sfheaders

[dcooley]

This isn't an issue, but I found this repo after your issue on jsonify and thought another of my libraries may be of use - sfheaders

like rapidjsonr and jsonify you can link to it and call the Rcpp code directly. Of interest may be the sfc_point() (plus a few overloaded functions), which will create the sfc object, without you needing to require sf.

Hope that helps.

[knapply]

Thanks @dcooley . I saw that on Twitter or elsewhere at some point, but couldn't remember the name when I was looking for it.

It's unclear how useful the spatial piece will be. All I do right now is take take the JSON bounding box...

[
  [
    -63.784544,
    -35.689402
  ],
  [
    -63.784544,
    -35.641737
  ],
  [
    -63.724717,
    -35.641737
  ],
  [
    -63.724717,
    -35.689402
  ]
]

... and flatten it into a vector.

[1] -63.78454 -35.68940 -63.78454 -35.64174 -63.72472 -35.64174 -63.72472 -35.68940

The only difference between what I do and what {rtweet} does is that I leave empty coordinates as empty vectors instead of having 8 NA_real_s hanging out in most rows.

I haven't pushed the latest work, but the only sf "functionality" I'm using in C++ is the prep_bbox_() function that turns non-empty vectors into matrices (and close the polygon), then slap the appropriate classes on it.

[[1]]
          [,1]      [,2]
[1,] -63.78454 -35.68940
[2,] -63.78454 -35.64174
[3,] -63.72472 -35.64174
[4,] -63.72472 -35.68940
[5,] -63.78454 -35.68940

attr(,"class")
[1] "XY"      "POLYGON" "sfg"    

All of that said, I'm trying to lock in some more critical things at the moment. My C++ is improving, but it's an uphill battle.

I'll eventually re-investigate what can be done spatial-wise and probably bring {sfheaders} in. Thanks again for pointing me to it.