TML-2787: M:N slice 3 — nested writes through the junction by tensordreams · Pull Request #683 · prisma/prisma-next

tensordreams · 2026-06-01T20:50:33Z

Slice 3 (final) of the SQL ORM: Many-to-Many End to End project (Linear project). Nested connect/disconnect/create through the junction + the required-payload safety rail.

Stacked PR. Base tml-2786 (#680) → tml-2785 (#679) → tml-2784 (#678) → tml-2597 (#673) → tml-2729 (#667) → main. Review/merge bottom-up.

Overview

db.orm.User.update/create({ tags: (t) => t.connect/disconnect/create(...) }) now routes to the UserTag junction (INSERT / DELETE / target-insert+link), under both create() and update(). The N:M not supported yet guard is gone. Junctions with a required non-FK payload column cant be written through the sugar, so create and connect on them are disabled with a clear error (disconnect stays).

Changes (5 commits)

74a778816 — runtime junction write path: partitionByOwnership gains a junctionOwned bucket (keyed on through presence); connect→INSERT, disconnect→DELETE, create→target-insert+link, both flows, composite-key AND-ed; the rejection unit test flipped positive. (getRelationDefinitions now carries through.)
926bdc849 — required-payload fixture: User ↔ Role via UserRole(user_id, role_id, level NOT NULL) (canonical CLI emit).
3bccd80b3 — runtime guard: nested create on a required-payload junction throws.
e6c641811 — design correction (decision Remove SQL -> Runtime dependency #9): extended the guard to connect too (connect also INSERTs a junction row it cant complete → DB NOT-NULL violation), flipped the unit test, finished the 10 write integration tests.

Integration tests (per the project standard)

mn-nested-write.test.ts — 10 tests, no skips: connect/create on the pure User.tags junction with whole-row readback via include(tags), both create()+update() flows; disconnect; connect AND create on User.roles throw the guard; disconnect on User.roles works. Whole-row toEqual, explicit .select in most, ≥1 implicit/default selection.

AC status

✅ AC-1 — connect/disconnect/create route to junction DML, both flows, guard removed, rejection test flipped (unit + integration).
✅ AC-3 — write integration tests per the standard.
⚠ AC-2 — partial. Runtime disable of create+connect on required-payload junctions: ✅ done + tested. Type-level disable: deferred — see below.

⚠ Open decision (blocks marking this slice done)

The type-level .create disable cant be built as specified: the generated contract.d.ts relation type carries only to/cardinality/on, not through — so a conditional type cant see which model is the junction or that it has a required column. requiredPayloadColumns exists only at runtime. Honouring the type-level disable needs the contract .d.ts type emitter to carry through (a contract-surface change reaching into slice-0 territory). Options (full detail in wip/unattended-decisions.md #8):

(a) extend the .d.ts emitter to emit through, then a follow-up adds the conditional-type disable + negative type test;
(b) emit a narrower requiredPayloadColumns/hasRequiredPayload marker into the typed relation;
(c) accept runtime-only (contradicts the in-slice non-negotiable).

I shipped the runtime guard and left this for you. Once you pick (a)/(b), a small follow-up dispatch completes the type-level disable.

Notes

connect-disabled-on-required-payload was a mid-flight spec correction (the original spec wrongly assumed connect was FK-pair-only safe). The reverse Tag.users/Role.users directions are deferred (one-directional fixture, decision #3). Refs: TML-2787.

Summary by CodeRabbit

Release Notes

New Features
- Enabled nested create, connect, and disconnect operations for many-to-many relationships via junction tables.
- Added compile-time type checking to prevent invalid operations on junction relations with required fields.
Improvements
- Enhanced runtime validation with clear error guidance for missing required junction payload fields.

coderabbitai · 2026-06-01T20:50:41Z

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'TML-2787: M:N slice 3 — nested writes through the junction' accurately and specifically describes the main change: implementing many-to-many nested write operations that route through junction tables.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch tml-2787-slice-3-write

