nxfvars 0.1.0

Installation

git clone git@github.com:grst/nxfvars.git
cd nxfvars
pip install .

PyPI and conda will follow at some point.

Usage

Add the nxfvars.nf script to your pipeline. Import the nxfVars function as follows:

import { nxfVars } from "./nxfvars.nf"

In each process where you want to use nxfvars, add the following line to the script section, right before executing the python script or notebook:

${nxfVars(task)}

For instance:

process TEST {
    cpus 2

    input:
    val bar
    path foo
    path nxfvar

    script:
    """
    ${nxfVars(task)}
    jupyter nbconvert notebook.ipynb --execute --to html
    """
}

In your python script, add

from nxfvars import nxf

You can now retrieve inputs, params and task variables as follows:

nxf.input('bar')
nxf.task('cpus')
nxf.params('some_param')

To execute a notebook interactively (i.e. outside of nextflow), you may define default values:

nxf.input('bar', "/path/to/some/file.foo")

For a full example pipeline, see the example directory.

How it works

The groovy library serializes the nextflow variables and encodes them into an environment variable, which is injected using ${nxfVars(task)}. See also my blog post about Nextflow’s implicit variables this and task.

The nxfvars Python library decodes the env var and makes the values conveniently accessible

Ideas for the future

• This can obviously be easily ported to other languages, e.g. R.

• The library could be extended to serve as a minimal argument parser, i.e. that a Python script containing nxf.input("foo") could also be executed (outside Nextflow) using scripyt.py --foo=bar.