Agenda 9/3/2021

[F-said]

Updates

• Created constant exception messages for reuse Dev fix test feat #305
• Merged in dev-feature-io-dan to dev-feature-io Update to import .mat, rather than .mff; loop over participants; log missing files.  #279 ty @DMRoberts
• Expanded code to exit on bad data path Dev fix test feat #305
• Allow users to exclude all subjects, tasks, or runs Dev fix test feat #305
• Docker container issue resolved and documented (user error) No write permission in docker  #306
• Prerelease edit pre-release notes & description  #283 ty @jessb0t
• ICA algo Dev feature ica yanbin #280 ty @yanbin-niu
• 2/9 releases back to tar format
• Expanded caught exceptions in:
  • filter_data() to include ValueError & AttributeError Fix FR3 Exception #293
  • ica_raw() to include ValueError & AttributeError Fix IR3 Code #294
• fixed test cases for:
  • filter_data() Fix FR3 Exception #293
  • ica_raw() Fix IR3 Code #294
  • reref_data() Fix RE3 Exception #295
  • segment_data() Fix SE3 Exception  #296
  • load_data() Fix Select Data "Select All" Bugs #299 & Fix Select Data Exception Bugs #297

Blocks/Challenges

• Discussion: when should we allow the program to exit and when should we allow the pipeline to progress? Maybe if bad path is given we want the pipeline to exit and not run-along w/o user knowing that something is wrong.
  • in the context of pipeline features:
    • Parameter Error: Skip to next subj
    • Bad data (somehow bad data got past mne_bids): hard stop
  • in the context of load functions
    • Data is missing in path: hard stop
    • Path is missing: hard stop
• how light-weight should user_params be? Should we cut down on any part?

Next Actions

• PR Dev fix test feat #305
• PR Added discrete sections for how to contrib to issues, docs, and code #286
• PR Dev feature ica yanbin #280
• User_param ignore preprocessed data feature Create user_param option to not re-run already processed data  #309
• User_param exit on error feature Create user_param option to exit on any exceptions  #308

[F-said]

Longest agenda to date probably

[yanbin-niu]

Discussion: when should we allow the program to exit and when should we allow the pipeline to progress? Maybe if bad path is given we want the pipeline to exit and not run-along w/o user knowing that something is wrong.
in the context of pipeline features:
Parameter Error: Continue pipeline
Bad data (somehow bad data got past mne_bids): hard stop?
in the context of load functions
Data is missing in path: continue w/no preprocessing since no data
Path is missing: hard stop?

Beyond the path checking(should be first prioritized), something off the top of my head, but not sure whether it is possible/feasible:

Can we do the checking of user_param and data at the very beginning of each loop (before calling features from preprocessing)? if any is invalid, we jump to (continue) the next loop and write error (which participant, which task, and which run) into output. (Each feature still needs its own input checking).

• Raw data - an instance of BaseRaw

This class is public to allow for stable type-checking in user code (i.e., isinstance(my_raw_object, BaseRaw)) but should not be used as a constructor for Raw objects

https://mne.tools/stable/generated/mne.io.BaseRaw.html?highlight=baseraw#mne.io.BaseRaw

Or we can use _validate_type (this can be used to check many other MNE types, such as "path-like")
https://github.com/mne-tools/mne-python/blob/maint/0.23/mne/utils/check.py#L383

• For user_param, should decide on items that must be provided. If any is missing, then we think the user_param is invalid.

[yanbin-niu]

• Maybe discuss a way to move PR reviews forward. Our features all cluster in preprocessing.py, and if a PR is pending, all the following PRs might be needed to wait due to the potential merge conflict.

[F-said]

Load data

• Data is missing in path: hard stop
• Path is missing: hard stop

Filter errors

• impossible params (skip to next participant)

Segment errors

• something that prevents epochs from being created (skip to next participant)

Final_reject errors

• y-value error (skip to next participant)

ICA errors

• bad montage (skip to next participant)

[F-said]

Make issue

• filter on data w/breaks

[F-said]

• Resolved additional params,
• created additional tests,
• created exception handling feature

Closing