[EHN] creating the metainformation file to allow lazy loading by default (#688)

* [EHN] creating the metainformation file to allow lazy loading by default

* [EHN] whats_new file

* [EHN] moving to the father class

* updating the whats_new.rst file

* fixing the tests

* [EHN] moving to utils and changing to tests

* Apply suggestions from code review

Signed-off-by: Bru <b.aristimunha@gmail.com>

* mistake with whats_new.rst file

---------

Signed-off-by: Bru <b.aristimunha@gmail.com>

Author: bruAristimunha

docs/source/whats_new.rst

@@ -18,6 +18,7 @@ Develop branch  - 1.2.1
 Enhancements
 ~~~~~~~~~~~~
 - Reordering the examples in the documentation (:gh:`807` by `Bruno Aristimunha`_)
+- Creating the meta information for the BIDS converted datasets (:gh:`688` by `Bruno Aristimunha`_)
 
 
 Bugs

moabb/datasets/utils.py

@@ -1,8 +1,10 @@
 """Utils for easy database selection."""
 
 import inspect
+from pathlib import Path
 
 import mne
+import mne_bids
 import numpy as np
 from mne import create_info
 from mne.io import RawArray
@@ -323,3 +325,29 @@ def stim_channels_with_selected_ids(
     raw_with_stim = raw.copy().add_channels([stim_raw], force_update_info=True)
 
     return raw_with_stim
+
+
+def bids_metainfo(bids_path: Path) -> dict:
+    """Create metadata for the BIDS dataset.
+
+    To allow lazy loading of the metadata, we store the metadata in a JSON file
+    in the root of the BIDS dataset.
+
+    Parameters
+    ----------
+    bids_path : Path
+        The path to the BIDS dataset.
+    """
+    json_data = {}
+
+    paths = mne_bids.find_matching_paths(
+        root=bids_path,
+        datatypes="eeg",
+    )
+
+    for path in paths:
+        uid = path.fpath.name
+        json_data[uid] = path.entities
+        json_data[uid]["fpath"] = str(path.fpath)
+
+    return json_data

moabb/tests/test_datasets.py

@@ -6,6 +6,7 @@
 
 import mne
 import numpy as np
+import pandas as pd
 import pytest
 
 import moabb.datasets as db
@@ -20,7 +21,7 @@
 from moabb.datasets.compound_dataset import CompoundDataset
 from moabb.datasets.compound_dataset.utils import compound_dataset_list
 from moabb.datasets.fake import FakeDataset, FakeVirtualRealityDataset
-from moabb.datasets.utils import block_rep, dataset_list
+from moabb.datasets.utils import bids_metainfo, block_rep, dataset_list
 from moabb.paradigms import P300
 from moabb.utils import aliases_list
 
@@ -208,6 +209,14 @@ def test_cache_dataset(self):
             assert len(expected) == len(cm.output)
             for i, regex in enumerate(expected):
                 self.assertRegex(cm.output[i], regex)
+
+            metainfo = bids_metainfo(tempdir)
+            dataframe = pd.DataFrame(metainfo).T
+            subjects = dataframe["subject"].unique()
+            self.assertEqual(len(subjects), 1)
+            self.assertEqual(subjects[0], "1")
+            self.assertEqual(len(dataframe["session"].unique()), 2)
+
         shutil.rmtree(tempdir)
 
     def test_dataset_accept(self):