Add dispatch mechanism for geocoding backends

[cosi1]

Folks,

here's a proposed geocoding mechanism that provides a set of predefined backends (currently Nominatim and Google), as well as the ability to use a custom geocoding function. It definitely may need some polishing (no unit tests or a detailed documentation so far) - think of it as of a proof of concept rather than an ultimate solution :-)

Best,
-p-

[zkamvar]

Folks,

here's a proposed geocoding mechanism that provides a set of predefined backends (currently Nominatim and Google), as well as the ability to use a custom geocoding function. It definitely may need some polishing (no unit tests or a detailed documentation so far) - think of it as of a proof of concept rather than an ultimate solution :-)

Best,
-p-

nominatim should not be in here since it's not on CRAN.

Because it seems that there are no geocoding APIs that will work anonymously, then I think we should just pick the smallest package that works and go with that. If users want the geocoding, then they will have to register for a key in any way they can.

[cosi1]

@zkamvar Well, it's in Suggested, so it's not a strict requirement, but I get your point.
If we assume that it's up to the user to geocode their locations (and get an API key if need be), we may as well stay with ggmap. The only downside of this approach is that it might be impossible (or at least difficult) to unit-test that function - unless we manually create a mockup response from Google's API and provide a testing switch in add_coordinates().

[zkamvar]

@zkamvar Well, it's in Suggested, so it's not a strict requirement, but I get your point.
If we assume that it's up to the user to geocode their locations (and get an API key if need be), we may as well stay with ggmap.

agreed.

The only downside of this approach is that it might be impossible (or at least difficult) to unit-test that function - unless we manually create a mockup response from Google's API and provide a testing switch in add_coordinates().

I believe this is the premise of the vcf or webmockr packages.

[cosi1]

@zkamvar Thanks, I didn't know that vcr had been ported to R!