Fix hyppo to 0.4.0

[AlonsoGuevara]

• Does this PR have a descriptive title that could go in our release notes?
• Does this PR add any new dependencies?
• Does this PR modify any existing APIs?
  • Is the change to the API backwards compatible?
• Have you built the documentation (reference and/or tutorial) and verified the generated documentation is appropriate?

Reference Issues/PRs

What does this implement/fix? Briefly explain your changes.

After 0.4.0 Hyppo changed its license from MIT to PolyForm Noncommercial License 1.0.0
This PR fixes the version to the last MIT licensed release

Any other comments?

[daxpryce]

With hyppo changing to non commercial, pulling in any version newer than 0.4.0 would be out of the question for anyone operating in a commercial space. I would assert this does not help graspologic's aims, but I do think we need to wait for @bdpedigo 's concurrence before I approve.

[bdpedigo]

I really do not want to pin to 0.4.0 in perpetuity - in fact I was about to update to the most recent version; i think we have a test failing because of an old version of hyppo.

i will check with hyppo to see why they made this change...

[bdpedigo]

@AlonsoGuevara @daxpryce I checked with JHU and it seems like this license change was intentional but perhaps they didn't fully understand what it meant for downstream stuff... so there's a chance it will be reversed. But regardless, I think we should assume we're stuck with this for the time being.

my only request on this PR is could you add a comment to the pyproject.toml (if that's possible) describing why we pinned hyppo?

looking forward, I expect someday pinning hyppo will get annoying as python versions/dependencies evolve. i think the future solution should be either dropping hyppo entirely or factoring the code that uses it out into a sub-project (e.g. graspologic-inference) which maybe has a more restrictive license. but i understand that that's currently a low priority item for all of us

[AlonsoGuevara]

@bdpedigo Thanks for getting back on this. I'll add the note and resubmit for approval

[daxpryce]

Another possible option is to add another set of dependencies under optional dependencies (so you would do something like `pip install graspologic[non-commercial]` or something to get hyppo support as is. it would mean any graspologic module that uses hyppo would have to be lazy loaded with a try/catch around failing to pull in hyppo with a warning message about what other deps you would need. But that's something you *can* do for the future, if the non commercial license is going to stick around for longer than we would like.

…

On Wed, Feb 12, 2025, at 10:15 AM, Ben Pedigo wrote: @AlonsoGuevara <https://github.com/AlonsoGuevara> @daxpryce <https://github.com/daxpryce> I checked with JHU and it seems like this license change was intentional but perhaps they didn't fully understand what it meant for downstream stuff... so there's a chance it will be reversed. But regardless, I think we should assume we're stuck with this for the time being. my only request on this PR is could you add a comment to the pyproject.toml (if that's possible) describing why we pinned hyppo? looking forward, I expect someday pinning hyppo will get annoying as python versions/dependencies evolve. i think the future solution should be either dropping hyppo entirely or factoring the code that uses it out into a sub-project (e.g. `graspologic-inference`) which maybe has a more restrictive license. but i understand that that's currently a low priority item for all of us — Reply to this email directly, view it on GitHub <#1095 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG3SU2MICC572QTXOYPHED2POFUXAVCNFSM6AAAAABWUIMMASVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNJUGUYDEOJQGM>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[AlonsoGuevara]

@bdpedigo @daxpryce Updated as requested

[daxpryce]

lgtm

[jovo]

Hi all - I'm sorry, this was my mistake. I misunderstood something communicated to me, we will update hyppo's license to be MIT, and then the plan is for it to stay there in perpetuity.

[gnowland]

fixing hyppo to 0.4.0 results in this bug neurodata/hyppo#435 (patched in hyppo v0.5.2) affecting graspologic installs:

Traceback (most recent call last):
...
    from graspologic.partition import hierarchical_leiden
  File "graspologic/__init__.py", line 8, in <module>
    import graspologic.inference
  File "graspologic/inference/__init__.py", line 6, in <module>
    from .latent_distribution_test import latent_distribution_test
  File "graspologic/inference/latent_distribution_test.py", line 7, in <module>
    from hyppo.ksample import KSample
  File "hyppo/__init__.py", line 5, in <module>
    import hyppo.kgof
  File "hyppo/kgof/__init__.py", line 2, in <module>
    from .fssd import FSSD, FSSDH0SimCovObs
  File "hyppo/kgof/fssd.py", line 4, in <module>
    from past.utils import old_div
ModuleNotFoundError: No module named 'past'

Merging #1092 to update hyppo to v0.5.2 would resolve, but if the plan is to pin hyppo to v0.4.0 please at least add future>=1.0.0 to the pyproject.toml to resolve this?

But, it sounds like the hyppo license was reverted to MIT by @jovo , so really this PR should be closed and #1092 merged & a version update published.