Many thanks for this highly useful effort. A couple of thoughts/questions that could be converted to tickets from my side:
GeoJSON is a subset of / extends JSON support, but by having to explicitly initialise it (rather than load as a separate module, let's say) it seems to be treated on par. Would we do the same for TSV? SQLite? GeoPackage? But I guess it's too early to worry about scalability when there are only 3 formats supported at this time.
On this note, the introduction of default_serializer to file_formats.py might warrant it's own PR, or at least a check through the other code and documentation.
What about auto-detecting a GeoJSON file?
We should link to RFC 7946 or similar. To me this effort is also closely related to the effort to create a Geo Data Package. Currently the README doesn't really cover file format support, but it should.
We might want to stick to a .geojson file extension in the example, though make it clear that this is not required.
Where does validation happen? I.e. where would you hand off to a more specialized library like python-geojson.
Many thanks for this highly useful effort. A couple of thoughts/questions that could be converted to tickets from my side:
default_serializer
to file_formats.py might warrant it's own PR, or at least a check through the other code and documentation..geojson
file extension in the example, though make it clear that this is not required.Originally posted by @loleg in https://github.com/datahq/dataflows/issues/153#issuecomment-749396962