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

github-actions · 2026-06-01T20:52:59Z

size-limit report 📦

Path	Size
postgres / no-emit	156.34 KB (+0.79% 🔺)
postgres / emit	123.64 KB (+1.02% 🔺)
mongo / no-emit	76.92 KB (0%)
mongo / emit	71.01 KB (0%)
cf-worker / no-emit	183.82 KB (0%)
cf-worker / emit	147.7 KB (0%)

pkg-pr-new · 2026-06-01T20:53:27Z

Open in StackBlitz

@prisma-next/extension-author-tools

npm i https://pkg.pr.new/@prisma-next/extension-author-tools@683

@prisma-next/mongo-runtime

npm i https://pkg.pr.new/@prisma-next/mongo-runtime@683

@prisma-next/family-mongo

npm i https://pkg.pr.new/@prisma-next/family-mongo@683

@prisma-next/sql-runtime

npm i https://pkg.pr.new/@prisma-next/sql-runtime@683

@prisma-next/family-sql

npm i https://pkg.pr.new/@prisma-next/family-sql@683

@prisma-next/extension-arktype-json

npm i https://pkg.pr.new/@prisma-next/extension-arktype-json@683

@prisma-next/middleware-cache

npm i https://pkg.pr.new/@prisma-next/middleware-cache@683

@prisma-next/mongo

npm i https://pkg.pr.new/@prisma-next/mongo@683

@prisma-next/extension-paradedb

npm i https://pkg.pr.new/@prisma-next/extension-paradedb@683

@prisma-next/extension-pgvector

npm i https://pkg.pr.new/@prisma-next/extension-pgvector@683

@prisma-next/extension-postgis

npm i https://pkg.pr.new/@prisma-next/extension-postgis@683

@prisma-next/postgres

npm i https://pkg.pr.new/@prisma-next/postgres@683

@prisma-next/sql-orm-client

npm i https://pkg.pr.new/@prisma-next/sql-orm-client@683

@prisma-next/sqlite

npm i https://pkg.pr.new/@prisma-next/sqlite@683

@prisma-next/extension-supabase

npm i https://pkg.pr.new/@prisma-next/extension-supabase@683

@prisma-next/target-mongo

npm i https://pkg.pr.new/@prisma-next/target-mongo@683

@prisma-next/adapter-mongo

npm i https://pkg.pr.new/@prisma-next/adapter-mongo@683

@prisma-next/driver-mongo

npm i https://pkg.pr.new/@prisma-next/driver-mongo@683

@prisma-next/contract

npm i https://pkg.pr.new/@prisma-next/contract@683

@prisma-next/utils

npm i https://pkg.pr.new/@prisma-next/utils@683

@prisma-next/config

npm i https://pkg.pr.new/@prisma-next/config@683

@prisma-next/errors

npm i https://pkg.pr.new/@prisma-next/errors@683

@prisma-next/framework-components

npm i https://pkg.pr.new/@prisma-next/framework-components@683

@prisma-next/operations

npm i https://pkg.pr.new/@prisma-next/operations@683

@prisma-next/ts-render

npm i https://pkg.pr.new/@prisma-next/ts-render@683

@prisma-next/contract-authoring

npm i https://pkg.pr.new/@prisma-next/contract-authoring@683

@prisma-next/ids

npm i https://pkg.pr.new/@prisma-next/ids@683

@prisma-next/psl-parser

npm i https://pkg.pr.new/@prisma-next/psl-parser@683

@prisma-next/psl-printer

npm i https://pkg.pr.new/@prisma-next/psl-printer@683

@prisma-next/cli

npm i https://pkg.pr.new/@prisma-next/cli@683

@prisma-next/cli-telemetry

npm i https://pkg.pr.new/@prisma-next/cli-telemetry@683

@prisma-next/emitter

npm i https://pkg.pr.new/@prisma-next/emitter@683

@prisma-next/migration-tools

npm i https://pkg.pr.new/@prisma-next/migration-tools@683

prisma-next

npm i https://pkg.pr.new/prisma-next@683

@prisma-next/vite-plugin-contract-emit

npm i https://pkg.pr.new/@prisma-next/vite-plugin-contract-emit@683

@prisma-next/mongo-codec

npm i https://pkg.pr.new/@prisma-next/mongo-codec@683

@prisma-next/mongo-contract

npm i https://pkg.pr.new/@prisma-next/mongo-contract@683

@prisma-next/mongo-value

npm i https://pkg.pr.new/@prisma-next/mongo-value@683

@prisma-next/mongo-contract-psl

npm i https://pkg.pr.new/@prisma-next/mongo-contract-psl@683

@prisma-next/mongo-contract-ts

npm i https://pkg.pr.new/@prisma-next/mongo-contract-ts@683

@prisma-next/mongo-emitter

npm i https://pkg.pr.new/@prisma-next/mongo-emitter@683

@prisma-next/mongo-schema-ir

npm i https://pkg.pr.new/@prisma-next/mongo-schema-ir@683

@prisma-next/mongo-query-ast

npm i https://pkg.pr.new/@prisma-next/mongo-query-ast@683

@prisma-next/mongo-orm

