BUG: Fix bug with split epochs and ICA

docs/source/v1.6.md.inc

@@ -20,15 +20,16 @@
 
 ### :bug: Bug fixes
 
-- Fix minor issues with path handling for cross-talk and calibration files (#834 by @larsoner)
-- Fix bug where EEG `reject` params were not used for `ch_types = ["meg", "eeg"]` (#839 by @larsoner)
-- Fix bug where implicit `mf_reference_run` could change across invocations of `mne_bids_pipeline`, breaking caching (#839 by @larsoner)
-- Fix bug where `--no-cache` had no effect (#839 by @larsoner)
-- Fix bug where the Maxwell filtering step would fail if [`find_noisy_channels_meg = False`][mne_bids_pipeline._config.find_noisy_channels_meg]` was used (#847 by @larsoner)
-- Fix bug where raw, empty-room, and custom noise covariances were errantly calculated on data without ICA or SSP applied (#840 by @larsoner)
-- Fix bug where multiple channel types (e.g., MEG and EEG) were not handled correctly in decoding (#853 by @larsoner)
-- Fix bug where the previous default for [`ica_n_components`][mne_bids_pipeline._config.ica_n_components] of `0.8` was too conservative, changed the default to `None` to match MNE-Python (#853 by @larsoner)
-- Fix bug where the events table for the average subject could overflow in reports (#854 by @larsoner)
+- Fixed minor issues with path handling for cross-talk and calibration files (#834 by @larsoner)
+- Fixed EEG `reject` use for `ch_types = ["meg", "eeg"]` in epoch cleaning (#839 by @larsoner)
+- Fixed bug where implicit `mf_reference_run` could change across invocations of `mne_bids_pipeline`, breaking caching (#839 by @larsoner)
+- Fixed `--no-cache` behavior having no effect (#839 by @larsoner)
+- Fixed Maxwell filtering failures when [`find_noisy_channels_meg = False`][mne_bids_pipeline._config.find_noisy_channels_meg]` is used (#847 by @larsoner)
+- Fixed raw, empty-room, and custom noise covariances calculation, previously they could errantly be calculated on data without ICA or SSP applied (#840 by @larsoner)
+- Fixed multiple channel type handling (e.g., MEG and EEG) in decoding (#853 by @larsoner)
+- Changed the default for [`ica_n_components`][mne_bids_pipeline._config.ica_n_components] from `0.8` (too conservative) to `None` to match MNE-Python's default (#853 by @larsoner)
+- Prevent events table for the average subject overflowing in reports (#854 by @larsoner)
+- Fixed split file behavior for Epochs when using ICA (#855 by @larsoner)
 
 ### :medical_symbol: Code health
 

mne_bids_pipeline/steps/preprocessing/_08a_apply_ica.py

@@ -113,7 +113,7 @@ def apply_ica_epochs(
 ) -> dict:
     bids_basename = in_files["ica"].copy().update(processing=None)
     out_files = dict()
-    out_files["epochs"] = in_files["epochs"].copy().update(processing="ica")
+    out_files["epochs"] = in_files["epochs"].copy().update(processing="ica", split=None)
     out_files["report"] = bids_basename.copy().update(
         processing="ica", suffix="report", extension=".html"
     )

mne_bids_pipeline/tests/conftest.py

@@ -58,6 +58,8 @@ def pytest_configure(config):
     always:constrained_layout not applied.*:UserWarning
     ignore:datetime\.datetime\.utcfromtimestamp.*:DeprecationWarning
     ignore:datetime\.datetime\.utcnow.*:DeprecationWarning
+    # pandas with no good workaround
+    ignore:The behavior of DataFrame concatenation with empty.*:FutureWarning
     """
     for warning_line in warning_lines.split("\n"):
         warning_line = warning_line.strip()

mne_bids_pipeline/tests/test_run.py

@@ -60,7 +60,13 @@ class _TestOptionsT(TypedDict, total=False):
 _n_jobs = {"preprocessing/_05_make_epochs": 1}
 """,
     },
-    "ds000248_ica": {},
+    "ds000248_ica": {
+        "extra_config": """
+_raw_split_size = "60MB"
+_epochs_split_size = "30MB"
+_n_jobs = {}
+"""
+    },
     "ds000248_T1_BEM": {
         "steps": ("source/make_bem_surfaces",),
         "requires": ("freesurfer",),