A better test data repository is needed.

[brettviren]

As a start, the new testing system introduced in #199 uses a "test data repository" that gets built into build/tests/ by downloading tar files from a web site (mine at BNL). This is easy but it means the versioning of the test data files is not tied to the version of WCT. At least two problems arise:

• Updating data files will break old software versions.
• It is easy to write/change tests requiring new data and forget to update the test data repo.

A likely solution is to serve the test data repo via git. Three likely options:

• plain git. This must contend with Git's inherent "large file problem".
• git-lfs This is probably the next simplest but results in at least 2 copies of every file on the client side.
• git-annex This is most efficient but has some learning curve.
• datalad A simplifying layer on git-annex but requires more software, though it be only a pip install away

Somewhat tied to the choice is how to serve

• Use www.phy.bnl.gov can accommodate plain git or Gitea + git-lfs or the branched gogs server gin for git-annex support
• Use git-annex via SSH or other remotes
• Rely on Gin for free cloud for git-annex
• Note: GitHub supports git-lfs but the allowed bandwidth under free tier is likely too restrictive.

[brettviren]

Status update

For now, a trivial, low-feature "repository" is provided by distributing tar files from

https://www.phy.bnl.gov/~bviren/tmp/wcttest/data_repo/

More info at:

https://github.com/WireCell/wire-cell-toolkit/blob/master/test/docs/datarepo.org

[brettviren]

#338