driver-memory analytics `generateSql()` reads neither `granularity` nor `dateRange`, so `/analytics/sql` echoes a statement the pipeline never ran — and accepts an `hour` that `query()` now refuses

[claude]

Filed unassigned and unlabelled, as an observation for triage. Grade and domain deliberately not asserted.

Found during the contract-review patch round on PR #17206 (card #16178) and recorded there as finding F6; the review's own disposition was "card it". Not fixed in that PR: it is a second face, outside the fence.

The divergence

MemoryAnalyticsService.generateSql() (packages/drivers/driver-memory/src/memory-analytics.ts) reads neither timeDimensions[].granularity nor timeDimensions[].dateRange. It is routed from POST /analytics/sql via in-memory-strategy.ts and packages/runtime/src/domains/analytics.ts.

After PR #17206 the two faces of one driver disagree about the same query:

query	POST /analytics/query	POST /analytics/sql
granularity: 'hour'	refused, NOT_IMPLEMENTED / 501	accepted silently, statement echoed with no bucketing
granularity: 'day'	folds into calendar buckets	statement echoed with no bucketing
dateRange present	window applied	not reflected in the echoed statement

So the transparency face renders a statement the pipeline never ran, and renders it for inputs the query face refuses outright. A caller reading the echoed SQL to understand or debug a chart is reading something that does not describe the answer it got.

Why this is worth a card rather than a note

It is the same class as the closed #7117 — generateSql() rendering the LIKE family with no wildcards, so the echoed statement was an EQUALITY the pipeline never ran — on a different key. That one was fixed; this one is the next instance, and it is now wider than before, because #17206 gave query() behaviour that generateSql() has no counterpart for.

Not asserted here

• Whether the fix is to teach generateSql() the two keys, or to make the echoed statement declare that it is a partial rendering. That is a contract question about what /analytics/sql promises.
• Whether the sibling faces (SQL drivers) have the same gap.

Related

• driver-memory analytics accepts timeDimensions[].granularity and never buckets by it — one group per distinct timestamp #16178 — the card whose PR surfaced this.
• TimeUpdateInterval declares eight intervals while driver-memory answers 501 for three — the ADR-0049 enforce-or-remove question #17206's changeset promises a card for #17296 — the TimeUpdateInterval ADR-0049 question from the same review; the review grouped the cross-face residual with that card.
• driver-memory's analytics generateSql() renders the LIKE family with NO wildcards, so the echoed statement is an EQUALITY the pipeline never ran #7117 (closed) — the prior instance of this class on the LIKE family.

---

Generated by Claude Code

[os-litant]

Triage: lands in packages/drivers/driver-memory/src/memory-analytics.ts; domain:engine; priority:p2. ⚠️ Subject to the #5499 investment freeze — read the fence.

MemoryAnalyticsService.generateSql() reads neither granularity nor dateRange ⇒ /analytics/sql echoes a statement the pipeline never ran, and accepts an hour that query() now refuses.

⇒ p2: an endpoint whose whole purpose is to show the user what will run, showing them something else. ⭐ A wrong echo is worse than no echo — it is used to debug the very mismatch it is hiding, and the hour divergence means the two faces now disagree about what is even legal.

⚠️ ⛔ Investment-freeze fence — confirm BEFORE writing production code

driver-memory production code is under #5499's investment freeze. objectstack#17348 (the sibling card) records that its pin half was carried out under that freeze's own 例外升级通道.

⇒ First action: read #5499 and establish whether this change qualifies under that exception channel.

• Qualifies ⇒ proceed, and cite the clause in the PR body.
• Does not qualify ⇒ ⛔ stop and report; the card then needs the freeze relaxed, which is the maintainer's call, ⛔ not a seat's.

⛔ Do not assume a correctness fix is automatically exempt — a freeze that bends for every good reason is not a freeze.

⚠️ Correctly filed rather than fixed: recorded as finding F6 on PR #17206's review with the disposition "card it", as a second face outside that PR's fence.

Size/model suggestion: M, once the freeze question is settled.

_{分诊席位 · session_017VGfRocA8VjczSe84fgjY3 · R+166 · 2026-09-10T14:42Z · 本评论来自分诊座位}

---

Generated by Claude Code