npm i https://pkg.pr.new/@prisma-next/mongo-orm@683

@prisma-next/mongo-query-builder

npm i https://pkg.pr.new/@prisma-next/mongo-query-builder@683

@prisma-next/mongo-lowering

npm i https://pkg.pr.new/@prisma-next/mongo-lowering@683

@prisma-next/mongo-wire

npm i https://pkg.pr.new/@prisma-next/mongo-wire@683

@prisma-next/sql-contract

npm i https://pkg.pr.new/@prisma-next/sql-contract@683

@prisma-next/sql-errors

npm i https://pkg.pr.new/@prisma-next/sql-errors@683

@prisma-next/sql-operations

npm i https://pkg.pr.new/@prisma-next/sql-operations@683

@prisma-next/sql-schema-ir

npm i https://pkg.pr.new/@prisma-next/sql-schema-ir@683

@prisma-next/sql-contract-psl

npm i https://pkg.pr.new/@prisma-next/sql-contract-psl@683

@prisma-next/sql-contract-ts

npm i https://pkg.pr.new/@prisma-next/sql-contract-ts@683

@prisma-next/sql-contract-emitter

npm i https://pkg.pr.new/@prisma-next/sql-contract-emitter@683

@prisma-next/sql-lane-query-builder

npm i https://pkg.pr.new/@prisma-next/sql-lane-query-builder@683

@prisma-next/sql-relational-core

npm i https://pkg.pr.new/@prisma-next/sql-relational-core@683

@prisma-next/sql-builder

npm i https://pkg.pr.new/@prisma-next/sql-builder@683

@prisma-next/target-postgres

npm i https://pkg.pr.new/@prisma-next/target-postgres@683

@prisma-next/target-sqlite

npm i https://pkg.pr.new/@prisma-next/target-sqlite@683

@prisma-next/adapter-postgres

npm i https://pkg.pr.new/@prisma-next/adapter-postgres@683

@prisma-next/adapter-sqlite

npm i https://pkg.pr.new/@prisma-next/adapter-sqlite@683

@prisma-next/driver-postgres

npm i https://pkg.pr.new/@prisma-next/driver-postgres@683

@prisma-next/driver-sqlite

npm i https://pkg.pr.new/@prisma-next/driver-sqlite@683

commit: 1f24648

tensordreams · 2026-06-02T13:08:49Z

Updated: the type-level .create/.connect disable is now implemented (escalation #8 resolved via the chosen approach — the .d.ts emitter now carries through, landed in slice 0; the type gate derives required-payload purely at the type level, no any). AC-2 is closed at both the runtime and type levels. The whole stack has also been rebased onto the latest origin/main (clean — no fixture re-emit cascade). Negative type test: junction-link-write-disable.test-d.ts. No longer a draft.

tensordreams · 2026-06-12T17:37:21Z

On it 👍 — re: the review-body finding "Duplicate-connect error message can be misleading": softening the connect unique-violation wrap so it no longer asserts the junction link is already present (a non-PK unique constraint on the junction could be the real cause); it will say a unique constraint on the junction was violated and that the link may already be present. Unit + integration message assertions updated accordingly.

tensordreams · 2026-06-12T17:40:11Z

Done in c274f1e95 — re: the review-body finding "Duplicate-connect error message can be misleading": the connect unique-violation wrap now says the mutation 'violated a unique constraint on junction "

"; the junction link may already be present' instead of asserting the link already exists. Unit assertions (duplicate-key wrap + SQLITE_CONSTRAINT_PRIMARYKEY wrap) and the integration already-linked-tag test updated to the new wording.

tensordreams · 2026-06-12T17:47:14Z

On it 👍 — re: the decision-log note: recording duplicate-connect-errors as an intentional divergence from Prisma-classic implicit-M:N connect (which is idempotent via ON CONFLICT DO NOTHING) in the slice 03 spec's decision surface.

tensordreams · 2026-06-12T17:48:07Z

Done in 07379d4f8 — re: the decision-log note: the slice 03 spec (projects/sql-orm-many-to-many/slices/03-nested-write-through-junction/spec.md) now records duplicate-connect-errors as an intentional divergence from Prisma-classic implicit-M:N connect (idempotent ON CONFLICT DO NOTHING): a note on the connect design bullet plus an edge-case row with the rationale (portable write path, surfaces caller bugs) and the revisit condition. The wording reflects the softened error message from A25c (c274f1e95).

