[Question] Representation of empty fields in lake package construction

[bdestombe]

I have a geojson that contains all the configurations for the lakes of my model. Many of the configurations are strings. When stored to geojson and later loaded, the empty fields are represented by empty strings. To be expected I would say.

However, in the code the methods .isna(), notna() are used and they don't mark "" as na.

pd.isna("")
    False

lake_gdf["lakeout"].isna() | lake_gdf["lakeout"].eq("") could be used instead. Or on the user side do gdf.replace("", None).

An example are the boundary names in lakeout column. How are you guys handling the columns with a string data type and missing values?

My goejson:
https://github.com/NHFLO/data/blob/main/src/nhflodata/data/mockup/lakes_pwn/v1.0.0/lakes_pwn.geojson

nlmod/nlmod/gwf/lake.py

Line 211 in 736bc46

if use_outlets and (not lake_gdf["lakeout"].isna().all()):

nlmod/nlmod/gwf/lake.py

Line 274 in 736bc46

if pd.isna(datavar): # None or nan

[OnnoEbbens]

I have not encountered this before but I like your suggestion of using lake_gdf["lakeout"].isna() | lake_gdf["lakeout"].eq(""). We could also leave it to the user but they will probably run into errors before realizing they have to replace empty strings with None.

Also I cannot think of a usecase for empty strings so I am in favor of treating empty str values the same as nan values.

[bdestombe]

Thanks! I'll leave the issue open, until implemented

[rubencalje]

I also think the check for "" is a good solution. This is implemented in 3bd6965 and a849456 right? Can we close the issue?