Repository navigation
spec(type-compat): lift the oracle RAW(16) → lookup / master_detail pairing from lossy once the driver binds uuid through HEXTORAW #31592
Description
domain:server#4 · os-steve · session_01ADHkmP8JJo4zF2sjTBRwnX
carries item 3 of #31390; #31390 keeps items 1, 2 and 4 (the driver-sql oracle bind, the malformed-uuid refusal, the live pins).
Filing gate: a self-split of a queued domain:server card (execution-duties.md: a queued card holding another lane's item is split by the seat that finds it). Item 3 edits packages/spec/src/data/type-compat.ts, and the server lane makes no packages/spec edit beyond prose. ⛔ Not a claim. No domain:* label: triage routes it (expected domain:spec).
Ruling-ref: ADR-0015 §4.6 (as cited on #31390)
Blocked-by: #31390
Dedupe (open AND closed): "HEXTORAW", "RAW(16) lossy oracle lookup" → #31390 only.
维护者速读
Oracle 的 RAW(16)(SYS_GUID())主键现在在类型兼容表里对 lookup / master_detail 标为 lossy,因为驱动还不能按 uuid 字符串过滤或写入它。#31390 补上驱动绑定后，这里把这一对从 lossy 改为兼容，并更新对应 pin 与 demoteExact 注释。放宽接受面，需要契约复审。
Task
-
packages/spec/src/data/type-compat.ts: drop the oracle /uuiddemoteExactrule (or narrow it) soRAW(16)→lookup/master_detailanswers compatible again, and rewrite the JSDoc that says "Binding throughHEXTORAWis driver-sql (oracle): bind uuid text to RAW(16) through HEXTORAW, so a lookup over a SYS_GUID key filters and writes by its uuid string, and type-compat can lift the pairing from lossy to compatible #31390". - Update the pin(s) that hold today's
'lossy'answer. - Only after driver-sql (oracle): bind uuid text to RAW(16) through HEXTORAW, so a lookup over a SYS_GUID key filters and writes by its uuid string, and type-compat can lift the pairing from lossy to compatible #31390's bind has landed on
main: the pairing must not be declared before the runtime honours it.
Clause-②: yes (widening) expected · Metadata-protocol: to be measured (a federation draft that carried a // REVIEW: marker no longer does).
Activity
objectstack-fleet commentedon Nov 8, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: none
Triage:
domain:spec· p3 ·pm:blockedon #31390 · 2026-11-08T16:58Z⛔ Not a claim.
domain:spec. Moving a type-compat pair fromlossyto compatible widens an acceptance set, which is a metadata-protocol change under [Decision] Keepdomain:devxanddomain:serveras separate lanes, or merge them? Measured load and boundary friction, 2026-11-08 #31424 rule 2. The self-split from driver-sql (oracle): bind uuid text to RAW(16) through HEXTORAW, so a lookup over a SYS_GUID key filters and writes by its uuid string, and type-compat can lift the pairing from lossy to compatible #31390 is right.lossyuntil the driver binds uuid.Generated by Claude Code
objectstack-fleet commentedon Nov 9, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: none
Triage: unblocked ·
pm:blocked→pm:queue· 2026-11-09T02:13Z⛔ Not a claim.
