[wktraster] Add out-db support in gdal2wktraster.py script

[jorgearevalo]

Attached is a patch adding support for creating out-db rasters in gdal2wktraster.py script. Please review and correct it where needed.

[jorgearevalo]

Replying to mloskot: I'm simplifying the patch. I'll post a new version ASAP

[mloskot]

Jorge, that's great, thanks! So, I'm deferring review until the new version is attached. No rush!

[pracine]

Now that Jorge has access to SVN he can apply the patch himself. If it's already done, please close the ticket.

[jorgearevalo]

I've implemented this (function wkblify_band):

• bandidx (band number) comes from caller method: wkblify_raster_level. 1-based index
• the path to the external file is the same provided with the option "-r <raster_name>" when calling the loader

After this, I've commented these lines in rtpostgis.sql.in.c:

— Verify out_db
— TODO: Waiting for out-db rasters
—IF ( p_out_db = true ) THEN
— RAISE EXCEPTION 'TODO: Outside database rasters not implemented yet';
— RETURN 'fail';
—END IF;

Once re-compiled, I've correctly loaded a raster with outdb support doing this. Do you think the approximation is correct? If yes, I'll finish the out-db raster reading support in GDAL, and I'll test it (this is one of my GDAL tickets: ​http://trac.osgeo.org/gdal/ticket/3234

[jorgearevalo]

I've just committed this change (r5102). I'm testing it by WKT Raster GDAL driver (working on it). Waiting for comments to resolve the case.

[jorgearevalo]

I forgot: I think this storage method only has sense if the raster table has 1 file, of size RasterXSize, RasterYSize. What do you think?

[pracine]

This storage method makes sence mostly when a big raster dataset is cut into (JPEG) tiles for a web server. The web application first grab the path to the file list and then forward the tiles directly from the file system without reading the whole tiles from the database. This is supposed to be faster…

[jorgearevalo]

Replying to jorgearevalo:

I forgot: I think this storage method only has sense if the raster table has 1 file, of size RasterXSize, RasterYSize. What do you think?

Sorry, I meant 1 row. Each out-db raster has to have 1 row, of RasterXSize, RasterYSize.

[mloskot]

Jorge, I haven't tested but the commit looks good to me.

Just one note, I'd suggest to remove TODO comments when a feature is completed, instead of commenting out such comment. I mean the comment in rtpostgis.sql.in.c:539​

[jorgearevalo]

Sure. I've deleted the commented code, TODO comment included (r5103). I close the ticket for now…