[QGIS-Developer] What to do about incorrect SAGA algorithms?

[Pedro Venâncio]

Hi Nyall,

I think this rises another problem, that is the use of a very old version
of SAGA (2.3.2) in QGIS.

I've just tested Difference algorithm with Lene dataset in native SAGA
2.3.2 and 7.2.0, and 2.3.2 gives the problem described, but 7.2.0 gives the
correct result:

https://cld.pt/dl/download/2dfb0db5-eb74-4396-ad18-7451fd1fe788/test_saga_232.jpg

https://cld.pt/dl/download/92182907-d37a-4441-891c-3c85064b3100/test_saga_720.jpg

So, we are faced again with the question of keep an old version of SAGA in
QGIS core, or make it as an external plugin, with someone keeping an eye in
the changes introduced by SAGA team between versions (as we saw in the
past).

Alexander Bruy had made an effort to keep a parallel SAGA plugin that
support newer SAGA versions: https://plugins.bruy.me/processing-saga.html
But it is not in the official repository. Maybe this is the time to discuss
again the use of SAGA as an external plugin?

Best regards,
Pedro Venâncio




Nyall Dawson <nyall.dawson at gmail.com> escreveu no dia segunda, 4/03/2019
à(s) 11:11:

> Hi list,
>
> Just raising the discussion about what we should do with SAGA
> Processing algorithms when the results generated by SAGA are known to
> be incorrect (i.e. there's a bug in SAGA itself).
>
> Specifically, the SAGA difference and symmetric difference algorithms
> generate incorrect results. See https://issues.qgis.org/issues/21354,
> and the results from
>
> https://github.com/qgis/QGIS/blob/master/python/plugins/processing/tests/testdata/saga_algorithm_tests.yaml#L318
> . This seems to be a general problem with the algorithms, not isolated
> to particular input data.
>
> My gut feeling is: we should "deprecate" these algorithms (which means
> they continue to work for existing models, but are hidden from the UI
> and cannot be run from the toolbox or added to new models). I don't
> see the value in exposing broken algorithms when we have robust
> alternatives (the native QGIS difference/symmetric difference
> algorithms are very fast, stable, and well tested), and potentially a
> lot of harm.
>
> Thoughts?
>
> Nyall
> _______________________________________________
> QGIS-Developer mailing list
> QGIS-Developer at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20190304/d6ef3c87/attachment-0001.html>