Code management and sharing

Code management

• All code should be managed using a version control system, preferably git.

  • There are many resources online for learning git. A good starting point is Git for beginners: The definitive practical guide.
  • Git questions can be addressed to the #git channel on the Poldracklab Slack.

• Code should be regularly pushed to a remote server, preferably Github.

  • Our default is that all repositories for research projects will be public, unless there is a strong reason to keep it private.

• Any substantive code copied or adapted from another source (e.g. stackoverflow) should be attributed to the source (including the URL of the source).

• Care must be taken to avoid sharing identifiable data, private passwords, AWS credentials, and other confidential information when using Github.

  • Any private information should be stored in a separate text file that is read in by the code, and those files should be added to the .gitignore file so that they will not be checked in.
  • Any accidental breach should be dealt with and reported to Dr. Poldrack and other relevant parties immediately.

Code review

• All coding projects should be reviewed prior to paper submission.

Code sharing

• All code should be made available via Github, upon submission of the preprint at the latest.
• All shared code should be accompanied by an open source license.
  • Our general preference is for permissive licenses such as MIT or Apache, but other licenses may be used (e.g. in collaborative projects that use GPL-style licenses).
  • If code is used from other projects, the licensing requirements for that code should be obeyed.
• A Github release should be generated for code related to any paper submission, containing the exact code used to implement all of the included analyses for the submitted version.
  
• The Github repository should be linked to Zenodo in order to generate a DOI for each release, and this DOI should be included in the manuscript (rather than the Github link).