[QGIS-Developer] Disappearing -180 / 180 longitude grid lines on certain zoom levels

[Andreas Neumann]

Hi Even,

I just created an issue: https://issues.qgis.org/issues/19746

and assigned you - if this is ok. Feel free to remove yourself if you 
can't work on it.

I also added Richard's related issues for reference.

Thanks a lot in advance!

Andreas


Am 29.08.2018 um 21:58 schrieb Even Rouault:
> On mercredi 29 août 2018 21:50:25 CEST Andreas Neumann wrote:
>> Hi Even,
>>
>> Thank you very much for the detailed analysis of the spatial filter
>> issue we have here, that causes the disappearing grid lines.
>>
>> Would you be interested to work on fixing this issue on our bug fixing
>> program? This problem annoyed me for quite some time. And it appears on
>> different world projections.
> I might have a look at this, if nobody else more familiar with the relevant
> code looks at it. Is there a ticket filed about that ?
>
> Even
>
>> Thanks,
>>
>> Andreas
>>
>> Am 28.08.2018 um 22:19 schrieb Even Rouault:
>>> Yes, seems restricted to reprojection cases
>>>
>>> This seems to be an issue with the spatial filter issued to OGR
>>>
>>> At the zooms where the lines disappear, there are requests like:
>>>
>>> Thread 23 "Thread (pooled)" hit Breakpoint 2, OGR_L_SetSpatialFilterRect
>>> (hLayer=0x7f81180c90a0, dfMinX=-179.79163612932865,
>>> dfMinY=-69.446164378986353, dfMaxX=179.90530755284408,
>>> dfMaxY=78.959077253477474) at ogrlayer.cpp:1223
>>>
>>> At the zooms where that work (even when zoomed in), there are like:
>>>
>>> Thread 29 "Thread (pooled)" hit Breakpoint 2, OGR_L_SetSpatialFilterRect
>>> (hLayer=0x7f81180c90a0, dfMinX=-180, dfMinY=-90, dfMaxX=180, dfMaxY=90) at
>>> ogrlayer.cpp:1223
>>>
>>>
>>> I haven't looked at the QGIS code that computes this bounding box, but
>>> from my experience with gdalwarp which has similar challenges, it is
>>> tricky to compute a source bounding box from a target bounding box,
>>> because sometimes the coordinates in the target bounding box do not
>>> correspond to a physical point on Eath, and hence inverse projection
>>> fails. So you have to resort to a grid sampling approach, but that makes
>>> you miss the exact boundaries. So probably that a band-aid fix would be
>>> to add some ad-hoc logic, like "if the source SRS is long/lat, and the
>>> computed extent is almost worldwide, then extend it to full worlwide (or
>>> do not emit a spatial filter at all)"
>>>
>>> Even
>>>
>>>> With EPSG:4326 it does not seem to happen, but I tried EPSG:3857 and the
>>>> -180,180 grid lines do appear and disappear at different zoom levels. I
>>>> also confirm this with the Robinson projection with the addition that the
>>>> -90, 90 degree latitude lines also appear and disappear.
>>>>
>>>> Calvin
>>>>
>>>> On Tue, Aug 28, 2018 at 3:31 PM, Andreas Neumann <a.neumann at carto.net>
>>>>
>>>> wrote:
>>>>> Hi again,
>>>>>
>>>>> For some time already I noticed that, depending on the zoom level, the
>>>>> -180 / 180 degree grid lines appear / disappear in QGIS. This is
>>>>> unrelated
>>>>> to the EqualEarth projection just discussed and also appears on other
>>>>> world
>>>>> projections, like the Robinson projection.
>>>>>
>>>>> Here is the testfile I used with world (countries) and gridlines:
>>>>> http://www.carto.net/neumann/temp/gridlines.gpkg
>>>>>
>>>>> Can you reproduce this behaviour?
>>>>>
>>>>> Any idea why this happens and how this could be fixed?
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Andreas
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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
>