Skip to contents

arcpbf is an R package that processes Esri FeatureCollection Protocol Buffers. It is written in Rust and powered by the extendr library.

arcpbf has functions for reading protocol buffer (pbf) results from an ArcGIS REST API result. pbf results are returned when f=pbf in a query.

The package is extremely lightweight and fast.

Limitation: this package does not support Z and M dimensions at this point.

TL;DR

Developer Note: Rust must be installed to compile the package. Run the one line installation instructions at https://rustup.rs/. To verify your Rust installation is compatible, run rextendr::rust_sitrep(). That’s it.

PBF support

Note that only the FeatureCollection pbf specification is supported by arcpbf. If you want to process OSM pbf files use osmextract::oe_read(). Or, if you want to create and read arbitrary protocol buffers directly in R, use RprotoBuf.

Basic usage

In most cases, we will be processing a protocol buffer directly from an http request created with {httr2}.

library(arcpbf)

# specify url to sent our request to
url <- "https://services.arcgis.com/P3ePLMYs2RVChkJx/arcgis/rest/services/ACS_Population_by_Race_and_Hispanic_Origin_Boundaries/FeatureServer/2/query?where=1=1&outFields=objectid&resultRecordCount=10&f=pbf&token="
req <- httr2::request(url)
resp <- httr2::req_perform(req)

resp
#> <httr2_response>
#> GET
#> https://services.arcgis.com/P3ePLMYs2RVChkJx/arcgis/rest/services/ACS_Population_by_Race_and_Hispanic_Origin_Boundaries/FeatureServer/2/query?where=1=1&outFields=objectid&resultRecordCount=10&f=pbf&token=
#> Status: 200 OK
#> Content-Type: application/x-protobuf
#> Body: In memory (60102 bytes)

We can process request responses with resp_body_pbf(). Post-processing steps are applied by default. The arguments post_process and use_sf are TRUE by default.

resp_body_pbf(resp)
#> Simple feature collection with 10 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17253470 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#>    OBJECTID                       geometry
#> 1         1 POLYGON ((-17264972 2244291...
#> 2         2 POLYGON ((-17264972 2244291...
#> 3         3 POLYGON ((-17264587 2241560...
#> 4         4 POLYGON ((-17263053 2239296...
#> 5         5 POLYGON ((-17261894 2236947...
#> 6         6 POLYGON ((-17262143 2241010...
#> 7         7 POLYGON ((-17261575 2233733...
#> 8         8 POLYGON ((-17263397 2234419...
#> 9         9 POLYGON ((-17269918 2239302...
#> 10       10 POLYGON ((-17265323 2239121...

Multiple response objects

When running multiple requests in parallel using httr2::req_perform_parallel() the responses are returned as a list of responses. resps_data_pbf() processes the responses in a vectorized manner.

# create a list of requests
reqs <- replicate(5, req, simplify = FALSE)
# perform them in parallel
resps <- httr2::req_perform_parallel(reqs)
#> Iterating ■■■■■■■ 20% | ETA: 7sIterating ■■■■■■■■■■■■■ 40% | ETA: 6sIterating
#> ■■■■■■■■■■■■■■■■■■■■■■■■■ 80% | ETA: 1s

# process the responses 
resps_data_pbf(resps)
#> Simple feature collection with 50 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17253470 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#> First 10 features:
#>    OBJECTID                       geometry
#> 1         1 POLYGON ((-17264972 2244291...
#> 2         2 POLYGON ((-17264972 2244291...
#> 3         3 POLYGON ((-17264587 2241560...
#> 4         4 POLYGON ((-17263053 2239296...
#> 5         5 POLYGON ((-17261894 2236947...
#> 6         6 POLYGON ((-17262143 2241010...
#> 7         7 POLYGON ((-17261575 2233733...
#> 8         8 POLYGON ((-17263397 2234419...
#> 9         9 POLYGON ((-17269918 2239302...
#> 10       10 POLYGON ((-17265323 2239121...

