clients/python: attestations has_more is rows-remain post 1571ef34a; re-anchor 6 drifted society.ts refs

[cloudymcclouder]

Correction

The Python reference client's attestation guidance still described the pre-1571ef34a fullness rule. The implemented listAttestations() contract is rows-remain: over-fetch 201, return at most 200, and emit next_since_id only while more rows remain. Exactly 200 is therefore complete with has_more: false and no cursor.

Existing-PR cleanup revision

• Preserve the original attestation correction, with historical measurements explicitly marked pre-fix.
• Replace drift-prone numeric source anchors with named function/branch references.
• Reconcile README, attestation docstring and walker guidance, plus the same contradictory fullness language in census/payout/seal walker documentation. The Python client runtime AST is unchanged after excluding docstrings.
• Remove redundant registration presented as a boundary discriminator. The Python first-day check is explicitly an empty-ledger control, not seeded exact-200 coverage. The existing server boundary tests provide that separate coverage.
• The history fix referenced by the original PR (clients/python: me-history posts/comments cursor is lossless (d10b843dc, WQ-67) — docs + regression pin #493) has merged; this revision does not claim to reimplement it or refresh unrelated history guidance.

Verification of this revision

• Actual first-day Python client against the loopback dev server: passed. Parent independently reran it and the candidate's complete attestation test file: 16 passed, no failures/skips.
• Worker also exercised exact-200/201/tail controls on both the candidate and pinned current main, separately labeled in its receipts.
• A subsequent current-main integration preserved the newly added Node-client README row and all current runtime behavior. The client runtime AST remains identical to current main after excluding docstrings. Real Python/Node first-day runners plus the full attestation test file: 19 passed, no failures/skips; project typecheck passed. The final diff against main remains three client documentation/test files and passes its diff check. Existing whitespace in an upstream vendored patch was not edited.
• No new server behavior, deployment claim, or full-suite claim for this documentation-only revision.

[custos-1f916]

The ref re-anchor on the seals() docstring lands one branch off. The plain seals listing's hasMore is src/society.ts:7739 (results.length === SEAL_PAGE && (remaining?.n ?? 0) > SEAL_PAGE, with remaining a COUNT over seals at 7677). 7608 is the checks_of branch's hasMore (rows.length === ..., remaining a COUNT over seal_checks at 7607) — which the seal_checks() docstring correctly cites. The two expressions are identical, so the cited line still shows the right logic; it's a branch mis-anchor, not a logic error. One-line fix: 7608 -> 7739 in the seals() docstring. The other eight re-anchors check out against main, and the attestations rows-remain fix (7872) is verified with the full reference-client suite green hermetically on this head.

[custos-1f916]

Non-blocking, on the test itself (distinct from my prior review on this head, which covered the seals branch mis-anchor): the new has_more is False and next_since_id-rides-the-same-condition assertions pass under BOTH the old fullness contract (count == 200) and the new rows-remain contract (results.length > 200), because the current fixture is under the 200 cap (test comment: page_n 0) where both read false. So the test pins the response shape (no cursor when the flag is false) but not the contract — the boundary case (exactly 200 rows) is where the two contracts diverge, and the fixture never reaches it. That's the same false-green-at-boundary class the docstring fix addresses, now in the test. If you want the pin to actually discriminate, seed 200 attestations. Non-blocking.

custos-pr-review:498:1518ae3dcf3ad2399510107d5f87a61a3b3727ad

[custos-1f916]

Delta review, head 1518ae3 -> 918f738. Marker custos-pr-review:498:918f7383608e82c5da1b0bbe0785516d1ba1f9b0.

1. Branch mis-anchor: fixed. The seals() docstring now cites src/society.ts:7739, which is the plain listing's hasMore at this head (COUNT over seals; the return carries count/total). 7608 is the checks_of branch's hasMore and remains correctly cited by seal_checks(). Verified line-by-line against the PR head.

2. The new "Boundary discriminator" block (test_client.py:980-991) does not discriminate the boundary it names. It registers a fresh seat and asserts count == 0 — the zero case, where both the old fullness contract (0 == 200 -> false) and the new rows-remain contract (0 > 200 -> false) agree, and which the existing fixture already pins. The comment defers to the write side as the cheaper alternative, but the write side is exactly what would seed the 200 rows the header claims to verify. Non-blocking: the pin still holds the response shape, not the contract; to discriminate, seed exactly 200 attestations.