Write slice: connect/disconnect/create through the junction + required-payload .create disable (types+runtime). 4 dispatches (write path / required-payload fixture / type+runtime disable / integration). Type-disable feasibility risk pre-named with a halt. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

Replace the partitionByOwnership N:M rejection guard with a junctionOwned bucket. connect/disconnect/create over a through relation now resolve to junction INSERT/DELETE (create = target-insert then link) in both the create() and update() graph flows, after the parent PK is known. Composite keys are AND-ed across all parent/child column pairs; disconnect stays update()-only. Flip the rejection unit test to a positive junction-DML assertion and add connect/disconnect/create coverage for both flows. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…umns exist Throw a clear runtime error when a nested `.create()` targets an M:N relation whose junction table has non-FK NOT-NULL columns with no default (i.e. `requiredPayloadColumns` is non-empty). The error names the relation, the offending column(s), and points to the junction model / SQL builder as the supported alternative. `connect` and `disconnect` are unaffected — they only touch the FK pair. Pure junctions (no required payload) pass through the create path as before. Adds `requiredPayloadColumns` to the local `JunctionThrough` interface and copies it from the already-resolved `relation.through` in `getRelationDefinitions`. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…finish write integration tests Both `connect` and `create` write a bare junction row the M:N sugar can't complete when the junction has required non-FK columns (e.g. `user_roles.level NOT NULL`). Extend the runtime guard in `applyJunctionOwnedMutation` to cover `connect` in addition to `create`; `disconnect` (DELETE path) is unaffected. Guard message is unified: "Cannot `<op>` on relation `<rel>`: its junction `<table>` has required column(s) `<col>` the relation API can't populate. Use the `<Model>` model directly or the SQL builder." Unit test that previously asserted connect-on-User.roles succeeds is flipped to assert rejection. Integration tests are fully enabled (no it.skip); the new test asserts connect on User.roles throws the guard, while pure-junction (User.tags) connect/create/disconnect paths continue to work end-to-end. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…-2787) Write slice artifacts. Spec corrected mid-flight (decision #9: connect also disabled on required-payload junctions). Type-level .create disable deferred (decision #8 — needs a contract .d.ts emitter change, operator decision). Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…sted-write test; re-emit fixtures with through blocks Plain `string` constants failed the integration typecheck (tsc) because Tag.id / Role.id are `Char<36>`. Brand TAG_RUST, TAG_TS, ROLE_ADMIN, ROLE_EDITOR at declaration (matching the pattern from create.test.ts). Runtime values are unchanged. Also commits the re-emitted contract.d.ts fixtures carrying the new `through` blocks on the N:M relations (User.tags / User.roles). Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…elations at the type level Nested `.create` on an N:M relation whose junction carries a required non-FK payload column now resolves its argument to `never`, surfacing a compile-time error instead of only the runtime guard in `mutation-executor.ts`. `connect`/`disconnect` typing is unchanged. The derivation resolves the relation `through` to its junction model, treats any non-join column as a payload column, and disables create when any payload column is required (not nullable, no default). `User.roles` (junction `user_roles` with required `level`) disables create; `User.tags` (pure junction) keeps it. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

Renames the `CreateDisabled` flag on `RelationMutator` to `LinkWritesDisabled` and applies the same `never`-arg guard to both `create` and `connect` overloads. `disconnect` is unaffected (it issues a DELETE, not an INSERT, so no payload column is needed). `HasRequiredJunctionPayload` is already threaded as the flag via `RelationMutationCallback` — no change required there. The runtime guard in `mutation-executor.ts` already blocks both operations; the type system now matches. Test file renamed to `junction-link-write-disable.test-d.ts`; the former "connect remains available" positive test is replaced with a `@ts-expect-error` assertion, and a new pure-positive disconnect test preserves coverage of the allowed path. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…ard tests The type-level .create/.connect disable (now in dist after the main rebase) makes those calls compile errors; cast the args to `never` to still exercise the runtime guard on a real DB (defense-in-depth). Surfaced by rebuilding against origin/main. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…fter TML-2841 rebase The write path gained namespace coordinates: insertSingleRow, findRowByCriterion, toFieldName, compileInsertCount, compileDeleteCount all take a leading namespaceId. Add namespaceId to the local JunctionThrough, thread parentNamespaceId into applyJunctionOwnedMutation / readJunctionParentValues, and use relation.relatedNamespaceId / through.namespaceId for the related-row and junction-DML calls. Add the namespaceId arg to the M:N write tests. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

