interest in modernizing?

[jGaboardi]

@carsonfarmer Do you have any interest in seeing fastpairs modernized (e.g., Python 312 compatibility, pyproject.toml, GHA for testing, structured formatting & linting, etc.)? If so (and as time permits), I can start making specific tickets and chipping away at it.

xref: #12

[carsonfarmer]

I mean, yeh awesome. That would be amazing!

[jGaboardi]

Plan of attack

0. ✅ having trouble with local install due non-conforming versioning (see here). seems to need some minor doctoring and potential implementation of (4.v) before continuing. will investigate – remove portion of setup.py to allow for install #23
1. ✅ initialize a single py312_latest bare-bones environment + GHA for testing current package / code base - ensure passing – remove portion of setup.py to allow for install #23
   1. .github/workflows/testing.yml
   2. ci/py312_latest.yml - Ubuntu
2. ✅ remove .travis.yml – remove old .travis.yml -- #20 #25
3. ✅ consolidate old infrastructure into pyproject.toml – consolidate infrastructure into pyproject.tml #30
   1. setup.py
   2. requirements.txt
   3. recommended.txt
   4. .coveragerc
   5. use setuptools-scm for versioning
4. ✅ add codecov.yml and turn on for the repo
   1. add CODECOV_TOKEN secret #26
   2. update testing workflow file for CODECOV secret #27
   3. [INFRA] add a codecov.yml #28
5. ✅ add ruff specifications to pyproject.toml for linting and formatting codebase – utilizing ruff for formatting & linting – #20 #31
   1. One PR but in the following order:
      1. format with ruff specifications
      2. push for prelim review
      3. lint with ruff specifications
      4. push for final review and iterate if needed
6. ✅ remove Python 2 adaptations in the code base – utilizing ruff for formatting & linting – #20 #31
7. ✅ add a .pre-commit-configuration.yaml + turn on pre-commit for repo – add pre-commit-config.yaml - #20 #32
8. ✅ review dependency min versions & fill out CI envs – Ubuntu / Windows/ macOS – fix #22 - update min req & boost CI suite #20 #33
   1. [COMPAT] fastpairs not compatible with scipy>=1.12 #22
   2. update pyproject.toml
   3. py310_oldest (based on SPEC000)
   4. py311_latest
   5. py312_dev
9. ✅ fill out testing with any edge cases that crop up
   1. start work on improving testing #50
   2. flesh out testing - #50 #61
10. ✅ add a notebooks/ or examples/ directory with some demonstrations
    1. adding first notebook - basic usage #40
    2. notebooks: refactor basic - add n-dimensional #46
    3. ideas for more notebooks #45
    4. etc.
11. 👀 add type hints – Add type hints - beef up docstrings - test examples, etc. #72
12. 👀 revisit docstrings for numpydoc conformity – Add type hints - beef up docstrings - test examples, etc. #72
13. ⚠️ ensure tests for all errors & warnings
    1. mixed dimensional errors
    2. min points warning (initialized via build()? #58)
    3. ...
    4. ...
14. ⚠️ test against several distance scipy.spatial.distance functions
15. ⚠️ test against custom distance function
16. ➿ revisit README for updates, badges, etc. :
    1. update README by section #35
    2. add matplotlib to dependencies #36
    3. first update to README - formatting + Overview, Installation, Testing #37
    4. adding first notebook - basic usage #40
    5. boil down basic_usage.ipynb further - hard code points #41
    6. advanced_usage demo w/ high-dimensional points #44
    7. review wording, etc. in README
17. ⚠️ (potential) see about docs/ + a GHPages site built from a GHA
18. ⚠️ (potential) see about release GHA based on pushed v* tags (automated publishing on PyPI)
19. ⚠️ Following release on PyPI, initialize release on conda-forge as a feedstock
    1. add PyPI & conda-forge badge (once available) #43
    2. start with an alpha, etc release sand iterate until all actions, etc are in order
20. ⚠️ Add add Release, PyPI, and conda-forge badges to README
21. ...
22. triage
23. ...
24. Proceed with first stable release of fastpair 🥳 🎉

@carsonfarmer This is a loosely sequential checklist of proposed activities. Some make sense to do in individual PRs and some may be grouped. Let me know if you have any questions or concerns with anything here, otherwise I'll see about getting started with (1.)

[carsonfarmer]

TL;DR: This all looks good to me!

I'm pretty unfamiliar with modern Python development flows at this stage @jGaboardi, so I would defer to your judgement here. Since I have, in fact, met you in person, I have done ahead and invited you to be a collaborator on this repo. You should have direct access to GitHub Actions etc as needed with that invite?

[carsonfarmer]

Totally separate note: Is PySAL using this library?

[jGaboardi]

Totally separate note: Is PySAL using this library?

Not currently, but I became aware of fastpair last summer when @ljwolf, @gegen07, and I were working with a student in a PySAL GSoC project for spopt. So there's potential for inclusion later down the line in our new KNearestPMedian model, but we'll need to do performance testing, benchmarking, etc. before that happens. So, my contributions here are for 3 reasons (in order of general to specific):

1. I love the packaging aspect of the Scientific Python ecosystem.
2. I believe your work on this has a lot value and making it available to a broader audience will hopefully make others' lives easier.
3. The specific pysal/spopt reason mentioned above.

[gegen07]

@jGaboardi I saw a lot of work done by you! Awesome! Any help wanted?
The spopt will be helped with this brilliant implementation, thanks @carsonfarmer!

[jGaboardi]

@carsonfarmer For context here, @gegen07 was an outstanding PySAL GSoC 2021 student several years back that brought the locate module into existence, and he has since joined us on the core team.

[jGaboardi]

Any help wanted?

Of course! Your insight is always valuable, specifically due to your CS background. Anything specific that has caught your eye in the issues?

I can't speak for @carsonfarmer but IMHO we should get fastpair modernized with an actual release before addressing enhancements like:

• info & background on merge_closest() method #34
• fastpair for points in disjoint sets? #39
• potential for more robust random point generation #42

[gegen07]

No problem with current issues. All good!
I'll take a look at the issues for in-depth enhancements.
Agree with you, before enhancing it, let's get it over the release cycle first.