[gdal-dev] Re: Slow writing into Spatialite with -skipfailures

Jukka Rahkonen jukka.rahkonen at mmmtike.fi
Tue Apr 10 09:51:42 EDT 2012


Even Rouault <even.rouault <at> mines-paris.org> writes:


> Hum, would be interesting to have a sample to examine one of the 
geometry that 
> triggers the error. The SRID hypothesis seems unlikely, and as far 
as the 
> geometry type is concerned, it is weird because you have specified -nlt 
> MULTIPOLYGON ...
> 
> I imagine once the reason for the error(s) is more clearly identified, 
the 
> possibility for a prefiltering in ogr2ogr could be studied to avoid
 trying to 
> insert such features, and thus benefiting for transactions.

I managed to trap one problematic case but there is something weird 
going on. The feature comes originally from Oracle and it is for sure 
a polygon there (GTYPE=2003). I have converted it into shapefile with 
GDAL 1.6 and OCI plugin from OSGeo4W because for some reason I have 
not managed to connect our Oracle with more recent versions.

Now GDAL 1.6 that I used for converting polygon from Oracle into
shapefile tells that geometry is a polygon, while GDAL 1.9 claims 
that it is a multipolygon.

Ogr2ogr of GDAL 1.9 with this shapefile created by GDAL 1.6 leads 
to an error:

ogr2ogr -f SQLite -dsco spatialite=yes -dsco init_with_epsg=yes
-a_srs epsg:2393  multi.sqlite multi.shp

ERROR 1: sqlite3_step() failed:
  multi.GEOMETRY violates Geometry constraint 
[geom-type or SRID not allowed] (19)
ERROR 1: ROLLBACK transaction failed: cannot rollback - no 
transaction is active

ERROR 1: Terminating translation prematurely after failed
translation of layer multi (use -skipfailures to skip errors)

The error is the same even if I use -nlt multipolygon.

If I convert the GDAL 1.6 shapefile into a new shapefile with GDAL 1.9 the
geometry is reported to be of type POLYGON and conversion into spatialite is
successful. I cannot say if the problem is in GDAL 1.6 or in the Oracle OCI
driver or what but anyway I believe that the shapefile is somehow corrupted or
at least GDAL 1.9 thinks so.

The one-feature shapefile (multi?)polygon shapefile created by 
GDAL 1.6+OCI is here
http://latuviitta.org/documents/multi.zip

-Jukka Rahkonen-



More information about the gdal-dev mailing list