Introduction

**GOAL: generate, from a YAML description file, workflows for Pegasus, nextflow or Snakemake. **

Existing WMS (Workflow management systems) generate specific learning curves for bioinformaticians which will also engage strategic choices of development¹.

hpc_runner offers to directly use bash script bricks in a declarative format to form a processing pipeline, drawing inspiration from the syntax and logic of the .gitlab-ci file, used to run CI/CD pipeline on GitLab.

The tool is meant to be simple: there is no conversion from bash to another language. Script bricks can be directly added simplifying the development and the sharing of workflows.

Installation

Requirements:

• Python 3.8 or above.
• graphviz package, with dot tool.

Create a Python virtual environment and install the hpc_runner package from this repository :

python -m venv venv
pip install 'hpc_runner @ git+ssh://git@gitlab.cnrgh.fr/lbi/hpc_runner.git'

Running

Once installed, you can run hpc_runner on the provided example to check that it runs correctly:

cd examples/prepare_gisaid
hpc_runner --hpc-config prepare_gisaid.yml

In this example, hpc_runner uses the default backend. You can also define a custom backend configuration file and use it with the --hpc-backend option as follows :

hpc_runner --hpc-backend BACKEND_PATH --hpc-config myappconfig.yml

See the official Documentation for detailed information.

For developers

You may create a conda environment like this:

conda create -n hpc python=3.8
conda activate hpc
python3 -m pip install .[dev]

Building hpc_runner

The following commands will build the package in an isolated environment, generating a source-distribution and wheel in the directory dist/.

pip install build
python -m build

Available scripts :

• make deps : install development dependencies of the hpc_runner package,
• make check : run checks on the source code (linter, type checker, ...),
• make build : build the package in an isolated environment, generating a source-distribution and wheel in the directory dist/,
• make install : install the package inside a virtual environment,
• make test : run the tests against the package, and generate a coverage report in htmlcov/,
• make uninstall : uninstall the package from the virtual environment,
• make clean : delete all files and directories created by the other targets,
• make docs : generate an HTML documentation in docs/build/.