Hi Nyall,
I compiled your branch / pull request. Quite impressive! I like it.
Is it planned to land in master, or are there issues with this improvement?
One small issue I had, but maybe this is due to KDE, is that if I pan,
and the mouse cursor goes off the canvas area, the redraw is
paused/blocked, whereas, if the cursor stays inside the canvas area,
redraws always happen as expected. Not sure if something can be done
about this. It is also not a real problem.
Thanks to Sourcepole and you!
Andreas
On 11.07.2017 11:26, Nyall Dawson wrote:
On 11 July 2017 at 19:22, Neumann, Andreas <[email protected]> wrote:
Hi Martin,
Sounds interesting. Of course this is not urgent. We now need to concentrate
on all the stuff that requires refactoring and API changes in the master
branch.
But it would really be nice, if such rendering improvements could be
introduced in a future 3.x version. Maybe something for the next QGIS grants
program?
Andreas - do you compile master builds? I'd love it if you could give
the branch from https://github.com/qgis/QGIS/pull/4777 a test and give
feedback on the open questions in that PR. I suspect you'll be pleased
with the behavior with that change.
Full credit for the changes goes to Sourcepole - I've just forward
ported the feature from their fork and modernised the code for 3.0.
Nyall
Andreas
On 2017-07-11 10:48, Martin Dobias wrote:
Hi Andreas
On Tue, Jul 11, 2017 at 9:38 AM, Neumann, Andreas <[email protected]>
wrote:
Hi Martin,
I noticed that loading and redraw only happens when the mouse button is
released after the panning operation. While I understand the reasoning for
this (definitely on the safe side and no unnecessary redraws), I wonder, if
for WMTS, loading and display could already happen while panning and the
mouse button isn't yet released, e.g. triggered whenever a certain threshold
of panning distance (e.g. in percent of the last extent) was surpassed.
This would be even nicer than the current behaviour. Or are there technical
reasons, not to do that? Google maps works like this. It redraws while
panning.
Good news - there are no technical reasons that would prohibit us from
such improvement. Background rendering could start earlier, we just
need to add logic for such behavior. There is already a pending pull
request from Nyall to render adjacent parts of map canvas once
rendering of the visible area has finished - that may already be a
sufficient solution - I have not had time to play with it yet though.
Maybe these two approaches (render on map dragging, render outside of
current view) can be combined somehow for the best UX while keeping
overhead of extra rendering low.
We could even think about moving more towards tile-based approach to
rendering of map canvas, with the advantage of being able to cache map
tiles from previous canvas redraws for the same scale. This would
allow instant appearance of areas previously rendered with lower CPU
use at the expense of consuming more memory and having more
complicated logic for labeling.
Regards
Martin
_______________________________________________
QGIS-Developer mailing list
[email protected]
List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
