Declarative pipelines based on Filetrees. A pipeline is defined by:

• A file-tree, which defines the directory structure of the inputs and outputs of the pipeline. A tutorial on these file-trees is available here.
• A set of recipes describing how all the pipeline outputs are produced. A tutorial on writing these recipes is available here. Fsl-pipe will stitch these recipes together to produce any user-selected output files. Resulting jobs will either run locally, run distributed using dask, or be submitted to a cluster using fsl-sub.

An example diffusion MRI pipeline using fsl-pipe with detailed comments is available here.

Installation

pip install fsl-pipe

Any bug reports and feature requests are very welcome (see issue tracker).

Setting up local test environment

This package uses uv for project management. You will need to install uv to develop this package.

First clone the repository:

git clone https://git.fmrib.ox.ac.uk/fsl/fsl-pipe.git

Then we can ask uv to set up the local envoriment.

cd fsl-pipe
uv sync

Running tests

Tests are run using the pytest framework. This will already be available in the uv virtual environment.

uv run pytest src/tests

Compiling documentation

The documentation is build using sphinx.

cd doc
uv run sphinx-build source build
open build/index.html

Contributing

Merge requests with any bug fixes or documentation updates are always welcome.

For new features, please raise an issue to allow for discussion before you spend the time implementing them.

Releasing new versions

• Ensure CHANGELOG.md is up to date.
• Edit the version number on pyproject.toml
• Create a new tag for the version number
• Push to gitlab (including tags).
  • The pipeline from the tag will automatically publish fsl-pipe to pypi.
• Upload code on conda-forge using their automated release detection.