fix: make sure to include areas with disputed regions

[12rambau]

Fix #73

Instead of loading only the identified ISO_3 I naively discover them from the geojson. I'm concerned with the ordering and need to verify that it's not introducing an artificial offset. It does not slow down the computation that much.

[julianwid]

I think by concatenating the DFs retrieved for each ISO in isos you create a mismatch between level_df and complete_df, right? Because in the latter the Z0X are only at the end of complete_df, but in level_df they may be at any row index.

[12rambau]

that's what I though, then I won't avoid the reordering.....

[julianwid]

How about this:

isos = level_gdf.GID_0.dropna().unique()

# workaround for the wrong naming convention in the geojson files
# https://gis.stackexchange.com/questions/467848/how-to-get-back-spaces-in-administrative-names-in-gadm-4-1
# it should disappear in the next version of GADM
# we are forced to retrieve all the names from the df (sourced from.gpkg) to replace the one from
# the geojson that are all in camelCase

df_list = [Names(admin=iso, content_level=content_level, complete=True) for iso in isos]
complete_df = pd.concat(df_list)
# GID columns to merge on; they (should) match exactly
shared_cols = [f"GID_{i}" for i in range(int(content_level) + 1)]
# Camel-case columns to drop
drop_cols = [f"NAME_{i}" for i in range(int(content_level) + 1)]
gdf = pd.merge(level_gdf.drop(drop_cols, axis=1), complete_df,
             how='inner', on=shared_cols)

return gdf

In my hands, this works.