Open yymao opened 6 years ago
Well - you could argue that the only things the table really needs are the notebook titles and the links and badges; the owner information can be read from the notebooks themselves, but in any case, I think we can scrape out the owner information straightforwardly if a template is followed to use the "Owner:" construct (and could declare the owner to be unknown if we fail).
I think the README with badges adds a lot of value to a directory listing - its much easier to see which notebooks are failing. So, I think I am still in favor of having beavis
optionally write a README file. It would be nice for this README to be on the master
branch, though, so you don't have to click through to see what's failing...
@drphilmarshall and I were just discussing (on a StackClub call) the idea of having beavis generate a top-level summary in a markdown file on the rendered branch, which would look something like this (with appropriate link targets added):
Notebook | Owner | Status | Exception |
---|---|---|---|
Basics/Calexp_guided_tour | David Shupe | Passing | |
Deblending/lsst_stack_deblender | Fred Moolekamp | Failing | ImportError: cannot import name 'MultibandExposure' |
The StackClub use case we had in mind was tracking the work needed to migrate all notebooks to a new pipeline release.
@drphilmarshall @yymao I have some work-in-progress to push, but am not a collaborator on this repo. Please add me, or let me know if you prefer that I make a PR from a fork when I am all done.
Thanks @dkirkby ! I just added you with Write access to the @LSSTDESC/beavis-developers team - looking fwd to seeing your proposed upgrade (esp how you pull out the error messages). I think this'll be helpful in the DESC DC2 analysis world as well.
Update after today's stack-club meeting:
--no-summary
option (summary is created by default now).summary.md
at top-level of rendered branch.For an example, see here.
Next step is to parse the log file to pick out the exception that prevented a notebook rendering. This might be much easier to do in a small python script than bash: @yymao @drphilmarshall are you hoping to avoid using any python in this repo?
@dkirkby: Personally I think everything would be easier to do in a python script :) Bash was Phil's original choice and it sticks around.
I am opening this issue to follow up this discussion, but now I am in fact less certain if we actually need to do this for the following reason:
Hence, I labelled this issue as a question. Thoughts?