Reading from a file

In some cases you may have a file on disk that you want to process a pbf from. Use read_pbf() to do so. Again, post-processing steps are applied by default.

fp <- system.file("small-points.pbf", package = "arcpbf")
read_pbf(fp)
#> Simple feature collection with 2 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17260020 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#>          County                       geometry
#> 1 Hawaii County POLYGON ((-17264972 2244291...
#> 2 Hawaii County POLYGON ((-17264972 2244291...

FeatureCollection Result Types

There are three types of PBF FeatureCollection responses that may be returned as a result of a Feature Service Query request.

  • Feature Results:
    • the default query response type. Contains individual features with their attributes and geometries if available.
  • Count Result:
    • returned when returnCountOnly=true in an API request. Returned as a scalar integer vector.
  • Object ID Result:
    • returned when returnIdsOnly=true. A data.frame containing object IDs where the column name is set to the object ID field name of the feature service.

Feature Results

Feature results can either omit geometry entirely, for example in the case of a Table or when the query parameter returnGeometry=false, or include it. When geometry is omitted entirely, the response is processed as a simple data.frame. However, if the response does contain geometry, the response is a bit more complex.

Unprocessed feature results with geometries return a named list with 3 elements:

  • attributes:
    • a data.frame of the fields and their values
  • sr:
    • a named list with elements wkt, wkid, latest_wkid, vcs_wkid, and latest_vcs_wkid. These determine the coordinate reference system of the response as well as the vertical coordinate reference system.
  • geometry:
    • an sfc object without a computed bounding box or coordinate reference system set or a CRS set.
# read an example pbf without post-processing
fc_fp <- system.file("small-points.pbf", package = "arcpbf")
res <- read_pbf(fc_fp, post_process = FALSE)

res
#> $attributes
#>          County
#> 1 Hawaii County
#> 2 Hawaii County
#> 
#> $geometry
#> Geometry set for 2 features 
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: NA ymin: NA xmax: NA ymax: NA
#> CRS:           NA
#> POLYGON ((-17264972 2244291, -17264988 2244297,...
#> POLYGON ((-17264972 2244291, -17264967 2244286,...
#> 
#> $sr
#> $sr$wkt
#> [1] NA
#> 
#> $sr$wkid
#> [1] 102100
#> 
#> $sr$latest_wkid
#> [1] 3857
#> 
#> $sr$vcs_wkid
#> [1] NA
#> 
#> $sr$latest_vcs_wkid
#> [1] NA

When post-processing is applied to a geometry Feature Result, the CRS is set and the bounding box is computed. This requires the sf package to be available.

post_process_pbf(res)
#> Simple feature collection with 2 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17260020 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#>          County                       geometry
#> 1 Hawaii County POLYGON ((-17264972 2244291...
#> 2 Hawaii County POLYGON ((-17264972 2244291...

Lower level functions

The function open_pbf() will read a pbf file into a raw vector which can be passed to process_pbf(). In general you will not need this function, but it is handy for the sake of example.

pbf_raw <- open_pbf(fc_fp)
head(pbf_raw, 20)
#>  [1] 12 ac fd 01 0a a8 fd 01 0a 08 4f 42 4a 45 43 54 49 44 12 0c

This raw vector can be turned into an R object using process_pbf(). The output will not be post processed.

res <- process_pbf(pbf_raw)
res
#> $attributes
#>          County
#> 1 Hawaii County
#> 2 Hawaii County
#> 
#> $geometry
#> Geometry set for 2 features 
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: NA ymin: NA xmax: NA ymax: NA
#> CRS:           NA
#> POLYGON ((-17264972 2244291, -17264988 2244297,...
#> POLYGON ((-17264972 2244291, -17264967 2244286,...
#> 
#> $sr
#> $sr$wkt
#> [1] NA
#> 
#> $sr$wkid
#> [1] 102100
#> 
#> $sr$latest_wkid
#> [1] 3857
#> 
#> $sr$vcs_wkid
#> [1] NA
#> 
#> $sr$latest_vcs_wkid
#> [1] NA

Post-processing can be applied to the result of process_pbf() using post_process_pbf().

