Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: pipe_base
-
Labels:
-
Story Points:16
-
Epic Link:
-
Team:Data Release Production
-
Urgent?:No
Description
Using UUIDs for all of the IDs in QuantumGraph seems like it makes sense, and is something we would ideally do before we start saving any QuantumGraph stuff (including post-execution provenance) to the data repository.
I bet it would also sidestep some of the thorny hashing problems inside QuantumGraph, at least if we can use UUIDs for comparison without changing comparison semantics.
I think it'd be okay to break backwards compatibility with old QGs, as we've never sold them as something that would have a long shelf-life, but it'd be ideal if we could avoid that or at least have a deprecation period.
Attachments
Issue Links
- relates to
-
- Done
Activity
Michelle, can you look at the BPS bit, I still need to do a bps run using this branch, but I am trying to get this out sooner than later to start getting peoples eyes on it.
Jim, can you take a look at the other bits.
Again there might be a few changes still coming if something else comes up, but this gets it out for people to start looking at. It has passed jenkins and ci.
I've taken a look at all of the PRs besides ctrl_bps, and left comments on GitHub. I'm not marking Reviewed yet here both because I don't know if ctrl_bps has been reviewed and because I have requested changes in daf_butler. But Nate Lust and I have also discussed this live, and I think we agree on the path forward.
I had just commented on the ctrl_bps pull request. Not all places in ctrl_bps were changed so it was definitely not ready to be merged.
I'd been considering this reviewed after a few out-of-band discussions with Nate Lust, and I hope he's been considering it reviewed as well. Past time to mark it as such.
QuantumGraphs have a version number in them, so we at least can have a forward path, or a message about needing to rebuild an old graph style instead of just getting failures.