MNT: declare dependency floors the package can actually run on

[thc1006]

Addresses #1107. Not Closes, since the keyword only fires when a pull request targets the default branch and this targets develop.

Pull request type

• Code maintenance (refactoring, formatting, tests)

Current behavior

numpy>=1.13
scipy>=1.0

Neither is true, and pyproject.toml reads this file through dynamic = ["dependencies"], so it is what an installer is told.

SciPy. rocketpy/simulation/monte_carlo.py imports scipy.stats.bootstrap at module level, and the package __init__ reaches that module. The SciPy 1.7.0 release notes are where the function arrives:

scipy.stats.bootstrap has been added to allow estimation of the confidence interval and standard error of a statistic.

So 1.0 through 1.6 satisfy the floor and then fail on import rocketpy, before anything runs.

NumPy. The binding constraint here is a sibling rather than NumPy itself. matplotlib>=3.9.0 is in this same file and requires numpy>=1.23, which I found the direct way:

ImportError: Matplotlib requires numpy>=1.23; you have 1.21.3

That is at 1.21.3, which is where NumPy's own Python 3.10 wheels begin, so the wheel-availability answer is not far enough either.

New behavior

numpy>=1.23  # matplotlib 3.9 requires it
scipy>=1.8   # first to allow numpy 1.23; bootstrap needs 1.7

SciPy is 1.8 rather than 1.7 because the two have to compose: scipy==1.7.2 declares numpy<1.23.0, so pinning 1.7 and asking for numpy 1.23 is unsatisfiable. 1.8.0 raised its cap to <1.25.0.

Both are old. SciPy 1.8.0 is February 2022 and NumPy 1.23.0 June 2022, against a package that already requires Python 3.10.

Breaking change

• No

It refuses installs that would have failed anyway, earlier and with a message about versions rather than an ImportError.

How this was checked

Installed rather than reasoned about, since the constraint that decided the NumPy floor was not the one I expected:

python 3.10, numpy 1.23.0, scipy 1.8.0 pinned, everything else current

import rocketpy                                   works
tests/unit/simulation + tests/unit/stochastic     162 passed, 5 skipped

And resolved from the file itself, taking the lowest each declaration allows:

uv pip install --resolution=lowest-direct -r requirements.txt
-> numpy 1.23.0, scipy 1.8.0, matplotlib 3.9.0

One thing I found and deliberately left alone

requests, pytz, simplekml and dill carry no floor at all, and under a lowest-version resolution they pick releases that cannot run on Python 3.10:

pytz 2018.3       from collections import Mapping
requests (old)    from collections import MutableMapping

pytz>=2018.4 is the first that imports, if anyone wants the number.

I have not touched them, because that is a different defect. These two lines make a claim that is false; those four make no claim at all, and deciding what they should claim is a policy call rather than a correction. It is also not reachable the way the SciPy one is: an ordinary install takes the newest of an unfloored dependency, whereas a deliberate scipy==1.6 is a thing someone might really have.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.55%. Comparing base (e0ff281) to head (5b91612).
⚠️ Report is 24 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1108      +/-   ##
===========================================
+ Coverage    82.18%   82.55%   +0.37%     
===========================================
  Files          122      128       +6     
  Lines        16355    16555     +200     
===========================================
+ Hits         13441    13667     +226     
+ Misses        2914     2888      -26     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[thc1006]

Heads up on a merge order point, found by merging my open PRs together onto develop rather than testing each on its own base.

This PR and #1054 both raise the numpy floor off the same numpy>=1.13, so they conflict textually in requirements.txt:

develop   numpy>=1.13
#1054     numpy>=1.17   SeedSequence, added in 1.17.0
#1108     numpy>=1.23   matplotlib 3.9 requires it

Whichever lands second hits the conflict. The resolution is to take 1.23, which already covers what #1054 needs, so nothing is lost either way and no code has to change.

I have not rebased either branch, since resolving it now would only move the conflict to whichever of us merges first. Happy to rebase this one onto #1054 if you would rather have it ready to go.

On the combined tree, develop + #1054 + #1102 + #1103 + this, with the floor at 1.23: ruff clean, 2198 passed and 45 skipped.