Kelsey Jordahl

@mrocklin I agree with @jorisvandenbossche that there is some low-hanging fruit for optimization. The initial implementation of GeoPandas focused on the API, and I believe that is fairly stable now....

@kuanb I am thinking of a dask-GeoDataFrame backed by PostGIS with something like [`dask.dataframe.read_sql_table`](http://dask.pydata.org/en/latest/dataframe-api.html#dask.dataframe.read_sql_table) to have a deferred implementation . That seems doable if the data are already in PostGIS....

[Travis says OK](https://travis-ci.org/enthought/chaco/builds/18108207)

@jonathanrocher Thanks for the ping, obvioussly this had dropped off my radar. I'll try to pick it up again.

Sorry I missed seeing this issue before the tutorial. Glad you figured it out.

We use astropy internally at Planet Labs. Not because we are doing astronomy (though we are imaging the Earth from space), but because the NaN treatment in [astropy.convolution.convolve()](http://docs.astropy.org/en/stable/api/astropy.convolution.convolve.html#astropy.convolution.convolve) is so...