[Qgis-psc] QGIS 2.18 bug fixing effort

[Nyall Dawson]

On 16 May 2017 at 18:57, Giovanni Manghi <giovanni.manghi at gmail.com> wrote:

> We *really* need to make something about the geoprocessing tools that
> *still* are returning incomplete/wrong results in a semi-transparent
> way (no, the common user do not look the log). This was last discussed
> in Las Palmas but afaik not much has been done in this sense. I have
> already proven that with purely SPATIAL SQL tools (via ogr2ogr) the
> results are more reliable and considerably faster (tests made with
> dissolve and clip, but I could add any other operation).

This is already fixed in 3.0. Now the default behaviour is to
terminate an algorithm when encountering an invalid geometry
(controllable by users - this behavior has a significant performance
impact!)

> Would each
> copy of spatialite shipped with QGIS be compiled with support for
> ST_Makavalid I could add a check to have the geometries fixed
> beforehand and in a complete transparent way (bad geometries are the
> real problem for the tools we used to call "ftools").

That's not necessary - makeValid is available in the QGIS geometry
class in 3.0 now. We could potentially add an option to "auto repair"
invalid geometries in processing too (...but performance impact etc
etc...).

On that note...

Unfortunately I've got some MAJOR breakage coming soon for processing
3.0 as I finish off the last part of the c++ conversion. It's
definitely going to break algorithms, and we'll need lots of testing
before release to make sure I can fix any regressions which occur.
Things are going to get a lot rougher here before they get better!

Nyall




>
>
> Cheers
>
> -- G --
>
>
>
>
>
>
>
>
>
>
> On Fri, May 5, 2017 at 10:02 AM, Andreas Neumann <andreas at qgis.org> wrote:
>> Hi Alex,
>>
>> Thank you for your reply.
>>
>> I'll cc Giovanni.
>>
>> Giovanni - can you please provide Alex a list of processing/analysis related
>> bugs/regressions - the most pressing ones?
>>
>> Alex: we'll be happy to pay for a couple of hours. We would be interested in
>> a regular relationship with core devs who are available to fix the most
>> pressing issues prior to every release and potentially also after.
>>
>> Can you estimate - how many hours or days you can dedicate to fixing issues?
>> I mean within the next 1-2 months and for QGIS 2.18
>>
>> I'll add you to our list of core devs who participate in our bug fixing
>> effort then -
>> https://docs.google.com/spreadsheets/d/1Cdh-6eAmVCdiZIQGMhu_Aqgg09tpxsC40BGXQJtYc54/edit#gid=1281977200
>>
>> Thanks,
>> Andreas
>>
>>
>> On 3 May 2017 at 18:39, Alexander Bruy <alexander.bruy at gmail.com> wrote:
>>>
>>> Hi Andreas,
>>>
>>> 2017-05-03 13:10 GMT+03:00 Andreas Neumann <andreas at qgis.org>:
>>> > I wonder if you would be interested / would have time - to assist us
>>> > with
>>> > our QGIS 2.18 bug fixing efforts - esp. with the processing issues?
>>>
>>> Sure I will be happy to help with this as much as I can.
>>>
>>> > We would pay some devs for up to 5 days (total 40h per dev) with a fixed
>>> > hourly rate of 100 €. Some devs can only contribute a day or two - which
>>> > is
>>> > also fine.
>>>
>>> I would like to donate my time to project.
>>>
>>> --
>>> Alexander Bruy
>>
>>
>>
>>
>> --
>>
>> --
>> Andreas Neumann
>> QGIS.ORG board member (treasurer)
> _______________________________________________
> Qgis-psc mailing list
> Qgis-psc at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/qgis-psc