Create GeoJSON serializer

[JosiahParry]

Pull Request

This pull requests implements a geojson serializer. The serializer utilizes geojsonsf. The serializer only supports objects of class sf and sfc.

Please note that devtools::check() fails due to hookable.R#3 and it cannot find the namespace. As such I cannot provide check results without fiddling with code that is not mine.

No code was modified, only added. As such there should be no breaking features or breaking backwards compatibility.

---

At present, the only way to provide geojson—that I am aware of—via plumber API is to utilize sf to write to a temporary file. read that tempfile via readLines() and use a cat serializer to return the exact json that was written by sf. This serializer removes the need to create hacky solutions for those who would like to return geojson from a plumber API.

library(sf)
library(plumber)

nc <- st_read(system.file("shape/nc.shp", package="sf"))

#* @apiTitle GeoJSON serializer

#* Test sfc object
#* @param id The row ID to print. Must be between 1 - 100.
#* @get /sfc
#* @serializer geojson
function(id = "") {
  nc[["geometry"]][as.integer(id)]
}

#* Test sf object
#* @param id The row ID to print. Must be between 1 - 100.
#* @get /sf
#* @serializer geojson
function(id = "") {
  nc[as.integer(id),]
}

#* Test parsing of geojson
#* @parser geojson
#* @post /parse
function(req, res) {
  st_coordinates(req$body)
}

PR task list:

• Update NEWS
• Add tests
• Update documentation with devtools::document()

[CLAassistant]

All committers have signed the CLA.

[meztez]

As an alternative. You can register a serializer or a parser directly on the fly. Here is a an example using xml.

library(plumber)
library(xml2)

serializer_xml <- function(..., type = "application/xml") {
  serializer_content_type(type, function(val) {
    as.character(xml2::as_xml_document(val, ...))
  })
}
register_serializer("xml", serializer_xml)

parser_xml <- function(...) {
  parser_text(function(txt_value) {
    xml2::as_list(xml2::read_xml(txt_value, ...))
  })
}
register_parser("xml", parser_xml, fixed = c("text/xml", "application/xml"), regex = "/xml$")

#* @post /score/xml
#* @serializer xml
#* @parser xml
#* @parser multi
#* @param info:file
#* @param report:file
#* @param debug
function(info, report, req) {
  process_input(info, report)
}

[schloerke]

Notes: https://datatracker.ietf.org/doc/html/rfc7946#section-12 defines application/geo+json as the standard Content-Type for geojson data

[schloerke]

Thank you @JosiahParry !

Is there a way to close the round trip loop by adding a geojson parser? (Be able to both parse and serialize geojson information). @meztez's comment above gives a quick example on how to implement a parser.

---

If you run into more serializers/parsers that you believe are commonly used, please feel free to add another PR 😄

[JosiahParry]

@schloerke thanks for the comments!
Will do. I'll make your changes and also add the parser.

[meztez]

geojsonsf use a different R json library. I wonder if this could be avoided to reduce plumber depencies?

[schloerke]

@meztez , do you think these codes should live within the geojson pkg?

I don't remember the setup for broom's integration with other packages. I'm guessing this is closed to that situation

[JosiahParry]

@meztez No libraries were added to Imports as such the dependency tree remains the same. There is no effect on plumber dependencies.

[meztez]

@meztez No libraries were added to Imports as such the dependency tree remains the same. There is no effect on plumber dependencies.

Great! In that case, should we prepend call to geojsonsf function with a requireNamespace check?
https://github.com/rstudio/plumber/search?q=requireNamespace&type=code

@JosiahParry thanks a lot for this contribution

[JosiahParry]

All requested changes should be addressed in most recent commit.

[JosiahParry]

Exciting!!! Thanks all for your guidance :)