How do we depend on `{XML}`?!

[IndrajeetPatil]

Currently, GHA are failing because of XML installation failures (https://github.com/r-lib/actions/issues/559). But I don't understand where this package is used.

pak::pkg_deps_explain("easystats", "XML")
#> ✔ Updated metadata database: 3.98 MB in 9 files.                          
#> ℹ R 4.2 aarch64-apple-darwin20 packages are missing from Bioconductor      
#> ✔ Updating metadata database ... done                                     
#> x XML  

Is there any soft or hard dependency that (recursively) depends on {XML}?

[bwiernik]

I think that's the new HTML 5 support in documentation

[IndrajeetPatil]

But all GHA fail due to this issue; i.e., all of them seems to need and install {XML}, not just the new HTML-5 GHA.

So I am wondering if there is any soft dependency we rely on that requires {XML}.

[etiennebacher]

Is it because of this line? Or did you add it because there were already some failures?

https://github.com/easystats/workflows/blob/86e36484d77c1406a24155aab53e4d8e09682abc/.github/workflows/R-CMD-check.yaml#L63-L66

(It was added in 65cfaf3)

[IndrajeetPatil]

Yeah, I added it when failures started to occur. That's the recommended solution in the original actions repo issue.

Lemme remove it and see if it makes any difference.

[etiennebacher]

Also, I think pak::pkg_deps_explain() does not take into account the "Suggests", but GHA installs them to run the tests. For example, datawizard has dplyr in "Suggests" but it doesn't show in the output:

> pak::pkg_deps_explain("datawizard", "dplyr")
x dplyr   

[IndrajeetPatil]

You can set dependencies = "all":

> pak::pkg_deps_explain("datawizard", "dplyr", dependencies = "all")
✔ Loading metadata database ... done
datawizard -> brms -> bayesplot -> dplyr                                   
datawizard -> brms -> shinystan -> bayesplot -> dplyr
datawizard -> dplyr
datawizard -> rstanarm -> bayesplot -> dplyr
datawizard -> rstanarm -> shinystan -> bayesplot -> dplyr
datawizard -> tidyr -> dplyr

But, even then, {XML} doesn't show up anywhere.

> pak::pkg_deps_explain("easystats", "XML", dependencies = "all")
x XML  

[IndrajeetPatil]

LMAO. We didn't need it after all. Things seem to be working now.

Closed in https://github.com/easystats/workflows/commit/c116bd1ce43e904edae71b23e32932a0be01a5cc

[etiennebacher]

Where do you see that it worked?

[IndrajeetPatil]

https://github.com/easystats/easystats/pull/303

[IndrajeetPatil]

I think only {parameters} needs it:

> pak::pkg_deps_explain("parameters", "XML", dependencies = "all")
✔ Updated metadata database: 4.82 MB in 12 files.                         
ℹ R 4.2 aarch64-apple-darwin20 packages are missing from Bioconductor      
✔ Updating metadata database ... done                                     
parameters -> EGAnet -> semPlot -> XML 

But including it in the shared workflow creates issues for other packages. Quite the conundrum.

[strengejacke]

Any ideas how to resolve this? Can we make an individual GHA workflow for parameters?

[IndrajeetPatil]

Replace this with the following:

# Workflow derived from https://github.com/r-lib/actions/tree/v2/examples
# Need help debugging build failures? Start at https://github.com/r-lib/actions#where-to-find-help

on:
  push:
    branches: [main, master]
  pull_request:
    branches: [main, master]

name: R-CMD-check

jobs:
  R-CMD-check:
    runs-on: ${{ matrix.config.os }}

    name: ${{ matrix.config.os }} (${{ matrix.config.r }})

    strategy:
      fail-fast: false
      matrix:
        config:
          #- {os: macOS-latest,   r: 'devel'}
          - {os: macOS-latest,   r: 'release'}
          - {os: macOS-latest,   r: 'oldrel'}
          #- {os: macOS-latest,   r: 'oldrel-2'}
          #- {os: macOS-latest,   r: 'oldrel-3'}

          - {os: windows-latest, r: 'devel'}
          - {os: windows-latest, r: 'release'}
          - {os: windows-latest, r: 'oldrel'}
          #- {os: windows-latest, r: 'oldrel-2'}
          #- {os: windows-latest, r: 'oldrel-3'}

          - {os: ubuntu-latest,  r: 'devel', http-user-agent: 'release'}
          - {os: ubuntu-latest,  r: 'release'}
          - {os: ubuntu-latest,  r: 'oldrel'}
          - {os: ubuntu-latest,  r: 'oldrel-2'}
          - {os: ubuntu-latest,  r: 'oldrel-3'}

    env:
      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
      R_KEEP_PKG_SOURCE: yes
      _R_CHECK_CRAN_INCOMING_: false
      _R_CHECK_FORCE_SUGGESTS_: false

    steps:
      - uses: actions/checkout@v2

      # Always try to use the latest pandoc version
      # https://github.com/jgm/pandoc/releases
      - uses: r-lib/actions/setup-pandoc@v2
        with:
          pandoc-version: '2.19.2'

      - uses: r-lib/actions/setup-r@v2
        with:
          r-version: ${{ matrix.config.r }}
          http-user-agent: ${{ matrix.config.http-user-agent }}
          use-public-rspm: true

      # TODO: Check which of the ignore conditions are still relevant given the
      # current suggested dependencies and the minimum supported R version.
      # Update if anything out of date or not needed anymore.
      - uses: r-lib/actions/setup-r-dependencies@v2
        with:
          extra-packages: |
            any::rcmdcheck
            any::XML
            metafor=?ignore-before-r=4.0.0
            emmeans=?ignore-before-r=4.1.0
            rmcorr=?ignore-before-r=4.1.0
            randomForest=?ignore-before-r=4.1.0
            MuMIn=?ignore-before-r=4.2.0
          needs: check

      # Don't error on "note" because if any of the suggested packages are not available
      # for a given R version, this generates a NOTE causing unnecessary build failure
      - uses: r-lib/actions/check-r-package@v2
        with:
          error-on: '"warning"'
          upload-snapshots: true

You will need to manually keep this file in sync with the version in workflows repo.

[etiennebacher]

Other possibility: pass the name of the package that is being tested as an input to the call of the common workflow. In the common workflow, make an if condition that installs XML only if the package tested is parameters. I think it should be feasible but I don't know how exactly.

Cf https://github.com/easystats/easystats/issues/234#issuecomment-1234230324

[IndrajeetPatil]

@etiennebacher Do you want to give it a try?

[etiennebacher]

Yes, maybe this weekend

[strengejacke]

I think only {parameters} needs it:

> pak::pkg_deps_explain("parameters", "XML", dependencies = "all")
✔ Updated metadata database: 4.82 MB in 12 files.                         
ℹ R 4.2 aarch64-apple-darwin20 packages are missing from Bioconductor      
✔ Updating metadata database ... done                                     
parameters -> EGAnet -> semPlot -> XML 

But including it in the shared workflow creates issues for other packages. Quite the conundrum.

Can we prevent XML from being installed?