Repository navigation
proofs: merge on partial sections is not a join for the information order - #380
hyperpolymath wants to merge 2 commits into
Conversation
…rder Late review finding on #370 (merged): the module header and the bridge-status row called the resolve-parametric merge a join-semilattice of partial sections. Under `_⊑_` every present value must stay unchanged, so two sections with different values at one index have no common upper bound at all; merge-comm/assoc/idem are algebraic laws of a resolver-parametric merge, not join laws. - EchoPartialSection.NoJoin.no-common-upper-bound proves it (Bool-indexed counter-model, `--safe --without-K --cubical-compatible`, no postulates); pinned in Smoke.agda. - Module header, docs/bridges/cross-repo-bridge-status.adoc and MAP.adoc now say "resolver-parametric merge algebra" and cite the lemma. - CHANGELOG entry. Checked: EchoPartialSection.agda typechecks with Agda 2.7.0.1, stdlib 8326c74, absolute-zero f486c29 (scripts/setup-agda.sh); the warning count is unchanged (12, all the existing UnsupportedIndexedMatch class, none in the new block). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01J4QvDizuhJtQDcxHxM2wj8
|
No actionable comments were generated in the recent review. Summary
Merge Risk:
|
|
The full local run has finished and passed. I used Agda 2.7.0.1, stdlib
Generated by Claude Code |
Brings in #376 (absolute-zero repinned to 8a3162a), #378 and #379 (doc reference fixes touching Smoke.agda, MAP.adoc and the bridge-status page), which this PR also edits. Merged cleanly; re-verified below. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01J4QvDizuhJtQDcxHxM2wj8
Follow-up to #370, from CodeRabbit's last review, which arrived after the merge.
The overclaim
The
EchoPartialSectionheader and the bridge-status row called the resolve-parametric merge a "join-semilattice of partial sections". Under_⊑_, every value present in a section must stay unchanged in anything above it. So two sections holding different values at one index have no common upper bound at all.merge-comm,merge-assocandmerge-idemare laws of a resolver-parametric merge, not join laws.Changes
EchoPartialSection.NoJoin.no-common-upper-boundproves this with a Bool-indexed counter-model. It is checked by Agda, not just stated in prose:--safe --without-K --cubical-compatible, no postulates. It is pinned inSmoke.agda.docs/bridges/cross-repo-bridge-status.adocanddocs/echo-types/MAP.adocnow say "resolver-parametric merge algebra" and cite the lemma.CHANGELOG.adoc.No new module, so
All.agdaand the kernel-note classification are unchanged.Validation
8326c74and absolute-zerof486c29, installed byscripts/setup-agda.shwith checksums verified:All.agda,Smoke.agdaandAuditSmoke.agdaexit 0;scripts/kernel-guard.shpasses;EchoPartialSection.agdakeeps the same 12 warnings, all of the existingUnsupportedIndexedMatchclass, none in the new block.main(cecc896).maingained CNO bridges: repin absolute-zero to 8a3162a (pc-driven machine) #376 (absolute-zero repinned to8a3162a), ci(hypatia): pin the scanner to hypatia main 6d8ab66 #378 and docs: fix references to files that no longer exist #379, which touchSmoke.agda,MAP.adocand the bridge-status page. The merge applied cleanly. Re-run with the toolchainscripts/setup-agda.shnow pins (absolute-zero8a3162a):All.agda,Smoke.agdaandAuditSmoke.agdaexit 0, andscripts/kernel-guard.shpasses.fe5f462. The Agda workflow passes (run 44129193750), as do governance, CodeQL, secret scans and Hypatia. CodeRabbit produced no actionable comments.Generated by Claude Code