Let nx-datasets respect XDG cache locations

[tmattio]

Why it matters

Right now Dataset_utils.get_cache_dir always expands to ~/.cache/ocaml-nx/datasets/... (see nx-datasets/lib/dataset_utils.ml:54-62). That ignores both XDG_CACHE_HOME and any custom cache path users set for sandboxed environments. Outreachy applicants working on shared machines or CI systems need to redirect downloads to writable directories without patching the code.

How to reproduce

1. Run XDG_CACHE_HOME=/tmp/outreachy-cache dune utop nx-datasets.
2. Evaluate Nx_datasets.Dataset_utils.get_cache_dir "iris";; — it still prints a path under ~/.cache/ocaml-nx/ instead of /tmp/outreachy-cache/....
3. The behaviour is the same even if you set NX_DATASETS_CACHE, because the code never checks that variable.

Your task

• Update nx-datasets/lib/dataset_utils.ml so cache resolution prefers, in order:
  1. NX_DATASETS_CACHE (if set)
  2. XDG_CACHE_HOME (if set)
  3. Fall back to $HOME/.cache
• Make sure the calculation happens each time we call get_cache_dir (a helper function returning a string is fine) so tests can override the env at runtime.
• Keep the trailing slash in the returned path, and keep creating directories the same way.
• Document the new lookup order in nx-datasets/lib/dataset_utils.mli.
• Add an Alcotest case (e.g. test_dataset_utils.ml) that temporarily sets the env vars with Unix.putenv, calls get_cache_dir, and asserts the path prefix matches the env you set. Remember to restore the old values at the end of the test so other cases are not polluted.

Tips

• Sys.getenv_opt (available since OCaml 4.06) is handy for optional environment variables.
• Prefer Filename.concat when stitching parts so Windows paths keep working.
• In tests, wrap env changes with Fun.protect to restore the original values even if the assertion fails.
• Run dune runtest nx-datasets to make sure the new test and the existing generator suite stay green.

Done when

• Calling get_cache_dir honours NX_DATASETS_CACHE/XDG_CACHE_HOME when they are set, and still falls back to the old behaviour otherwise.
• The updated .mli explains the precedence clearly.
• The new unit test fails on the current main branch and passes after your changes.

[adesinaadeola]

@tmattio Kindly assign to me

[tmattio]

Hi @adesinaadeola, have you already joined the Discord? If not please do join before starting work to say hi and introduce yourself, we're also asking applicants to confirm they can build and run the tests before starting work on an issue.

[adesinaadeola]

Hello tmattio, The discord link has expired when I checked. Kindly send a new link so I can join

…

On Wed, Oct 8, 2025 at 5:43 PM Thibaut Mattio ***@***.***> wrote: *tmattio* left a comment (raven-ml/raven#77) <#77 (comment)> Hi @adesinaadeola <https://github.com/adesinaadeola>, have you already joined the Discord? If not please do join before starting work to say hi and introduce yourself, we're also asking applicants to confirm they can build and run the tests before starting work on an issue. — Reply to this email directly, view it on GitHub <#77 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A2GRWP24UXME7F4XR7ARB6T3WU5KTAVCNFSM6AAAAACITWRJDGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGOBSGM4TKNRXGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[tmattio]

@adesinaadeola this is the link: https://discord.gg/kkb4gwQ3Xm

[adesinaadeola]

Thank you. I have introduced myself.

…

On Thu, Oct 9, 2025 at 3:50 AM Thibaut Mattio ***@***.***> wrote: *tmattio* left a comment (raven-ml/raven#77) <#77 (comment)> @adesinaadeola <https://github.com/adesinaadeola> this is the link: https://discord.gg/kkb4gwQ3Xm — Reply to this email directly, view it on GitHub <#77 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A2GRWP3CI5O4C6ISLVE6AGT3WXENZAVCNFSM6AAAAACITWRJDGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGOBTHA2TONZXGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[SusanGithaigaN]

@tmattio I’d like to work on this. Please assign it to me.

[tmattio]

Sure go ahead @SusanGithaigaN!

@adesinaadeola unless I'm mistaken I was waiting for you to confirm you can build + share your test run, hopefully I didn't get mixed up, but if so please do ping me on Discord.

[adesinaadeola]

Hello tmattio, I was already looking at the issue and waiting for you to assign it to me.

…

On Fri, Oct 10, 2025 at 3:08 PM Thibaut Mattio ***@***.***> wrote: *tmattio* left a comment (raven-ml/raven#77) <#77 (comment)> Sure go ahead @SusanGithaigaN <https://github.com/SusanGithaigaN>! @adesinaadeola <https://github.com/adesinaadeola> unless I'm mistaken I was waiting for you to confirm you can build + share your test run, hopefully I didn't get mixed up, but if so please do ping me on Discord. — Reply to this email directly, view it on GitHub <#77 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A2GRWPZUGRAVM4B3CXNBEYD3W64VRAVCNFSM6AAAAACITWRJDGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGOJQGM2TCNRVG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[tmattio]

@adesinaadeola are you able to build the project and had you shared your running tests on Discord?

[adesinaadeola]

@tmattio I have built the project and shared my running test on Discord

[adesinaadeola]

@tmattio Kindly assign this ticket to me

[Arsalaan-Alam]

hi @tmattio , can I work on this please?

[adesinaadeola]

@tmattio I have the fix before you assigned it to someone else. Kindly assign so I can submit my fix