Sunsetting splot
As per PySAL dev team discussions, splot is scheduled to be archived in future. Functionality for plotting of Moran statistics has been moved to esda to the module where the statistics itself is implemented. Same is happening over in giddy.
A few questions on what should happen next.
- What should happen with VBA plotting and related tools? There's no clear home for it anywhere else in the PySAL I think.
- What about convenience functions for subplots?
-
plot_local_autocorrelation- https://github.com/pysal/esda/pull/360 -
plot_moran<- this is easy subplot, probably not worth exposing -
plot_moran_bv<- as above -
moran_facet- https://github.com/pysal/esda/pull/359
- How to approach sunsetting? Deprecation notices in splot pointing to esda? Deprecation notice on import?
cc @pysal/devs
- What should happen with VBA plotting and related tools?
Perhaps libpysal if nothing else fits?
- What about convenience functions for subplots?
Is there any drawback to including at least [1] & [4] in esda?
- How to approach sunsetting? Deprecation notices in splot pointing to esda? Deprecation notice on import?
I'd vote for deprecation notice on import.
Is there any drawback to including at least [1] & [4] in esda?
Not really, just more work :).
I just realised that there is also some stuff that uses Bokeh for plotting in _viz_bokeh.py that is not present in the API reference. Should that be moved somewhere?
I'd vote for deprecation notice on import.
same
-
I believe that esda is the best fit for VBA plotting. While it's not tied to ESDA packages, exploratory spatial data analysis aligns conceptually with VBA plotting.
-
Adding both [1] and [4] would be great enhancement for esda
-
I agree with James and Eli.
My two cents, in case they’re helpful in any way:
What should happen with VBA plotting and related tools? There's no clear home for it anywhere else in the PySAL I think.
My first thought was mapclassify with all functionality optional and conditional on the viz libs installed (I know there was a lot of work to keep the library pretty lightweight and all focused on classification only. But my sense would be that VBA is a way of classifying that is bi-variate.
If that doesn’t work for some reason (e.g., too much work), esda as suggested would be a fitting home too.
What about convenience functions for subplots?
My sense would be esda as most (all?) functionality falls within exploratory data analysis and, in particular, analysis of spatial autocorrelation patterns.
+1 on moving the vba classification into mapclassify
Alright, I have opened remaining PRs for esda and am happy to finish draft in giddy as soon as I get some feedback there.
Happy with any decision regarding VBA but will probably leave that for someone else to tackle.
I just realised that there is also some stuff that uses Bokeh for plotting in
_viz_bokeh.pythat is not present in the API reference. Should that be moved somewhere?
Just remembering this. I suppose we should decide whether in integrate this bokeh functionality into the "home" packages (where possible) or simply sunset it.
I'd vote for deprecation notice on import.
same
Also seems like we never got to the deprecation notice on import. I can put in a PR for that, but perhaps we should wait until we handle creating migration notebook(s) and figure out the Bokeh situation.