post_process_pbf(res)
#> Simple feature collection with 2 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17260020 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#>          County                       geometry
#> 1 Hawaii County POLYGON ((-17264972 2244291...
#> 2 Hawaii County POLYGON ((-17264972 2244291...

post_process_pbf() can also be applied to a list of processed pbf responses.

multi_res <- list(res, res, res)

post_process_pbf(multi_res)
#> Simple feature collection with 6 features and 1 field
#> Geometry type: POLYGON
#> Dimension:     XY
#> Bounding box:  xmin: -17298700 ymin: 2216212 xmax: -17260020 ymax: 2261306
#> Projected CRS: WGS 84 / Pseudo-Mercator
#>          County                       geometry
#> 1 Hawaii County POLYGON ((-17264972 2244291...
#> 2 Hawaii County POLYGON ((-17264972 2244291...
#> 3 Hawaii County POLYGON ((-17264972 2244291...
#> 4 Hawaii County POLYGON ((-17264972 2244291...
#> 5 Hawaii County POLYGON ((-17264972 2244291...
#> 6 Hawaii County POLYGON ((-17264972 2244291...

Benchmarking

Below is a bench mark that compares processing pbfs to the current approach of processing raw json in arcgislayers and arcgisutils. The below recreates the example from the README of arcgislayers.

jsn <- function() {
  json_reqs <- c(
    "https://services.arcgis.com/P3ePLMYs2RVChkJx/ArcGIS/rest/services/USA_Counties_Generalized_Boundaries/FeatureServer/0/query?outFields=%2A&where=1%3D1&outSR=%7B%22wkid%22%3A4326%7D&returnGeometry=TRUE&token=&f=json&resultOffset=0",
    "https://services.arcgis.com/P3ePLMYs2RVChkJx/ArcGIS/rest/services/USA_Counties_Generalized_Boundaries/FeatureServer/0/query?outFields=%2A&where=1%3D1&outSR=%7B%22wkid%22%3A4326%7D&returnGeometry=TRUE&token=&f=json&resultOffset=2001"
  )
  reqs <- lapply(json_reqs, httr2::request) 
  
  resps <- httr2::req_perform_parallel(reqs) |> 
    lapply(function(x) arcgisutils::parse_esri_json(httr2::resp_body_string(x))) 
  
  do.call(rbind.data.frame, resps) |> 
    sf::st_as_sf()
}

# protobuff processing 
pbf <- function() {
  
  pbf_reqs <- c(
    "https://services.arcgis.com/P3ePLMYs2RVChkJx/ArcGIS/rest/services/USA_Counties_Generalized_Boundaries/FeatureServer/0/query?outFields=%2A&where=1%3D1&outSR=%7B%22wkid%22%3A4326%7D&returnGeometry=TRUE&token=&f=pbf&resultOffset=0",
    "https://services.arcgis.com/P3ePLMYs2RVChkJx/ArcGIS/rest/services/USA_Counties_Generalized_Boundaries/FeatureServer/0/query?outFields=%2A&where=1%3D1&outSR=%7B%22wkid%22%3A4326%7D&returnGeometry=TRUE&token=&f=pbf&resultOffset=2001"
  )
  
  reqs <- lapply(pbf_reqs, httr2::request)
  
  httr2::req_perform_parallel(reqs) |> 
    resps_data_pbf()
}

bench::mark(
  jsn(),
  pbf(),
  check = FALSE,
  relative = TRUE,
  iterations = 5
)
#> Iterating ■■■■■■■■■■■■■■■■                  50% | ETA:  1s                                                           Iterating ■■■■■■■■■■■■■■■■                  50% | ETA:  1s
#> Iterating ■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■  100% | ETA:  0s
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#>   expression   min median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <dbl>  <dbl>     <dbl>     <dbl>    <dbl>
#> 1 jsn()       3.06   3.27      1         3.94     1.39
#> 2 pbf()       1      1         3.94      1        1

Internals

Internally, there is a rust crate esripbf which is a a Rust library built with prost to handle the FeatureCollection Protocol Buffer Specification.

Future Notes

Alternatively, it may make sense to write to a geoarrow array and convert to sfc using {wk}. These are just thoughts.