Now we abort and advise the user if insufficient memory is available
to perform the calculation.
At some future stage it would be nice to perform the calculations
in blocks to allow this scenario, but for now it's better not to
crash.
copy/= operators or making them private
This revealed (and fixes) some issues, including a potential crash
using server access control (refs #13919), and a potential crash with
diagrams
Now all classes and members are either exposed to bindings or marked
as "not available in Python bindings" in the docs.
Drop test thresholds to 0. Now it should be much easier to determine
what missing members have been added which are causing test
failures.
Rationale:
- there was a lot of large objects passed by value, so potentially
there's a speed bump from this
- even for implicitly shared classes like QString/QList there's still
a (small) cost for copying the objects when there's no reason to
- it's the right thing to do!
...rather than reading input layers directly through GDAL.
Benefits include more robust handling of nodata/data type conversions,
less code duplication, also being able to take advantage of features
in QGIS raster code like handling gain/offset in rasters. (fix#12450)
Also, add a choice of output projection to the raster calculator.
Previously the output CRS would be taken from the first raster, with
no guarantees that the output extent matched the output CRS. This
resulted in empty/misplaced rasters. (fix#3649)
...for more accurate calculations (fix#9081)
Also:
- Add test suite for raster calculator
- Fix errors with log/log10 calculation and inputs <= 0
- Fix loss of nodata values in certain circumstances
* GDALGetGeoTransform failures can be ignored (partly reverts 0080f9e)
* handle quoted raster references
* don't complain if the output file is not yet set
