add speed up demonstration of datashader
Little benchmark to demonstrate in which cases it is helpful to use datashader to increase rendering speed
Check out this pull request on ![]()
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
Thanks for the PR, this is very informative! Just a few minor comments:
- [x] I'd remove the variables
mpl_sdsandds_sdsas they are not used (also, maybe you have already seen, the developers don't recommend to compute these statistics https://docs.python.org/3/library/timeit.html#timeit.Timer.repeat). - [x] The empirical indication that for ~1000 polygons on datashader is faster is very useful! Could you please also add a section to check this for points and circles?
Thanks a lot!
- [x] Regarding the failing docs, it seems that it can be fixed by using this
mystconfiguration setting: https://github.com/executablebooks/MyST-Parser/issues/519#issuecomment-1037239655, inconf.py.
- [ ] Regarding the failing docs, it seems that it can be fixed by using this
mystconfiguration setting: WARNING reference target not found, even though reference exists executablebooks/MyST-Parser#519 (comment), inconf.py.
We actually already seem to have that. Will try to remove the HTML logic
@Sonja-Stockhaus @LucaMarconato I fixed some things that were causing warnings in other notebooks. Apparently, this fixed it.
@Sonja-Stockhaus I modified your notebook slightly to have less redundant code. For testing I reduced the number of points. Will rerun over-night with enough Points to see the effect.
However, for merging I'd wait until your main PR is merged because I want these to be real version accessible from PyPi.
@timtreis ready to merge??
Not yet. On the one hand is datashader not faster for points, so that's weird. And then I wanted to first release sonjas feature branch so that the watermark has an actual release version attached and not a GitHub commit