Trajectory.is_latlon() update for GeoPandas 0.7

[martinfleis]

Geopandas 0.7 will come with pyproj.CRS class under gdf.crs, so following will not work in the next release as it expects self.crs to be a dict. See geopandas/geopandas#1101 for details.

https://github.com/anitagraser/movingpandas/blob/01417e5b2ab8e3b06f8060cbcee934673ea7b4e4/movingpandas/trajectory.py#L115-L126

This fix ensures compatibility with future releases of geopandas as well as older ones. crs.is_geographic checks for any lon/lat projection, not only WGS84. I assume that it is expected behaviour anyway.

cc pyOpenSci/software-submission#18

[anitagraser]

Excellent, thank you!

[anitagraser]

I just noticed that this change significantly slows down the creation of Trajectory objects. Cell 20 in 0_getting_started.ipynb used to finish in < 1s but with this change it takes > 1min. (Similar slow-downs occur in all other notebooks.)

[anitagraser]

I'm replacing all calls to this function with access to an object variable which is computed once in the constructor ...

[martinfleis]

That's because the check is done for ever single row instead of the whole gdf at once. Solution might be having Trajectory.is_latlon attribute doing this check at _init and then just checking for True/False instead of if self.is_latlon().

[martinfleis]

I'm replacing all calls to this function with access to an object variable which is computed once in the constructor ...

@anitagraser yeah, you were faster :)