[Qgis-developer] Shape repack issue

[Even Rouault]

Andreas,

I cannot confirm. But the bug has been reopened without specifying the 
operating system, the GDAL version and the exact procedure followed (possibly 
with the dataset in case it might make a difference), so it might hit some 
particular case.

So here's my setup :
- versions tried : up-to-date versions of QGIS master and 2.14 branch
- OS: Linux
- GDAL >= 2.1

Steps :
1) Download:
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.shp
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.shx
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.dbf

2) Open the shapefile in QGIS

3) Turn on edition mode

4) Select the feature with EAS_ID = 165 (the one roughly at the middle)

5) Delete it

6) Turn off edition mode and confirm saving the changes

7) Quit QGIS and do "ogrinfo -al poly.shp". You'll see "Feature Count: 9" and 
9 features numeroted without hole from OGRFeature(poly):0 to 
OGRFeature(poly):8


Sure, a test could be added to simulate that (looking at 
test_provider_shapefile.py it doesn't seem there's one), but there's probably 
something we miss here w.r.t. the reporter setup.


On the reverse, to see the effect of *no* repacking, if you do :

1) Download:
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.shp
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.shx
http://svn.osgeo.org/gdal/trunk/autotest/ogr/data/poly.dbf

2) ogrinfo poly.shp -sql "delete from poly where eas_id = 165" -dialect sqlite

3) ogrinfo -al poly.shp. You'll see "Feature Count: 10" and features numbered 
from  OGRFeature(poly):0 to OGRFeature(poly):9 with a hole (OGRFeature(poly):8 
missing) --> the shapefile hasn't been repacked (as "expected" from current OGR 
behaviour)

Even

> Hi,
> 
> There is (hopefully was) this annoying ESRI Shape Repack issue - see
> http://hub.qgis.org/issues/11007
> 
> I see that this bug was re-opened - Antoine SIG claims that it worked in
> 2.14.1 and 2.14.2 (Windows) but fails again in 2.14.3. Can someone
> confirm? How can that happen? Can a test be added to ensure this doesn't
> fail again?
> 
> How about 2.16? Does it work there if it is built with GDAL 2.x?
> 
> I can't reproduce, since I don't have another GIS besides QGIS to check.
> 
> 
> Thanks,
> 
> Andreas

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com