Zulip Chat Archive
Stream: Equational
Topic: CI task suggestion: Report changes in known facts
Joachim Breitner (Oct 02 2024 at 07:27):
Here is a nice task for those who like writing Github Actions:
Write a bot that builds the project at the base branch and the PR branch (maybe with debug.byAsSorry
to be fast), collects the results, and reports how our summary numbers (implications, anti-implications, implied implications, implied anti-implications) have changed in this PR.
Shreyas Srinivas (Oct 02 2024 at 09:10):
Is there already a script in the main branch that outputs this info?
Shreyas Srinivas (Oct 02 2024 at 09:10):
I might get to it today or tomorrow
Shreyas Srinivas (Oct 02 2024 at 09:10):
I see several threads on this topic, hence the question
Joachim Breitner (Oct 02 2024 at 09:11):
extract_implications
kinda does it, but produces huge output (i.e you'd still have to tally the lists).
Maybe add a --stats
mode to it that just prints the stats in json format,and post-process that.
Daniel Weber (Oct 02 2024 at 09:33):
There is lake exe extract_implications outcomes --hist
Joachim Breitner (Oct 02 2024 at 09:54):
Thanks!
For finding the right base branch,
git merge-base origin/${{ github.base_ref }} HEAD
is used for a similar task in other projects where I have seen that.
May need
- uses: actions/checkout@v4
with:
# fetch full history so that git merge-base works
fetch-depth: 0
# fetch PR commit, not predicted merge commit
ref: ${{ github.event.pull_request.head.sha }}
in the workflow
Terence Tao (Oct 02 2024 at 13:22):
Should I create an explicit task for this, or is this something that someone can just do directly with an issue or pr?
Shreyas Srinivas (Oct 02 2024 at 13:23):
I think this is easiest for project contributors because we already have the permissions.
Shreyas Srinivas (Oct 02 2024 at 13:23):
I am on a different time crunch which is why I am not doing it right away
Terence Tao (Oct 02 2024 at 13:33):
Ok then i will not make an explicit project task for this.
David Renshaw (Oct 02 2024 at 13:51):
and reports how our summary numbers (implications, anti-implications, implied implications, implied anti-implications) have changed in this PR
Do you mean mean a bot that would post on zulip? Or something that updates the webpage? (Both sound to me like good ideas.) Or something else?
Joachim Breitner (Oct 02 2024 at 14:37):
David Renshaw said:
and reports how our summary numbers (implications, anti-implications, implied implications, implied anti-implications) have changed in this PR
Do you mean mean a bot that would post on zulip? Or something that updates the webpage? (Both sound to me like good ideas.) Or something else?
I’d expect a comment on the PR
David Renshaw (Oct 02 2024 at 18:12):
Here's my go at making something that updates the website: https://leanprover.zulipchat.com/#narrow/stream/458659-Equational/topic/proposal.3A.20add.20a.20dashboard.20page.20to.20the.20website/near/474364688
Shreyas Srinivas (Oct 02 2024 at 18:13):
Will merge right away
Shreyas Srinivas (Oct 02 2024 at 18:14):
This helps people see the graph live and then if there are suggestions for improvement we can change it later
Shreyas Srinivas (Oct 02 2024 at 18:14):
(It passes CI)
Terence Tao (Oct 02 2024 at 18:22):
Suggestions for additions to the dashboard:
- Links to files for each of the statistics given (i.e., a link to a text file of explicitly true implications, a list of implicitly true implications, etc.).
- Stats about equivalence classes: how many equivalence classes of equations are there? How many unknown implications are there after quotienting by equivalence? See equational#176.
- A list (and count) of "equivalence-closing" implications - implications that, if proven, would generate an equivalence because the converse is already established. See equational#188.
- Similar stats, but only for the equations considered in
Equations.lean
. Here one can also add a Hasse diagram, as it is small enough to be human digestible. See equational#183.
Last updated: May 02 2025 at 03:31 UTC