TST: fix inheritance class type error #646
Conversation
NinaWie
left a comment
There was a problem hiding this comment.
Amazing that you found this! Your argument makes sense to me and this is a proper fix (you won over ChatGPT ;) )
I'm slightly confused about the sp = gpd.GeoDataFrame(staypoints.copy()) - I don't understand how this additional casting can avoid calling validate twice, but I guess you tested this. However, it seems like something that would also happen for positionfixes / locations etc in other parts of the code.
Thanks for the check! The casting is for avoiding errors for later processing functions, e.g. I found that validate will be called when calling dissolve function I.e. line 145, which will raise an error as the intermediate data frames does not necessary fulfill our sp definition (missing columns for example). I thus think casting to geodataframe makes sense, but have to admit we need to check why certain functions will trigger validate while others don’t. |
|
Okay makes sense. Yes it's a bit mysterious right now why validate is triggered, but we can look at this separately. |
closes #644
The issue is raised due to a recent update in the GeoDataFrame function
_geodataframe_constructor_with_fallback()(see here), where instead of the hardcodedGeoDataFrame(line 54), a class instanceclsis passed to the function and implicitly initialized (line 2026). This means we cannot passself, but instead need to explicitly pass and callGeoDataFrameto prevent initializing a trackintel class. The latter will raise an error because the geometry link is only assigned at the very end of theGeoDataFrame._constructor_from_mgr()function (see here).The PR also contains a minor fix to explicitly convert sp into GeoDataFrame before processing, otherwise the
validatefunction will be called multiple times.