[gdal-dev] VRT OGRVRTDataSource filter & python function

[Peter Schmitt]

Hi,

I have two GDAL questions that are Mapserver adjacent (or perhaps
vice-versa?).

1.  Does OGRVRTDataSource allow Python code similar to raster VRT Python
Pixel Functions?
2.  Is there a reasonable way to pass custom arguments to a VRT?

I have a VRT that executes some SQL:

    <!-- file.vrt -->
    <OGRVRTDataSource>
      <OGRVRTLayer name="0">
        <SrcDataSource>PG:host=db user=me dbname=datastore
password=abc</SrcDataSource>
        <SrcSQL>
          SELECT some_table.location, some_table.id, some_table.geom FROM
some_table
        </SrcSQL>
      </OGRVRTLayer>
    </OGRVRTDataSource>

Is there a reasonable way to pass some argument to filter the VRT?  For
example, a condition like `WHERE some_table.id = 123`.  Imagine the query
has a very large number of records. I dont want to process the filter
"client side":  Ideally the condition gets executed in SQL before OGR gets
any results.  I am aware of ogr2ogr's -sql statement
https://gdal.org/en/stable/programs/ogr2ogr.html#cmdoption-ogr2ogr-sql

Ultimately this will happen via MapServer, which is referencing the VRT
layer.  I would like the client to pass a QUERYSTRING argument `foo_id=123`
to perform a filter on the data in the VRT.  Usually I do this via
MapServer Runtime substitution.  This works if I embed the contents of
file.vrt directly in my Mapfile

    LAYER
        NAME my_layer
        TYPE POLYGON
        EXTENT -180 -90 180 90
        PROJECTION
            "init=epsg:4326"
        END
        CONNECTIONTYPE OGR
        TILEITEM 'location'
        TILEINDEX "<OGRVRTDataSource>...WHERE some_table.id =
%foo_id%</SrcSQL></OGRVRTLayer></OGRVRTDataSource>"

        VALIDATION
            'foo_id' "^[0-9]+$"
            'default_foo_id' "-1"
        END
    END

However, my actual OGRVRTDataSource file is large.  If I inline a large VRT
Mapserver complains "msBuildPath(): Unable to access file. Path is too
long.  Check server logs.".  I could use a Mapserver FILTER
https://mapserver.org/mapfile/layer.html#filter, but I think that occurs
AFTER the VRT is evaluated. . . so if there are millions of rows, that data
gets filtered by the client (Mapserver) and not by SQL, so there's a lot of
data moving around and very slow performance.

Thanks!
Pete
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20250728/54d49b25/attachment.htm>