Blocked-by:target is done: driver-sql (oracle): bind uuid text to RAW(16) through HEXTORAW, so a lookup over a SYS_GUID key filters and writes by its uuid string, and type-compat can lift the pairing from lossy to compatible #31390 (closed).Generated by Claude Code
objectstack-fleet commentedon Nov 9, 2026
domain:spec#2 · zhuangjianguo · session_01QsR2jUe9c6YdtNkrzUnMUd
Claim: PM loop round 1 · 2026-11-09T02:54Z
Session:
session_01QsR2jUe9c6YdtNkrzUnMUd(seat); dev = card threadsession_01QsR2jULf9M1RFFq4dq9ioHAccount:
zhuangjianguoBranch:
claude/project-thread-h15c0w(a project card thread can push only its assigned branch, a measured platform exception toclaude/issue-<n>-<slug>)Worktree:
objectstack-issue-31592Domain:
domain:specSeat:
domain:spec#2File surface:
packages/spec/src/data/type-compat.ts+type-compat.test.ts; prose this card makes false:content/docs/data-modeling/external-datasources.mdx(theRAW(16)row L558 and the "not supported" bullet L576-580),docs/adr/0015-external-datasource-federation.md§4.6 L431-434 plus a datedAmended byentry;service-datasourcetest pins only where they hold today's'lossy'answer; a changeset (stop on breach; explain in the report)Container & model:
S,mode:thread,model: default tier — dispatch-gates --tier on origin/main 43fb4822 says MANDATORY claude-fable-5-1 (the docs/adr/** hit, no one-line exit); reason: the maintainer's standing rule 「后续不再派 fable 卡」 (2026-10-27) puts builds on the default tier; the review is owed at CONTRACT_REVIEW_TIER on an isolated sub-agent (rule 2026-11-08 16:12Z 「有 fable 车道就可以契约复审啊」),effort: xhighClause-②: yes (widening)
Metadata-protocol: semantic
Responsibility: n/a — not a defect card
Thread-read: 6730608741
Premise: holds (origin/main 43fb482:
type-compat.tsL454oracle: { uuid: { demoteExact: true } }, test L393-402 pins'lossy'; #31390's bind landed as PR #31642, 836d408,HEXTORAW(?)for comparands, writes and by-id keys)Serial constraints cleared: none (no open PR touches
type-compat.ts,external-datasources.mdxor ADR-0015; the seat's other card #31669 does not touch them).Tier H (
docs/adr/**). Contract review owed at CONTRACT_REVIEW_TIER. ⛔ Nodriver-sqlcode change.objectstack-fleet commentedon Nov 9, 2026
domain:spec#2 · zhuangjianguo · session_01QsR2jULf9M1RFFq4dq9ioH
os-dev-report
{ "issue": 31592, "status": "done", "branch": "claude/project-thread-h15c0w", "pr": "https://github.com/objectstack-ai/objectstack/pull/31786", "head": "e090be93", "session": "session_01QsR2jULf9M1RFFq4dq9ioH", "premise_still_valid": true, "summary": "Narrowed, not dropped: DialectLossyRule.demoteExact is now the list of exact pairings it demotes, and oracle/uuid names only 'text'. isCompatible('RAW(16)', lookup|master_detail|user, 'oracle') now answers true (was 'lossy'); text stays 'lossy' because #31390 binds comparison comparands, written values and by-id keys through HEXTORAW but pattern operators ($contains/$icontains/$startsWith/$like) still compile as a plain LIKE against the RAW, which Oracle converts to upper-case undashed hex (pinned by the #31390 control test). JSDoc, external-datasources.mdx (RAW(16) row + the 'not supported' bullet rewritten with the #31390 refusals + a pattern-operator bullet), ADR-0015 §4.6 and its Amended-by line, and a minor @objectstack/spec changeset.", "premise_readings": { "type_compat_main_43fb4822": "held: JSDoc L430-436, rule L454, test L393-402 (re-read at 9e1f1f90, unchanged)", "bind_coverage": "filter comparands (=,$eq,$ne,$in,$nin,ranges) bound; writes bound; by-id keys bound; lookup expand = { id: { $in } } (objectql engine-summaries-and-expansion.ts L593) = comparand path, bound. NOT bound: pattern operators (sql-driver-oracle-dialect.test.ts 'control: a pattern operator ... unjudged').", "demoteExact_member": "kept, narrowed to a FieldType list; oracle still uses it for text, so it is not dead. mssql/text addLossy unchanged.", "consumers": "service-datasource L957 draft REVIEW: unchanged (suggested type for RAW(16) is text, still lossy). L1113 lookup gate: unchanged for oracle (only OData navigationProperties reach it). L1339 validateObject: lookup/master_detail/user over oracle RAW(16) lose the type_mismatch warning; text keeps it. Pins holding 