[MAINT] Update pyriemann mean import

[bruAristimunha]

Summary

• Try the legacy pyriemann mean import first for older releases.
• Fall back to pyriemann.geometry.mean.gmean when the legacy module emits the deprecation warning or is unavailable.

Why

Newer pyriemann versions deprecate pyriemann.utils.mean, which can fail when deprecations are treated as errors.

Validation

• python -W error::DeprecationWarning -c "import meegkit.trca"
• python -W error::DeprecationWarning - <<PY ... trca_regul(...) ... PY
• python -m pytest tests/test_trca.py -q
• python -m ruff check meegkit/trca.py

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.12%. Comparing base (1138a60) to head (a99ee0d).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #103      +/-   ##
==========================================
+ Coverage   83.02%   83.12%   +0.10%     
==========================================
  Files          25       25              
  Lines        2833     2827       -6     
==========================================
- Hits         2352     2350       -2     
+ Misses        481      477       -4     

☔ 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.

[nbara]

Hey @bruAristimunha , thanks! Hope you're well.

Don't you think we could just enforce a newer version of pyriemann?

[sappelhoff]

Thanks for tackling the pyriemann import breakage — the mean_covariance → gmean fallback in asr.py / trca.py looks right.

One suggestion for the block_covariance hunk specifically. check_function / cov_est_functions have had an unstable availability history: they are absent in pyriemann 0.5 (which exposes only the private _check_est) and are present again in 0.12. So there is a release range in between where the new try/except still raises ImportError from both branches — pyriemann.geometry.covariance doesn't exist yet, and pyriemann.utils.covariance no longer carries the two names. (I verified 0.5 and 0.12 directly; I didn't pin down exactly which 0.6–0.11 releases are affected.)

The public covariances() function sidesteps this — it is present in both pyriemann.utils.covariance and pyriemann.geometry.covariance on 0.5 and 0.12 alike. Since covariances(X, estimator=...) just applies the named estimator to each matrix along the first axis, block_covariance can stack the windows and call it directly, which also lets the try/except go:

def block_covariance(data, window=128, overlap=0.5, padding=True, estimator="cov"):
    """..."""
    from pyriemann.utils.covariance import covariances

    assert 0 <= overlap < 1, "overlap must be < 1"
    blocks = []
    n_chans, n_samples = data.shape
    if padding:  # pad data with zeros
        pad = np.zeros((n_chans, int(window / 2)))
        data = np.concatenate((pad, data, pad), axis=1)

    jump = int(window * overlap)
    ix = 0
    while (ix + window < n_samples):
        blocks.append(data[:, ix:ix + window])
        ix = ix + jump

    return covariances(np.array(blocks), estimator=estimator)

This is behaviour-equivalent to the old per-window est(...) loop (matches np.cov per window on 0.5), and the estimator strings ('scm', 'lwf', 'oas', 'mcd', …) are unchanged. The mean_covariance parts of this PR are orthogonal and unaffected. Happy to push this to your branch if useful.

[bruAristimunha]

hey @sappelhoff, sure!

I will adjust!

And yes, @nbara, for sure, I will just increase Pyriemann and solve the problem :)

[nbara]

(@bruAristimunha updated minimum python version and rebased)

[bruAristimunha]

all good for me @nbara :) Please let me know if you want me to work on this PR or even something else, as I made the modification of Pyriemann, I decided to create this PR

[sappelhoff]

@nbara can we get the open ASR PRs in? 🙏 🙂 I am planning to send over a few more AI-assisted ones in the next week or so.

[nbara]

Yes feel free to merge those I approved, otherwise I will do it later tonight

[sappelhoff]

Thanks, could you invite me as a collaborator then? I currently do not have the rights for it.

[larsoner]

Was this not in 0.11? Maybe should have been try/except for backward compat

https://github.com/conda-forge/meegkit-feedstock/actions/runs/29774919450/job/88461881957?pr=4