetail is not that crucial for us. If the other cn be implemented we have something I can use in production.

[sheinbergon]

Do you plan on using ST_MakeValid in production?

[blikij]

Do you plan on using ST_MakeValid in production?

Yes, It would allow us to perform additional validations on the data. If the errors are minor... The result would not be written back, instead we would write into an additional table... (using a create table... select...)

[sheinbergon]

OK. Got it. I'll make an effort to add that to the release plan

[sheinbergon]

@blikij I can confirm that an implementation of ST_MakeValid, based on JTS's GeometryFixer will make it into the upcoming release.

You will be able to call

ST_MakeValid(geom)
ST_MakeValid(geom,'keepcollapsed=true|false')

bear in mind this is somewhat experimental as GeometryFixer is somewhat of a new feature for JTS. Results seem to differ from the GEOS / PostGIS implementation

You can see further discussion here locationtech/jts#959

[sheinbergon]

Released in 0.8.8