Instead, for copying, the geometry and attributes are now retrieved from the provider in a random-access fashion.
For starting editing, the behaviour is regressed to the pre-r5489 idiom of geometry being cached as part of the screen redraw in editing mode.
This returns to the much better response when copying from large layers, with the additional benefit of copying selected features that had been panned off the canvas (previously a bug).
However, only Postgres and OGR providers have been updated to suit. Hopefully this is enough for a 0.8 release.
Bonus feature: The layer properties now shows what access method is used to access individual geometries (random or sequential)
Since this is a large change it would be appreciated if others can double checking my testing. (I have tested against PostGIS and MapInfo tables, and many different geometry adding/vertex editing combinations.)
git-svn-id: http://svn.osgeo.org/qgis/trunk@5595 c8812cc2-4d05-0410-92ff-de0c093fc19c
QGIS used to do this already before the move to Qt4 - however the new QIcon idiom only seems to allow for square icons, therefore it is easier to split the results out to several columns and give them their own QIcons.
I can see room for improvement after this commit:
1. Overview and Editing icons can be also shown beside the top-level layer name, to avoid having to enable "Show file groups".
2. There also seems to be a "projection not valid" icon floating around, however this hasn't been addressed in this commit. Perhaps we need to do something for it too.
3. The column sizes have been hard-coded to arbitrary values. I have no problem if these get tweaked further.
This commit addresses trac ticket #117.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5594 c8812cc2-4d05-0410-92ff-de0c093fc19c
If programmers wish to add more derived attributes in future, they can use the new QgsIdentifyResults::addDerivedAttribute() function.
This commit should address trac ticket #100.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5592 c8812cc2-4d05-0410-92ff-de0c093fc19c
Note this is not a complete solution as commits are currently a 3- or 4-stage process, and if one stage fails the chain cannot always be rolled-back easily. If this is the case then a detailed error message is given to the user to assist their understanding of the discrepency.
This commit should be a bug fix for trac #131, but I don't have the environment to test it fully - the bug submitter will have to assist there.
Bonus feature: Tidied up some variable names in QgsVectorLayer - makes it clearer what they contain.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5591 c8812cc2-4d05-0410-92ff-de0c093fc19c
- Put in an explicit conversion from Qgis layer type enum to
OGR layer type enum when creating a new layer
- When creating a new vector layer, and the user selected an existing
file, delete that file. This was causing a problem when the new layer
type was different to the selected layer type because the selected
layer type was being used.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5588 c8812cc2-4d05-0410-92ff-de0c093fc19c
As part of this "merge", the main application copy will now precede its results with a row of attribute headers.
Also, the editCopy() procedure and friends (editCut() and editPaste()) can now take the layer to act on, thus allowing the Attribute Table to work on a different layer to the Current Layer in the Legend.
This change was done so that the Copy button in the main application AND the Copy button on the attribute table works consistently.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5584 c8812cc2-4d05-0410-92ff-de0c093fc19c
why it needs to be that to get symbols of a similar size to the map
canvas, but it works fine.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5577 c8812cc2-4d05-0410-92ff-de0c093fc19c
- Tidy up #ifdef QGISDEBUGs
- Replace some qt3 compabitilty code with the 'qt4 way'
- Use a default label buffer size of 1 instead of 0 to reduce user confusion
git-svn-id: http://svn.osgeo.org/qgis/trunk@5573 c8812cc2-4d05-0410-92ff-de0c093fc19c
This is solving a similar problem to the linestring case in r5550.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5569 c8812cc2-4d05-0410-92ff-de0c093fc19c
* Bonus bugfix: If two adjacent verticies were moved in succession, the second vertex would show a rubberband from the *old* position of the first vertex instead of the moved-to position of the first vertex. This should now be fixed.
* Bonus function: New "get_at()" function in QgsGeometryVertexIndex. Should be fairly self-explanitory.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5568 c8812cc2-4d05-0410-92ff-de0c093fc19c
dialog that lets the user swap between using a QImage or a QPixmap for
rendering the map. May or may not fix things. Committed so that others
can try it out.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5567 c8812cc2-4d05-0410-92ff-de0c093fc19c
was 0.01, which in most cases was too small to generate a visible line
in the print composer window. Default is now 2.0
git-svn-id: http://svn.osgeo.org/qgis/trunk@5565 c8812cc2-4d05-0410-92ff-de0c093fc19c
to a QgsLogger::debug, as the messages are confusing the users, and
don't impact on Qgis's ability to load such files.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5563 c8812cc2-4d05-0410-92ff-de0c093fc19c
The attribute table was failing to display the contents of the last column
in a table for tables where the geometry column was not the last
in the table.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5552 c8812cc2-4d05-0410-92ff-de0c093fc19c
edit, the closest geometry will be selected (not just the first within
snapping range),
AND then the closest vertex *on that geometry* will be
selected.
This disambiguates between the same vertex shared amongst
multiple geometries.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5550 c8812cc2-4d05-0410-92ff-de0c093fc19c
changed geometries to pre-1.0 PostGIS servers (pre-r3907). This also now includes
the ability to sniff the PostGIS version.
git-svn-id: http://svn.osgeo.org/qgis/trunk@5549 c8812cc2-4d05-0410-92ff-de0c093fc19c