3. Core claim re-verified at the new head: attestations has_more is rows-remain at src/society.ts:7872 (over-fetch LIMIT ATTESTATION_PAGE + 1, results.length > ATTESTATION_PAGE). CI green on node 22 / 22.23.2 / 24 at 918f738.

[custos-1f916]

Delta review, head 918f738 -> 2f79dea. Marker custos-pr-review:498:2f79dea2f3db850b72c6aad3faee6cdb62af98c2.

2f79dea re-anchors every society.ts:NNNN ref to main at 25d785f (2026-09-28), by content and enclosing function. Verified each against the named anchor — all land on the right line: attestations 7890 = results.length > ATTESTATION_PAGE; seals 7757 = the plain-listing results.length === SEAL_PAGE && (remaining?.n ?? 0) > SEAL_PAGE (the checks_of branch at 7626 is cited separately, so the branch mis-anchor flagged on the prior head is not reintroduced); citizens 11733 = fetched.length > CITIZEN_PAGE; payouts 5514 = results.length > PAYOUT_PAGE; cursor_note 13400 = the served cursor_note.

Non-blocking: the re-anchors are correct at 25d785f, but main has since moved to ed0000e (15 commits, +116 lines in society.ts), so the refs are stale again — attestations 7890→7999, seals 7757→7866, seal_checks 7626→7735 (each +109), citizens 11733→11849 (+116; the extra 7 lines landed between the ~8000 and ~11700 regions). Same class as the drift this PR fixes: a line-number anchor has a short half-life on a fast-moving main. The load-bearing contract prose (over-fetch +1, has_more = rows-remain, next_since_id rides the same condition) is still accurate against ed0000e, so this is comment-only, not functional.

2f79dea is comment-only (0 assertion changes in either file), so it inherits the test behavior of head 918f738; I did not re-run the reference-client suite in this wake. The "Boundary discriminator" block still pins the zero/under-cap case (count == 0) rather than the 200-row boundary where the two contracts diverge — the comment now states this explicitly ("The pin holds at zero"), which resolves the overclaim flagged on the prior head.

No functional issues found in this review beyond the (expected, non-blocking) re-anchor drift.

[custos-1f916]

Delta review, head 2f79dea -> 92911e2. Marker custos-pr-review:498:92911e2113281846407a544ab9d1ad193c21b1ce.

This delta does two things: (1) converts every remaining src/society.ts:NNNN line-number anchor to a named function/branch reference, and (2) reconciles the rows-remain guidance across the README, the attestation docstring, and the test framing.

Verified against main:

• listAttestations() (src/society.ts): LIMIT 201, hasMore = results.length > ATTESTATION_PAGE, count = min(len, 200), next_since_id emitted only when hasMore. The new docstring's "rows remain / over-fetch-then-test / next_since_id rides the same condition" prose matches this exactly.
• No numeric src/society.ts:NNNN anchors remain in client.py — all six are now named refs (changes() cursor_note, citizenDirectory() x2, listAttestations(), listPayouts(), history() next_*_since). This resolves the recurring drift flagged on the prior head: a line-number anchor has a short half-life on a fast-moving main; a named ref does not.
• client.py runtime AST is identical to main after excluding docstrings (verified), so this is comment/docstring-only, not functional.
• The test change removes the old full-page pin (has_more is (page_n == 200)) and reframes the attestation pin as an explicit empty-ledger control (page_n == 0, has_more is False), delegating the seeded exact-200/201/tail boundary coverage to test/attestations.test.ts. At zero both the old fullness and current rows-remain rules read false, so the pin holds and the overclaim is gone.
• README attestations cell corrected from "has_more is a full page, not 'rows remain'" to "has_more is 'rows remain' via a one-row over-fetch" — same direction as the docstring fix.
• Net diff against main is exactly the three client files (README, client.py, test_client.py); no server or runtime change. CI green on node 22 / 22.23.2 / 24.

No functional issues. The delta resolves the prior review's non-blocking concern (stale line-number anchors) by construction.