Packaging R for WebAssembly

[IsabelParedes]

r-base

• Figuring out configuration and cross-compilation
  • https://github.com/IsabelParedes/emscripten-forge-recipes/tree/r-base-flang
• lang: R, C, Fortran
• Package r-base for emscripten-forge
  • Add r-base #1287
  • Requires:
    • Patched flang-new available on conda-forge/label/llvm_dev (for linux)
      • [dev] Add patches for WASM support conda-forge/flang-feedstock#69
    • A recipe to build the Fortran Runtime on emscripten-forge (for wasm)
      • Add libflang #1285
    • libblas
      • lang: Fortran
      • Builds with patched flang-new
      • Add libblas and liblapack #1284
    • liblapack
      • lang: Fortran
      • Builds with patched flang-new
      • Add libblas and liblapack #1284

Other dependencies

• xz add xz recipe #1055
• libcurl (NOT NEEDED / wouldn't work!)
• cairo Add cairo #1109
  • deps: glib, pixman
• pixman add pixman #1058
• glib add glib #1051
• pango fancy text rendering (optional) Add pango #1398
  • deps: cairo, harfbuzz
• harfbuzz Add harfbuzz #1067
  • deps: graphite2 Add graphite2 #1057
• icu Add icu #1043
• libiconv
• zlib
• libtiff
• libpng
• bzip2
• pcre2 Add pcre2 #1041
• libjpeg-turbo jpeg Add jpeg (alternative to libjpeg-turbo) #1052
• tk
• bwidget
  • deps: tk
• tktable
  • deps: tk
• readline
  • dep: ncurses Add ncurses #882

R packages

All of the following packages depend on r-base and are dependencies for xeus-r.

• r-rlang Add r-rlang #1303
• r-evaluate noarch
• r-jsonlite Add r-jsonlite #1307
• r-glue Add r-glue #1306
• r-cli Add r-cli #1364 Only <v3.0.1 noarch
• r-repr noarch
  • dependencies:
    • r-base
    • r-base64enc Add r-base64enc #1314
    • r-htmltools Add r-htmltools #1345
      • lang: R, C
      • dependencies:
        • r-base
        • r-base64enc
        • r-rlang
        • r-digest Add r-digest #1308
        • r-ellipsis Add r-ellipsis #1343
          • lang: R, C
          • dep: r-rlang
        • r-fastmap Add r-fastmap #1309
    • r-pillar noarch
      • dependencies:
        • r-base
        • r-cli
        • r-ellipsis
        • r-rlang
        • (optional) r-crayon noarch
          • Superseded by r-cli
          • lang: R
          • deps: r-base
        • r-fansi Add r-fansi #1311
        • r-lifecycle noarch
        • r-vctrs Add r-vctrs #1342
          • lang: R, C
          • dependencies:
            • r-base
            • r-rlang
            • r-cli
            • r-glue
            • r-lifecycle
        • r-utf8 Add r-utf8 #1313
• r-IRdisplay noarch
  • dependencies:
    • r-base
    • r-repr
• r-ggplot2 noarch
  • dependencies:
    • r-base
    • r-rlang
    • r-cli
    • r-glue
    • r-lifecycle
    • r-vctrs
    • r-mass Add r-mass #1315
    • r-gtable noarch
    • r-isoband Add r-isoband #1316
    • r-mgcv Add r-mgcv #1340
      • nlme Add r-nlme #1338
      • matrix Add r-matrix #1339
      • lattice Add r-lattice #1317
    • r-scales noarch
      • r-farver Add r-farver #1323
    • r-tibble Add r-tibble #1347
    • r-magrittr Add r-magrittr #1324
    • r-withr noarch

r-coursekata

• r-coursekata (noarch)
  • r-purrr >=0.3.4 Add r-purrr #1348
  • r-metrics (noarch)
  • r-dslabs (noarch)
  • r-ggformula (noarch)
  • r-lsr (noarch)
  • r-mosaic (noarch)
  • r-pak (noarch)
  • r-palmerpenguins (noarch)
  • r-supernova (noarch)
  • r-viridislite (noarch)

[DerThorsten]

maybe we can use clapack instead of liblapack https://github.com/emscripten-forge/recipes/tree/main/recipes/recipes_emscripten/clapack

[IsabelParedes]

maybe we can use clapack instead of liblapack https://github.com/emscripten-forge/recipes/tree/main/recipes/recipes_emscripten/clapack

I like this idea. This could simplify a lot of things. Worth trying it out!

[SylvainCorlay]

https://github.com/conda-forge/r-coursekata-feedstock

[IsabelParedes]

As a side note, I tried building r-base with clapack and openblas which are already available on emscripten-forge. However, r-base still has an internal appl lib which requires having a Fortran to WebAssembly toolchain.

[adamblake]

@IsabelParedes @SylvainCorlay I got a new requirement for this project which is the neuralnet package. It looks like it is noarch, but has a dependency on MASS (already packaged on this issue) and Deriv (not yet packaged). I checked and Deriv only depends on r-base, so I think we are good, right?

• r-neuralnet meta: https://github.com/conda-forge/r-neuralnet-feedstock/blob/main/recipe/meta.yaml
• r-deriv meta: https://github.com/conda-forge/r-deriv-feedstock/blob/main/recipe/meta.yaml

[IsabelParedes]

@IsabelParedes @SylvainCorlay I got a new requirement for this project which is the neuralnet package. It looks like it is noarch, but has a dependency on MASS (already packaged on this issue) and Deriv (not yet packaged). I checked and Deriv only depends on r-base, so I think we are good, right?

• r-neuralnet meta: https://github.com/conda-forge/r-neuralnet-feedstock/blob/main/recipe/meta.yaml

• r-deriv meta: https://github.com/conda-forge/r-deriv-feedstock/blob/main/recipe/meta.yaml

Hi @adamblake! Correct, these two packages do not need compilation so they're good to go.