'lossy' in service-datasource: 0 (git grep), suite green unchanged." }, "tests": "spec type-compat.test.ts 42/42; spec full `pnpm --filter @objectstack/spec test` 824 files / 26081 passed + 1 todo, VERDICT command-exit 0; spec typecheck VERDICT 0; consumer `pnpm --filter @objectstack/service-datasource test` 74 files / 1299 passed, VERDICT 0. Ablation 1 (flag reading restored in isCompatible): red 'lookup: expected lossy to be true', 1 failed/41; ablation 2 (oracle row demoteExact: []): red 'expected true to be lossy', 1 failed/41; both via ablation-replace.mjs, landed and restored by blob hash (src-imported suite, no dist preflight).", "gates": "dispatch-gates --commands: 108 derived; 107 exit 0; check:dual-build-cjs-loads exit 3 PREREQUISITE NOT MET (full build) = NOT MEASURED, CI's. First-run exit 3 on check:doc-formula-expressions, check:doc-security-posture, check:skill-examples cleared by a locked turbo build of formula+lint+client-react, then exit 0. After merging main (f6474306) the five gates whose scripts moved re-ran exit 0 (ci-filter-parity, cross-package-test-inputs, pm-governed-merges, engine-double-contract, adr-anchors); spec build + check:generated exit 0 (14 artefacts up to date). --ran: 108 derived, 107 run, 1 NOT-MEASURED, 0 UNRUN. Body gates on the PR body: check-changeset-no-major exit 0 (LEVEL AXIS accepts yes (widening) + semantic with spec at minor), check-adr-0087-registration exit 0, check-closing-keyword-parity exit 0, check-partof-closing-keyword exit 0. PR check-runs: in_progress at report time.", "check_changeset_reading": "Metadata-protocol: semantic accepted with '@objectstack/spec': minor (LEVEL AXIS ✓). Not tried at patch.", "dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm", "line_budget": "94 changed lines (+67/-27) over 5 files, under 3000", "files_changed": [".changeset/31592-oracle-raw16-reference-exact.md", "content/docs/data-modeling/external-datasources.mdx", "docs/adr/0015-external-datasource-federation.md", "packages/spec/src/data/type-compat.test.ts", "packages/spec/src/data/type-compat.ts"], "tier": "Tier H (docs/adr/**); built at the default tier per the standing no-Fable rule; the CONTRACT_REVIEW_TIER review is the seat's", "deviations": [ "Narrowed instead of lifting every uuid exact pairing: text stays lossy on oracle RAW(16) (pattern operators unbound). The brief allows this ('lift only what is honoured and say which'); the card ruling allows 'or narrow it'.", "ADR-0015: besides L431-434 and Amended-by, the §4.6 DIALECT_LOSSY_RULES sentence ('oracle demotes the exact pairings of uuid') was corrected to 'the exact text pairing', because this change makes it false. Same file, same section.", "spec test:repo not run locally (CI Test Core).", "Live Oracle NOT MEASURED (no OS_TEST_ORACLE_URL)." ], "mcp_calls": "0 — no MCP GitHub calls", "api_writes": "2 — relay pr_create (#31786, run 42558565520, body read back identical 11459 bytes), this post-stamped comment on #31592; plus 2 git pushes", "open_questions": [], "security_findings": "none", "out_of_scope_findings": [ "carrier: none · noted, not filed — base-row observation: CANONICAL_TO_FIELD.uuid keeps 'user' exact on every dialect, yet a platform user id (better-auth 32-char string / 16-char nanoid) is never a uuid, so any uuid column refuses a real platform user id on write. Same judgement on every dialect, not this card's." ] }Generated by Claude Code
objectstack-fleet commentedon Nov 9, 2026
domain:spec#2 · zhuangjianguo · session_01QsR2jUe9c6YdtNkrzUnMUd
ACCEPT → PR #31786 comment 6732302802 (head
e090be93463b7273a813ee3cdfcb2c7863d6619b; contract review PASS 6732289043). Tier H: waits underpm:awaiting-mergefor aGOVERNED_APPROVERSapproval, reviewers requested.objectstack-fleet commentedon Nov 9, 2026
domain:spec#2 · zhuangjianguo · session_01QsR2jUe9c6YdtNkrzUnMUd
Landing note · 2026-11-09T05:56Z
os-zhuang(merge commit75d7fb11d7) through the merge queue, heade090be93as reviewed (contract review PASS 6732289043, ACCEPT 6732302802).Fixesclosed this card;pm:dispatchedremoved.Blocked-by: #31592. Atextfield over an oracleRAW(16)column stays'lossy'until pattern operators bind throughHEXTORAW, which is adriver-sqlquestion and not this lane's.