GDAL thinks string field is an integer field in GML layer

[jbronn]

The attached GML file has features that have three fields, 'dbl', 'int', and 'str' that are of real, integer, and string types, respectively. The 'str' field is reported as being of integer type instead of being string type.

>>> from osgeo import ogr
>>> ds = ogr.Open('test_point.gml')
>>> lyr = ds[0]
>>> ldefn = lyr.GetLayerDefn()
>>> [ldefn.GetFieldDefn(i).GetFieldTypeName(ldefn.GetFieldDefn(i).GetType()) for i in range(ldefn.GetFieldCount())]
['Real', 'Integer', 'Integer'] # Should be ['Real', 'Integer', 'String']

The GML file (and xsd) were generated from using ogr2ogr -F GML on an existing  shapefile.

[warmerdam]

Chaitanya,

I *think* the GML driver will read the .xsd file if it is in the OGR produced "OGR simple features GML" format. Assuming that is true, could you look into whether we can exact the attribute field types from it in such a way that the field types are preserved with datasets like this?

We would need to add appropriate tests to the python test suite.

[chaitanya]

The GMLReader::ParseXSD() method was not handling attribute values in the .xsd file that have a prefix. I am attaching a patch that ignores the prefix.

By the way, the attached test_point.xsd file has the second type listed as decimal(Real) and will be treated as such.

Frank, I have tried to handle the issue during type-name verification, but there are more such verifications further down the road. So, I am just skipping the prefix part right after extracting the 'type' attribute value.

[chaitanya]

Fixes the problem with ignoring the .xsd file with 'type' attribute values with prefixes

[warmerdam]

The patch looks fine, but now the hard part is to extend the GML test script to test this capability - presumably including some new (small) test data.

[chaitanya]

Added test for ticket#2493

[chaitanya]

Data file for the test to be added to 'autotest/ogr/data' folder

[chaitanya]

Data file for the test to be added to 'autotest/ogr/data' folder

[chaitanya]

Committed fix in trunk (r16644). Added test for the issue in trunk (r16641, r16642 and r16643).

Frank, Should I add this fix in the 1.5 branch too?

[rouault]

Chaintanya,

I've re-added ogr_gml_6 in r16645 that apparently got accidently removed in r16641

[warmerdam]

Chaitanya,

Yes, please apply the update in the 1.6 and 1.5 branch. We do not normally port back the test suite additions - just the code change.

[chaitanya]

Applied the fix in 1.5 (r16649) and 1.6 (r16650) branches.