Separate out data into a separate package

[hadley]

Maybe rworldmap data? Or naturalearth?

It could also contain the code used to do processing from the raw data to the appropriate sp objects. If you could point me to that code, I'd be happy to start the ball rolling

[andysouth]

I had previously wondered about calling it rcountries, but happy to consider other options. I had thought it would be good to include the country synonyms joining stuff with modernised code.

I did have a reproducible workflow from the natural earth data to the sp objects, it's here : https://github.com/AndySouth/rworldmapSetup/blob/master/saveMapPolygons.r

I recently broke that convention for a user request about Western Sahara & Somaliland.
6fa5db1

What has stopped me from updating to the more recent Natural Earth boundaries is that the field names changed which could break a bunch of rworldmap examples, and I don't really have time to fix all that.

Suggestions appreciated.

[hadley]

I quite like naturalearth as a name as it makes it clear where the data is coming from - if you're aware of the natural earth data it's useful to be able to quickly find the corresponding R package.

If the field names have changed, then maybe it's reasonable to make a clean break and create a data only package, and then later figure out how to update rworldmap to use the new data?

I can help out with some of the steps in the data processing. I think it should be possible to automate the downloads, and hence maybe include some more of the natural earth datasets. If the package is data only, CRAN is usually not so strict about the size limitations. I think there's probably also a better solution to the encoding problem, and I'm happy to look into it.

I'd also be tempted to remove as much of the fix-up code as possible, instead relying on the quality of Natural Earth data to improve over time. (And encourage people to report problems upstream)

[hadley]

I think it's also worth thinking through what to do with the attached data - is it better to keep everything, or just keep (say) the ISO3 code, and then provide another data frame that makes it easier to link to other naming schemes.

[andysouth]

That all sounds good. Particularly the clean break and making naturalearth findable. Also as you say there are a whole load of fields from the shapefiles that aren't needed and could be left out. For the name how about rnaturalearth ? Partly because that word doesn't exist yet so would be easy to track it.

[hadley]

Sure, sounds good to me.

[andysouth]

Plus would it be good to see if the rOpenSci folks would like to be involved ? I've been wanting an excuse to get involved. What's best to do, get it started and go from there ?

[hadley]

@karthik how do we get started?

[karthik]

Plus would it be good to see if the rOpenSci folks would like to be involved?

@andysouth 👍

@hadley Sounds great. If you start a repo on either of your accounts and drop me a note, we can go from there. Once the package is completed, we have a quick review and onboarding process to get it into the ropensci suite.

[hrbrmstr]

So, would this also move the map data out of ggplot2? Count me in for this either way. I can fire up an rnaturalearth repo this weekend and start some of the migration.

[andysouth]

Thanks @karthik @hadley @hrbrmstr.

I just created the repo at https://github.com/AndySouth/rnaturalearth.

Do add youself in as you want. I'll start to outline tomorrow what I think should be in. Is the readme a good place to do that or are there any other recommended practices for planning ?

[andysouth]

See : ropensci/rnaturalearth#2