Improve SNIRF Sampling Rate Estimation by Replacing Rounding with Pre…

[hiijoshi]

Overview
This PR improves the robustness of sampling rate extraction from SNIRF files by replacing a potentially unsafe .round() operation with a more reliable np.allclose() check. This change ensures compatibility with high-precision data and avoids unintended misclassification of uniformly-sampled data as jittered.

Problem
In _extract_sampling_rate(), the existing implementation uses:

python
Copy
Edit
uniq_periods = np.unique(periods.round(decimals=4))
This approach may incorrectly group near-equal values due to rounding, leading to false positives when detecting jitter.

Solution
This PR:

Replaces the rounding + uniqueness check with np.allclose(periods, mean_period, rtol=1e-6)

Retains jitter detection logic with improved precision

Keeps the warning message for users when jitter exceeds 1% threshold

Benefits
Better support for high-resolution SNIRF datasets

More accurate sampling rate detection

Fewer false warnings about jitter

Cleaner code and better alignment with scientific precision standards

🔬 Test Strategy
Although the function is part of a data I/O pipeline, the following unit tests are proposed:

✅ New Tests:
Test Perfectly Uniform Sampling

Input: Linearly spaced time_data

Expected: No warning, correct sampling rate

Test Sampling With Very Small Precision Jitter

Input: time_data with tiny noise added (below 1% jitter)

Expected: Sampling rate computed, info logged, no warning raised

Test Clearly Jittered Data

Input: time_data with jitter > 1%

Expected: Sampling rate computed, warning raised

✅ Example Test Snippet (using pytest + mne.utils._logging.catch_logging):
python
Copy
Edit
import numpy as np
from mne.io import read_raw_snirf
from mne.utils._logging import catch_logging

def test_uniform_sampling_rate_extraction():
class DummyDat:
def get(self, path):
if path == "nirs/data1/time":
return np.linspace(0, 10, 1001)

sr = _extract_sampling_rate(DummyDat())
assert np.isclose(sr, 100.0), "Expected sampling rate 100 Hz"

def test_jitter_warning(caplog):
class DummyDat:
def get(self, path):
if path == "nirs/data1/time":
t = np.linspace(0, 10, 1001)
t[500] += 0.02 # Introduce artificial jitter
return t

with caplog.at_level("WARNING"):
    _extract_sampling_rate(DummyDat())
    assert "Found jitter" in caplog.text

[welcome]

Hello! 👋 Thanks for opening your first pull request here! ❤️ We will try to get back to you soon. 🚴

[drammock]

duplicate of #13184