Fixes#38092 by adding an optional QgsFeedback argument to
the executeSql method and by implementing the PQCancel
method in the PG provider internals.
While the cancellation works well for all supported provider while
fetching results in the loop, the cancellation of a running query is now
implemented for the postgres provider connection only because the GPKG
and GDAL both rely on GDALDatasetExecuteSQL which cannot be interrupted.
This PR also introduce a few optimizations in the PG DB-Manager
code that should probably fix also other "slowness" issues that
were reported after 3.x during PG query execution.
A small UX change in th SQL dialog makes it evident to the user that
a cancellation request has been sent to the backend: the button text
is changed to "Cancellation requested, please wait..." so that for
provider connections that are not able to interrupt the running query
and must wait for the fetching loop to exit from the exeuteSql call
the user knows that something is happening and that a cancellation
request has been successfully sent.
This prevents the Python GIL from being released before calling the method
(which is the default behaviour). For very cheap to call c++ methods the
cost of releasing the GIL can outweigh the cost of the c++ call, which means
it's more efficient to retain the hold on the GIL.
Ideally we'd do this everywhere, and switch to an explicit ReleaseGIL
annotation on functions which are slow or risky (raise exceptions, or
do something which can cause a GIL deadlock). But those are very tricky
to identify, so instead just explicitly hold the gil on cheap methods
which are likely to be called many times and could have an impact on
script performance.
are to be clipped from the project, instead of just tracking this
by the presence of any checked layers
Avoids inconsistencies between the layers which are visibly clipped
on the map vs the options which are set in the GUI.
* QgsField::ConfigurationFlags: use a negative form for the flags so that default flags is None
this will make this futureproof, so newly added flags won't cause trouble for compatibility with old projects
* follow up
* remove template class
* fix order of init
* follow up
* Revert "update test projects (open and save)"
This reverts commit c738609d9177ed398ba2a0ae0e1c16448accd5f6.
* DoNotExposeVia -> HideFrom
rename the confusing "Capitalize First Letter" option to
"Force First Letter to Capital"
This change is intended to clarify the role of the "capitalize
first letter" option, and to provide an option which actually
does what users expect the "capitalize first letter" option
to do.
Fixes#16539
Instead of ALWAYS converting all features in a tile to QGIS representations,
now we intelligently skip over any layers which aren't required for
rendering or labeling (e.g. because the current renderer/labeling
configuration is disabling these layers or doesn't have a rule for
them).
This improves rendering speed with sources like the OS ZoomStack tiles,
which have a LOT of detail even at small map scales (e.g. building
data is present in very zoomed out tiles!!).
When saving a raster expression the generated expression didn't contain square brackets around the letters used for band identifiers. Sticking with the NDVI example:
```
("NIR@1" - "Red@1") / ("NIR@1" + "Red@1")
```
becomes
```
(a - b) / (a + b)
```
Due to the way the expression is parsed these would not be interactie parameters for the user to set as the parameters requre square brackets around the layer tags. This change simply updates the string replacement to include the square brackets so you would get instead:
```
([a] - [b]) / ([a] + [b])
```
