Skip to content

Refinement note: content-accessor-guards Rule 1 gap (docs only) #276

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
dk107dk wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Refinement note: content-accessor-guards Rule 1 gap (docs only) #276

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 47 additions & 0 deletions specs/references_v3/notes/deferred_work_bucket_list.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,53 @@ candidates for that re-audit, all still unconditional/immediate raises in
same "was a pointer actually applied within each partition" reasoning
the literal-root fix used, not a blind port.

**Ready for a quick decision, then a well-specified build — found
2026-08-29/30 while checking whether this was safe to build solo
overnight, deliberately NOT built without a check-in first.** FILES'
own GROUP-mode + field-accessor fix (2026-08-29, this session) narrowed
`query()`'s `if partitioned and (has_range or (pointers and has_
manifest)): raise` to keep `has_manifest` (whole-resource) rejected,
precisely BECAUSE the existing `ambiguous_content_read=has_manifest and
len(selected_candidates) > 1` flag just below it (in the shared return)
is a NAIVE count check — it does not know whether `partitioned` is
`True`, so it cannot currently tell "several groups, each already
reduced to exactly one via its own pointer" (legitimate, per this very
bucket-list item's own CSVPATHS precedent) apart from "several raw,
unreduced candidates sharing one accessor" (a genuine Rule 1
violation). Confirmed field accessors are unaffected by this — they
never set `has_manifest`, so `ambiguous_content_read` is always
`False` for them regardless of count, which is why narrowing the
rejection to field-accessors-only was safe to ship without touching
this flag at all (three PRs, 2026-08-28/29, none of which needed to
touch `ambiguous_content_read`).

To also let `:manifest()` (and any other whole-resource content
accessor) through here, the flag itself needs to become partition-
aware — something like `ambiguous_content_read = has_manifest and
len(selected_candidates) > 1 and not partitioned` (or, more precisely,
"and no per-partition pointer reduction actually happened") — mirroring
the exact reasoning CSVPATHS' own already-shipped fix used for the
identical shape. The SAME naive-count-check gap almost certainly also
affects: `ResultsReferenceFinder3._star_group_and_reduce()`'s own
`accessor is not None and pointer is not None` check (still an
unconditional raise, not a deferred count check at all yet); and
possibly RESULTS' own literal-root `match_all and accessor is not
None` check (the two are explicitly linked above, "should be decided
together"). FILES' `'*'`-traversal `:manifest()` rejection (this
session's other two PRs) is the SAME question at one more remove
(GROUP mode there also produces one candidate per file_home group).

**What would unblock building this**: confirm the fix direction above
is right (a partition-aware flag, not a blind removal of the
`has_manifest` guard), then decide scope — FILES only (continuing this
session's own thread) vs. also RESULTS/CSVPATHS in the same pass
(bigger, but closes the whole cross-datatype question the "should be
decided together" note above already flags). Once scoped, this is
concrete, mechanical work with a clear correctness bar (the CSVPATHS
"several groups, each pointer-reduced" test case already describes the
must-not-regress behavior) — not open design the way the predicate-
argument mechanism or RESULTS' `:groups()` semantics still are.

## Predicate-argument field accessors (`:on_arrival(:not_none())`) — filter half built for `:idchain()`, generic mechanism still not built

David, 2026-08-21, drafting the compendium's replacement `:manifest()`/
Expand Down