[Qgis-developer] ogr osm provider seems broken

[Even Rouault]

Le vendredi 06 juin 2014 08:55:39, Richard Duivenvoorde a écrit :
> Hi Devs,
> 
> to me the way qgis handles .osm files seems broken, see:
> 
> http://hub.qgis.org/issues/10000
> 
> http://hub.qgis.org/issues/10427
> 
> apparently ogr returns -1 on the number of features, but more important
> the feature interator seems broken
> 
> Problems: loading an osm file and viewing the attribute table only shows
> 2 records (while all points are shown on the map).
> Iterating over the features with python also raises errors.
> 
> Not sure if we should make this a blocker, because I'm not convinced of
> the importance of .osm files.

Richard,

Jukka pretty much answered rightly the questions in the tickets. The OGR OSM 
driver is rather tuned to handle multi-gigabyte .osm/.pbf files, and due to the 
structure of those files, it is impossible to count efficiently the number of 
features without reading it entirely. Hence -1 returned as the feature count. 
If you issue an explicit SELECT COUNT(*) FROM, it means that you know what you 
want, and you're ready to wait for it...

And, also related to the structure of the data, the driver is a bit particular 
in the sense that the iteratation over features with GetNextFeature() might 
not work if the osm file is of a "sufficient size". See the "Interleaved reading" 
paragraph of http://www.gdal.org/drv_osm.html.

For such cases, it might be interesting to develop a new concept in OGR, a 
whole dataset feature iterator, that could return features from different 
layers as soon as they are available.

Best regards,

Even


-- 
Geospatial professional services
http://even.rouault.free.fr/services.html