[raster] raster_columns out-db field should be an array

[pracine]

There should be one boolean value per band for out-db.

[mloskot]

Pierre,

Are you referring to the RASTER_COLUMNS specification or its implementation?

I'm not the author of the specification, I'm the author of its implementation (still incomplete). I'm not sure I'm the right person to fix the spec. I think it's a good idea to discuss it on the list, so Frank and others have chance to give comments.

I'm assigning myself.

[mloskot]

Pierre, what's status of this report? Have you had chance to update the spec?

[pracine]

I updated the final specification and the documentation.

[jorgearevalo]

So, finally the out-db field is a boolean, not an array. Isn't it?

I think that has more sense, because I don't think a raster with one band in-db and other one out-db is a common situation… All bands in-db or all bands out-db.

Am I right?

[pracine]

The problem is that out-db (actually offline in rt_api.c) is a per-band variable. So I'm trying to find in situations where we could end with a raster having one or many out-db bands and one or many in-db bands. Up to now the only case I found would be something like a ST_RasterFromBands(r1, r2, r3) function constructing a multi-band raster from raster columns originating from different tables. But we could decide that such a function always produce in-db rasters…

I would say lets leave it a boolean until we get in a situation where we can not avoid having out-db and in-db bands at the same time in a raster.

I set it to WKTRaster Future so we kepep track of it.

[dustymugs]

The field out_db field no longer exists in raster_columns. It may return once out-of-db raster support is fully part of PostGIS Raster.

[robe]

Milestone renamed