Wrong data returned for some OGRSQL queries combined with OGR_L_SetNextByIndex

[theroggy]

The following combination of factors leads to wrong data being returned:

• the arrow API is used
• to execute an "OGRSQL" sql query that contains a "LIMIT" clause (e.g. SELECT * FROM naturalearth_lowres LIMIT 1). When the "SQLITE" dialect is used it seems to work fine.
• and this is combined with using OGR_L_SetNextByIndex to skip some rows

In this combination the rows aren't skipped correctly.

I'm afraid I don't have a code snippet ready to reproduce it... it was noticed while working on the arrow interface in pyogrio, so in cython.

However, it can be triggered by eg. running "SELECT * FROM LIMIT 1" on any dataset using the "OGRSQL" dialect and setting OGR_L_SetNextByIndex to index 1. This should return 0 rows, but it returns one row.

Tested on gdal 3.7 and 3.8, and this gives the same result.

reference: geopandas/pyogrio#306

[rouault]

This is not specific to the Arrow interface.

$ python
>>> from osgeo import ogr
>>> ds = ogr.Open('poly.shp')
>>> lyr = ds.ExecuteSQL("SELECT * FROM POLY LIMIT 1")
>>> lyr.SetNextByIndex(1)
>>> lyr.GetNextFeature()
OGRFeature(poly):1
  AREA (Real) = 247328.172
  EAS_ID (Integer64) = 179
  PRFEDEA (String) = 35043423
  POLYGON ((480035.34375 4765558.5,480039.03125 4765539.5,479730.375 4765400.5,479647.0 4765369.5,479690.1875 4765259.5,479819.84375 4765180.5,479779.84375 4765109.5,479681.78125 4764940.0,479468.0 4764942.5,479411.4375 4764940.5,479353.0 4764939.5,479208.65625 4764882.5,479196.8125 4764879.0,479123.28125 4765015.0,479046.53125 4765117.0,479029.71875 4765110.5,479014.9375 4765147.5,479149.9375 4765200.5,479639.625 4765399.5,480035.34375 4765558.5))

[theroggy]

Sorry for the misinformation about it being specific for Arrow. I don't see where, but somewhere in the pyogrio code it must be "handled" for the non-arrow case...