rmapshaper

An R package providing access to the awesome mapshaper tool by Matthew Bloch, which has both a Node.js command-line tool as well as an interactive web tool.

I started this package so that I could use mapshaper’s Visvalingam simplification method in R. There is, as far as I know, no other R package that performs topologically-aware multi-polygon simplification. (This means that shared boundaries between adjacent polygons are always kept intact, with no gaps or overlaps, even at high levels of simplification).

But mapshaper does much more than simplification, so I am working on wrapping most of the core functionality of mapshaper into R functions.

So far, rmapshaper provides the following functions:

• ms_simplify - simplify polygons or lines
• ms_clip - clip an area out of a layer using a polygon layer or a bounding box. Works on polygons, lines, and points
• ms_erase - erase an area from a layer using a polygon layer or a bounding box. Works on polygons, lines, and points
• ms_dissolve - aggregate polygon features, optionally specifying a field to aggregate on. If no field is specified, will merge all polygons into one.
• ms_explode - convert multipart shapes to single part. Works with polygons, lines, and points in geojson format, but currently only with polygons and lines in the Spatial classes (not SpatialMultiPoints and SpatialMultiPointsDataFrame).
• ms_lines - convert polygons to topological boundaries (lines)
• ms_innerlines - convert polygons to shared inner boundaries (lines)
• ms_points - create points from a polygon layer
• ms_filter_fields - Remove fields from the attributes
• ms_filter_islands - Remove small detached polygons

If you run into any bugs or have any feature requests, please file an issue

Installation

rmapshaper is on CRAN. Install the current version with:

install.packages("rmapshaper")

You can install the development version from github with remotes:

## install.packages("remotes")
library(remotes)
install_github("ateucher/rmapshaper")

Usage

rmapshaper works with geojson strings (character objects of class geo_json) and list geojson objects of class geo_list. These classes are defined in the geojsonio package. It also works with Spatial classes from the sp package, and with sf and scf objects from the sf package.

We will use the states dataset from the geojsonio package and first turn it into a geo_json object:

library(geojsonio)
#> Registered S3 method overwritten by 'geojsonsf':
#>   method        from   
#>   print.geojson geojson
#> 
#> Attaching package: 'geojsonio'
#> The following object is masked from 'package:base':
#> 
#>     pretty
library(rmapshaper)
#> Registered S3 method overwritten by 'geojsonlint':
#>   method         from 
#>   print.location dplyr
library(sp)
library(sf)
#> Linking to GEOS 3.10.2, GDAL 3.4.2, PROJ 8.2.1; sf_use_s2() is TRUE

## First convert to json
states_json <- geojson_json(states, geometry = "polygon", group = "group")
#> Assuming 'long' and 'lat' are longitude and latitude, respectively

## For ease of illustration via plotting, we will convert to a `SpatialPolygonsDataFrame`:
states_sp <- geojson_sp(states_json)

## Plot the original
plot(states_sp)

## Now simplify using default parameters, then plot the simplified states
states_simp <- ms_simplify(states_sp)
#> Warning in sp::proj4string(sp): CRS object has comment, which is lost in output; in tests, see
#> https://cran.r-project.org/web/packages/sp/vignettes/CRS_warnings.html
plot(states_simp)

You can see that even at very high levels of simplification, the mapshaper simplification algorithm preserves the topology, including shared boundaries:

states_very_simp <- ms_simplify(states_sp, keep = 0.001)
#> Warning in sp::proj4string(sp): CRS object has comment, which is lost in output; in tests, see
#> https://cran.r-project.org/web/packages/sp/vignettes/CRS_warnings.html
plot(states_very_simp)

Compare this to the output using rgeos::gSimplify, where overlaps and gaps are evident:

library(rgeos)
#> rgeos version: 0.5-9, (SVN revision 684)
#>  GEOS runtime version: 3.10.2-CAPI-1.16.0 
#>  Please note that rgeos will be retired by the end of 2023,
#> plan transition to sf functions using GEOS at your earliest convenience.
#>  GEOS using OverlayNG
#>  Linking to sp version: 1.4-7 
#>  Polygon checking: TRUE
states_gsimp <- gSimplify(states_sp, tol = 1, topologyPreserve = TRUE)
plot(states_gsimp)

The package also works with sf objects. This time we’ll demonstrate the ms_innerlines function:

library(sf)

states_sf <- st_as_sf(states_sp)
states_sf_innerlines <- ms_innerlines(states_sf)
plot(states_sf_innerlines)

All of the functions are quite fast with geo_json character objects and geo_list list objects. They are slower with the Spatial classes due to internal conversion to/from json. Operating on sf objects is faster than with Spatial objects, but not as fast as with the geo_json or geo_list. If you are going to do multiple operations on large Spatial objects, it’s recommended to first convert to json using geojson_list or geojson_json from the geojsonio package. All of the functions have the input object as the first argument, and return the same class of object as the input. As such, they can be chained together. For a contrived example, using states_sp as created above:

library(geojsonio)
library(rmapshaper)
library(sp)
library(magrittr)

## First convert 'states' dataframe from geojsonio pkg to json
states_json <- geojson_json(states, lat = "lat", lon = "long", group = "group", 
                            geometry = "polygon")

states_json %>% 
  ms_erase(bbox = c(-107, 36, -101, 42)) %>% # Cut a big hole in the middle
  ms_dissolve() %>% # Dissolve state borders
  ms_simplify(keep_shapes = TRUE, explode = TRUE) %>% # Simplify polygon
  geojson_sp() %>% # Convert to SpatialPolygonsDataFrame
  plot(col = "blue") # plot

Using the system mapshaper

Sometimes if you are dealing with a very large spatial object in R, rmapshaper functions will take a very long time or not work at all. As of version 0.4.0, you can make use of the system mapshaper library if you have it installed. This will allow you to work with very large spatial objects.

First make sure you have mapshaper installed:

check_sys_mapshaper()
#> mapshaper version 0.5.88 is installed and on your PATH
#>                  mapshaper-xl 
#> "/usr/local/bin/mapshaper-xl"

If you get an error, you will need to install mapshaper. First install node (https://nodejs.org/en/) and then install mapshaper with:

npm install -g mapshaper

Then you can use the sys argument in any rmapshaper function:

states_simp_internal <- ms_simplify(states_sf)
states_simp_sys <- ms_simplify(states_sf, sys = TRUE, sys_mem=8) #sys_mem specifies the amout of memory to use in Gb.  It defaults to 8 if omitted. 

par(mfrow = c(1,2))
plot(st_geometry(states_simp_internal), main = "internal")
plot(st_geometry(states_simp_sys), main = "system")

Thanks

This package uses the V8 package to provide an environment in which to run mapshaper’s javascript code in R. It relies heavily on all of the great spatial packages that already exist (especially sp and rgdal), the geojsonio package for converting between geo_list, geo_json, and sf and Spatial objects, and the jsonlite package for converting between json strings and R objects.

Thanks to timelyportfolio for helping me wrangle the javascript to the point where it works in V8. He also wrote the mapshaper htmlwidget, which provides access to the mapshaper web interface, right in your R session. We have plans to combine the two in the future.

Code of Conduct

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

LICENSE

MIT