when snapping line layers
Because the default behavior of the snapper is to insert extra
vertices into the snapped geometry in order to make it 'follow'
the reference geometries exactly, this can result in unwanted
results for line layers where the resultant snapped layer
has overlapping line segments.
Since we can't always know what the desired result is that the
user wants (maybe they do want overlapping lines), instead
give them control over the result by exposing extra enum
options which never insert extra vertices.
Because:
- it's unused in master, and is more code to maintain
just for possible use by plugins
- it's unmaintained, and has had no work done (beside compilation
fixes) in the recent past
- there's no unit tests or detailed documentation to show
how the class should be used
Instead of just forcing writing the triangulation to a shapefile (boo!)
change the parameter to use a QgsFeatureSink, so that anything
which implements the QgsFeatureSink interface can be used for
storing the triangulation.
These classes still need a lot of cleanup (e.g. use of proper
3d geometry classes instead of their own 3d line classes, etc)
and we don't want them locked into the 3.0 API.
Better to remove them from the API and reintroduce them after
they have been cleaned up (in >3.0). They are mostly implementation
details anyway, and unlikely to be used outside of the high
level interpolation classes.
We don't want these part of stable API - this namespace really shouldn't
exist, as it mostly contains code which is duplicated in other
parts of the API (with better maintained and tested versions).
These classes are unused in the master QGIS code, and are
unmaintained and with no unit tests or other QA, and
have inflexible API (e.g. always requiring writing outputs
to shapefiles)
They all have equivalent algorithms available via Processing
(where the algorithms are unit tested and maintained). We should
be pushing all QGIS api users to use the Processing algorithms
instead.
Previously this was only enabled for optional arguments (i.e.
those with default values). Enabling them for all arguments
allows for more readable PyQGIS code, and there seems no
downside given that we already have this support partly enabled.
The consequence of this change is that when 3.0 API is frozen
the freeze must also include the naming of function arguments,
since that's effectively now part of public API.
