Make `nipoppy init` create an empty code directory too

[michellewang]

From a discussion with @nikhil153 today.

• The user would specify both a DATASET_ROOT and a CODE_ROOT (for example). The code directory should have a structure that mirrors the dataset structure, except they should contain code/scripts instead of data:
  • tabular: for manifest-generation, tabular bagel generation
  • dicom_reorg: for wrangling raw, pre-BIDS imaging data
  • analysis: for downstream analyses
  • Others?
• Maybe also a requirements.txt file pre-populated with Nipoppy
• Maybe also run git init if the user has Git installed

I think the idea would be that this is a starting point for "good practices", but that users are not obligated to use these directory names exactly. Unlike the dataset layout which is mandatory since the Nipoppy tools expect it.

[michellewang]

Alternative: follow abstractions/"steps" listed in the slides

• curation
• processing
• tracking
• extraction
• analysis etc.

[Remi-Gau]

• Maybe also run git init if the user has Git installed

related to this:

I already have the cohort creator turning the nipoppy "hierarchy of folders" into a datalad sub-datasets with a code diretory.

Maybe this could be an option of the init CLI to create datalad datasets if datalad is installed and if requested by the user.