Reporting dependencies

Reporting dependencies

All modules should have clear documentation explaining how module dependencies are managed and how to activate any software environment(s).

In addition to this documentation, we recommend also recording your software dependencies directly from your code using some of the strategies detailed below.

Reporting dependencies from R code

Using sessionInfo() in R

The R programming language includes a helpful function called sessionInfo() that prints information about the R version and loaded packages.

In R Notebooks

Include sessionInfo() in the final chunk or cell of any computational notebook using R to ensure that the report output will include information about R and package versions.

In R scripts

Not using a notebook? No problem! You can capture the output of sessionInfo() in a text file by including the following in your script:

sink("sessionInfo.txt")
sessionInfo()
sink()

This will result in the contents being written to a file called sessionInfo.txt in your working directory, so be sure to use unique filenames if multiple scripts in your analysis are writing out the contents of sessionInfo().

You can post the contents of these files on your pull requests or commit them to the repository.

Reporting dependencies from Python code

Using session_info.show() in Python

The session-info Python package can be used to report version information about Python and loaded modules. If you created a module using --use-jupyter or --use-python, session-info was automatically included in the module's conda environment.

Import session-info by placing the following in the Setup section of your Jupyter notebook or in the load packages section of your script:

import session_info

In Jupyter notebooks

Include the following in the final cell of any Jupyter notebook using Python to ensure that the notebook reports version information:

session_info.show()

In Python scripts

Using a script instead?

Import contextlib by placing the following in the load packages section of your script:

import contextlib

Then, you can use the following to write the output of session_info.show() to a text file called session_info.txt in your current directory:

session_info_path = "session_info.txt"

with open(session_into_path, "w") as f:
    with contextlib.redirect_stdout(f):  # direct the session_info output to a file
        session_info.show(dependencies=True)

If you're using this approach with multiple scripts in your module, be sure to use unique file names when naming your session info text files.

You can post the contents of these files on your pull requests or commit them to the repository.