#R in GeoSpaX

GeoSpaX runs R for the analysis and figures R is best at: tmap and ggplot2 maps for publication, terra rasters, rayshader relief, species distribution models with biomod2 and MaxEnt, tidymodels, dbplyr and Google Earth Engine through rgee. The data stays in GeoSpaX: layers and rasters on the map go into R as sf and terra objects, and what R returns comes back to the map and the report.

#The R tab

Every button opens the R console on a template. The template is ordinary R that you can read, change and copy into your own scripts. Pick the inputs, then press Run:

• Layer (sf) → layer: any layer on the map, as an sf object.
• Raster (terra) → raster: any raster on the map, as a terra SpatRaster (all its bands).
• Packages: checked before the code runs. If one is missing, the console offers to install it.

What the code returns comes back. Return a named list, or call the helpers:

Printed output (print, summary, show_query) shows under the results. Figures and tables are also in the run's files and its report. Each console run is kept in the run history (the recipe r-console). It is replaced by the next run, so save code you want to keep as a recipe (below).

Online data (Explorer › Online data: Climate, Species …) feeds straight in. For a species distribution model, get occurrences from GBIF and bioclim layers from WorldClim, then open Explorer › R & Python › R › biomod2 or › MaxEnt with the occurrences as layer and the climate raster as raster.

#R setup

R setup checks whether R is installed and which packages are there, and installs what is missing.

• R itself: if R is missing, Install R and selected packages installs the official CRAN build.
  • Windows: for the current user, without admin rights.
  • macOS: CRAN's installer; macOS asks for your password.
  • Linux: your package manager (apt, dnf, zypper or pacman); the system asks for your password.
  • On a server without a desktop to ask on, R setup gives the command to run instead.
• Package kits: tick the kits you need; Spatial data, Publication maps and Spatial statistics are ticked by default.

Any other CRAN package can be named under Other packages.

How the packages are installed:

• Where: into GeoSpaX's own library (<state>/r-library). Your own R setup and library are not changed, and the R console and recipe R steps look there first.
• Ready-built: the packages are binaries, so nothing is compiled. They come from CRAN on Windows and macOS, and from Posit Public Package Manager for your distribution on Linux (Ubuntu, Debian, RHEL and relatives, openSUSE).
• Progress and failures: installation needs internet and shows progress package by package. A package that fails is named at the end; the others are still installed.

rgee also needs a Google Earth Engine account and Python. Run rgee::ee_install() and rgee::ee_Initialize() once in R; after that, the Earth Engine template works.

#R in the web edition

In the web edition, R runs on the server. People using GeoSpaX in a browser get full R with every package, and nothing is installed on their computers. R setup installs R and the packages on the server, once.

Running R inside the browser itself (webR) is possible for light work, but it is slower, downloads a lot, and rayshader, the heavier biomod2 models and rgee do not run there. It is not part of GeoSpaX yet.

#R in recipes

The console is a one-step recipe; any recipe can have R steps (use: r) between built-in steps. The step's inputs can be layers or rasters, and its outputs go on to the next steps (AUTOMATION §6).

  - id: sdm
    use: r
    with:
      inputs: { layer: "${occurrences.layer}", raster: "${climate.raster}" }
      packages: [maxnet, terra, sf]
      code: |
        library(maxnet)
        ...
        list(suitability = terra::predict(raster, fit, type = "cloglog", na.rm = TRUE))

Recipes with R or Python code run code on your computer, so GeoSpaX asks before saving one that came from elsewhere (AUTOMATION §8). Code typed in the R console is yours, so it runs without that question.

#Over HTTP

curl localhost:8080/api/r                                  # R, its version, the kits, installed packages
curl -X POST localhost:8080/api/r/install -d '{"kits": ["maps", "sdm"], "packages": ["lidR"]}'
curl localhost:8080/api/jobs/1                             # progress of that installation

#Charts and code style

Charts follow ggplot2. GeoSpaX sets ggplot2::theme_set() and the discrete colour palette before your code runs, from Settings › Charts (theme_grey by default; theme_bw, minimal, classic, light or dark; the ggplot2 hue palette, viridis, Okabe–Ito, Brewer Set2 or Dark2; the text size). The charts GeoSpaX draws itself (report charts, the Attributes chart, raster histograms) use the same look, so a report reads as one piece. Add your own + theme(...) to change one chart.

Code follows the tidyverse style guide. The templates (R tab) use the native pipe |>, dplyr and purrr verbs, snake_case names, .data[[field]] for a field chosen at run time and one step per line; write your own steps the same way (styler::style_text() tidies a script).