[Qgis-developer] python/ogr problems reading large gml file

[Richard Duivenvoorde]

Thanks Even!

so
 os.utime(os.path.join('/tmp/foo/data.gfs'), None)
fixed all my strange problems :-)

apparently it was a some race-problem of the closing time of the gml
file on disk and the modification time of the gfs: smaller files just
worked etc etc..

anyway: thanks!
Maybe a warning about ignoring the 'old' gfs would be nice...

I tried:
$ ogrinfo --debug data1.gml
but that did not show me more info

Then I did:
$ ogrinfo --debug
ERROR 1: --debug option given without debug level.

Should I be able to use --debug from cli anyway?

Regards,

Richard


On 02-05-16 15:50, Even Rouault wrote:
> Le lundi 02 mai 2016 15:38:45, Richard Duivenvoorde a écrit :
>> Hi Devs,
>>
>> (below all via python/plugin on Debian/Linux):
>>
>> Via a WFS I retrieve 10000 features as GML as data.gml
>> To be able to read (attr+geometry) the GML I copy a data.gfs next to it.
>>
>> Now all seems ok, untill the number of features returned is > 1500.
>> Strange errors like:
>>
>> ogrinfo -ro data.gml reports a (None) geometry
>> BUT if I 'touch' the data.gfs ogrinfo, then suddenly all is ok
>> Seems like ogrinfo does not like the copied gfs at first?
>> Permissions are ok. A 'touch /tmp/foo/data.gfs' is enough.
> 
> The OGR GML driver only uses the .gfs file if its modification time is equal or 
> greater than the the one of the .gml file (to avoid using an outdated .gfs 
> file). If you run ogrinfo with --debug on, you'll see a message about that. 
> Should probably be a more verbose warning.
> 
>>
>> Off course the same holds for QGIS, if I open the retrieved and saved
>> gml file as a gml-vectorlayer, a 1000 feature containing one is probably
>> ok, but <1500 often fails and the 10000 one ALWAYS fails (untill I
>> 'touch' the gfs file).
>> IF something goes wrong: there is no exception, and 'layer.isValid()'
>> returns true, but geometry/attributes are None at such moment.
>>
>> I'm pretty carefull with closing/opening the files (I think), but it
>> smells like either files that are still open, OR python loosing it's
>> handles or so?
>>
>> Anybody some idea on what I am doing wrong?
>> Sorry, cannot sent you a wfs-url it's an internal service here.
>>
>> Thansk for any pointers
>>
>> Regards,
>>
>> Richard Duivenvoorde
>> _______________________________________________
>> Qgis-developer mailing list
>> Qgis-developer at lists.osgeo.org
>> List info: http://lists.osgeo.org/mailman/listinfo/qgis-developer
>> Unsubscribe: http://lists.osgeo.org/mailman/listinfo/qgis-developer
>