Vectorize bandpass filtering in NoisyChannels._get_filtered_data

[sappelhoff]

Summary

Vectorize the bandpass filter in NoisyChannels._get_filtered_data by filtering all channels at once with scipy.signal.filtfilt(..., axis=1) instead of looping filtfilt over each channel into a pre-allocated np.zeros_like buffer.

The per-channel loop recomputed the FIR initial conditions (lfilter_zi → linalg.solve) once per channel; axis=1 computes them once and runs the filter in C across all rows.

Why

_get_filtered_data dominates the runtime of find_bad_by_correlation (and any path that needs the filtered signal). On an 18.5-min, 99-channel @ 500 Hz recording:

• filtfilt step: 9.98 s → 2.40 s (4.16×) (clean wall-clock, median of 5)
• find_bad_by_correlation: 12.14 s → 5.06 s (2.40×)
• It also drops the explicit full-size np.zeros_like output buffer.

Correctness

Bit-for-bit identical output — filtfilt along axis=1 applies the same per-row algorithm, and lfilter_zi depends only on the filter coefficients, not the data:

• np.array_equal with the old per-channel loop, maxdiff 0.0, on the real recording and random arrays of shapes (64, 9601), (10, 1000), (1, 5000), (99, 553759).
• Full test suite passes (47 passed), including tests/test_matprep_compare.py (numeric MATLAB-PREP equivalence).
• get_bads(as_dict=True) and all _extra_info arrays unchanged on the real recording (reject "omit" and None) and the eegbci fixture (matlab_strict False and True).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.85%. Comparing base (54e2554) to head (cf6173f).
⚠️ Report is 21 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #195      +/-   ##
==========================================
- Coverage   97.86%   97.85%   -0.01%     
==========================================
  Files           7        7              
  Lines         842      839       -3     
==========================================
- Hits          824      821       -3     
  Misses         18       18              

☔ 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.