GeoPandas: Contribute enhancements, look up bugs already fixed

[emiliom]

• It'd be great if a new GeoPandas release were issued at least a week before GeoHackweek! Last release was July 2016. Sigh. Can we help make that happen? I doubt it, but still. The new release is being discussed in this issue
• Implement this enhancement, ".from_features() does not accept FeatureCollection". Submitted PR ENH: GeoDataFrame.from_features to accept FeatureCollection as well geopandas/geopandas#507 on 8/22
• I wonder if the plotting overlay problem we found is related to this issue. 8/22: it looks like it's fixed in master! See Better plotting defaults geopandas/geopandas#318 (comment)
• choropleth plot legend glitch: first legend class always defaulted to 0 rather than the actual min value. The bug is fixed in the GeoPandas 0.3
• The bug with crs not carried over by the dissolve method has been fixed!
• "Bring SRID in from_postgis()"
• Double check some methods (eg, to_file(), from_postgis vs read_postgis()) to confirm that:
  1. I'm using them consistently across the notebooks
  2. Their usage matches currently accepted preferences, rather than deprecated methods or sub-optimal usage pattern

I should probably make a list of the bugs or questions I identified in my tutorial notebooks last year ...

[jorisvandenbossche]

If that reassures you, I am planning to do a release before EuroScipy (which is end of August, I am also giving a talk there about GeoPandas).
BTW, help is certainly appreciated! (geopandas can use more contributors) Maybe not necessarily anymore for making this release come sooner, but for the next one :-)
(although this issue can use feedback: geopandas/geopandas#318)

And contributing doesn't necessarily be to do actual coding, also giving feedback / opening issues about bugs / usability problems / enhancement requests you encounter during tutorials are very useful. (both are welcome!)

Further, if you are worried/interested in the performance of geopandas, you can check out geopandas/geopandas#473 (the issue is about initial version of complete rewrite of internals of geopandas that should make it much faster)

(sorry to hijack the issue here :-))

[emiliom]

@jorisvandenbossche, thanks for chiming in! Please, hijack away.

If that reassures you, I am planning to do a release before EuroScipy (which is end of August, I am also giving a talk there about GeoPandas).

Yes, this is very reassuring! Thanks so much. I look forward to seeing your presentation materials after EuroSciPy, and seeing if I can borrow something. Please feel free to grab anything from my tutorial materials (two notebooks plus markdown pages).

I've gone ahead and commented on geopandas/geopandas#318. Hopefully it'll be of some use, and even better if someone points out a way to solve my problem.

I'll actually try to submit a PR to address geopandas/geopandas#225 early next week. I've been looking into it and I think I know exactly what to do.

Further, if you are worried/interested in the performance of geopandas, you can check out geopandas/geopandas#473

For my current usage patterns, performance is not a big deal. But it's great to hear about this cython work, and I've also heard about the Dask work (https://github.com/geopandas/geopandas/issues/461). My usage could definitely change in the future, depending on projects I'm involved with.

[mrocklin]

For my current usage patterns, performance is not a big deal. But it's great to hear about this cython work, and I've also heard about the Dask work (geopandas/geopandas#461). My usage could definitely change in the future, depending on projects I'm involved with.

@emiliom if the cython work is easily usable by that time (which I think is a decent bet) then benchmarking and performance testing would be very useful for the project. It would also be interesting to have people compare performance against other common tools like PostGIS, QGIS, and ArcGIS.

[emiliom]

@emiliom if the cython work is easily usable by that time (which I think is a decent bet) then benchmarking and performance testing would be very useful for the project. It would also be interesting to have people compare performance against other common tools like PostGIS, QGIS, and ArcGIS.

Ok, will keep it in mind. I doubt I can offer any real benchmarking and performance testing other than anecdotal, qualitative feedback, by the time of our event (mid September) or during the event. But maybe afterwards we can think of something