updateFirstGraph now mirrors createGraph and runs preflightJunctionOwnedCreateMutation over the junction-owned bucket before the scalar-update block. Without it, an update()-flow junction rejection (required-payload guard, metadata shape, unresolved connect target) landed after the scalar UPDATE had already persisted; ordering is the only partial-write protection on runtimes without transaction(). Pinned by a unit test asserting no UPDATE executes when the junction guard fires, and an update()-flow integration test asserting the scalar update does not survive a junction rejection. Addresses PR 683 review thread r3404036584. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…raph The junction-owned create bucket was the only one that did not recurse: parent-owned and child-owned nested creates route their payloads through createGraph, but the junction bucket called insertSingleRow directly, so a junction-created target carrying its own relation mutation (type-legal per RelationMutationCreate.data) passed through mapModelDataToStorageRow untouched and died in normalizeInsertRows with an unknown-column error. Swapping insertSingleRow for createGraph is drop-in (same returned row shape readJunctionTargetValues needs) and buys bucket parity. Pinned by a unit test where a junction-created Child carries an N:1 owner connect that now resolves into the child INSERT. Addresses PR 683 review thread r3404036572. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…flight Duplicate criteria in a single connect([...]) (or two criteria resolving to the same row) defeated the preflight: each criterion resolved independently and passed, the parent INSERT landed, the first link INSERTed, and the second threw the duplicate-link error — leaving the parent plus the first link behind, the exact partial-write shape the preflight exists to prevent. The preflight now keys each resolved target tuple and fails fast on a duplicate before any write, consistent with duplicate connect being a domain error in this slice. Pinned by a unit test asserting the rejection lands with zero INSERTs executed. Addresses PR 683 review thread r3404036592. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

Adds the create()-flow ordering edge case from review: a connect preflight that resolves its target successfully, followed by a parent INSERT failure (duplicate users pkey), must leave no junction rows and no parent mutation behind. Ordering is the only partial-write protection on the integration runtimes, so this pins the guarantee the rest of the suite leans on. Addresses the ordering finding in PR 683 review 4486308660. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…nctionLink Both the type gate and the runtime guard treat execution-defaulted junction payload columns as satisfied, but the junction INSERT built the row from the FK pairs only — the exempted case passed both gates and then hit a NOT NULL violation on the database. Mirror insertSingleRow: merge context.applyMutationDefaults({ op: create, table: through.table }) into the junction row before compileInsertCount. The reworked unit test builds its ExecutionContext from the patched contract (the defaults applier is a closure over the contract the context was created from, so spreading a patched contract over an existing context never sees the new default) and asserts the junction row includes the defaulted column. A new integration test drops the DB-side default on user_tags.created_at and covers an execution-defaulted junction payload column end-to-end. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…ons only The constraint failed message arm matched every SQLite constraint class — NOT NULL, FOREIGN KEY, and CHECK failures during a junction connect INSERT were rewritten into the duplicate-link error, pointing users at a duplicate that is not the problem. The arm was also redundant for genuine unique violations: SQLite says UNIQUE constraint failed, which the unique constraint arm already matches. Drop the over-broad arm and recognize SQLITE_CONSTRAINT_PRIMARYKEY next to SQLITE_CONSTRAINT_UNIQUE (better-sqlite3 reports junction-PK duplicates — the common M:N shape — with the PRIMARYKEY code). Unit tests pin both boundaries: a NOT NULL constraint failure passes through unwrapped, and the PRIMARYKEY code is recognized on its own with a message no string arm matches. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…onnect The shared-column conflict guard existed on the INSERT side only: insertJunctionLink merges parent/target values through writeJunctionColumn and throws on mismatch, while deleteJunctionLink emitted one predicate per side — a mismatched shared column produced contradictory predicates (tenant_id = 7 AND tenant_id = 8), so the DELETE matched nothing and disconnect silently no-oped where connect throws. deleteJunctionLink now builds the merged row via writeJunctionColumn and derives the WHERE from it: the mismatched case throws the same conflict error as connect (before any DELETE executes), and the values-equal case drops the redundant duplicate predicate. Both pinned by unit tests. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

The connect unique-violation wrap asserted "the junction link is already present", but when the junction table carries unique constraints beyond its PK the underlying violation may not be about the link being connected at all — the message attributed every unique violation to a duplicate link. Reword to state what is actually established: a unique constraint on the junction was violated, and the link may already be present. Unit and integration message assertions updated to the new wording. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…e gate DisconnectMutator reshaped disconnect for every relation, but nothing pinned that the bare disconnect() overload stays accepted for plain 1:N relations — if HasJunctionThrough ever mis-resolved toward true for non-junction relations, 1:N disconnect-all would silently drop out of the typed surface. A positive update-input test on User.posts pins it. HasRequiredJunctionPayload was only exercised on the required-to-disabled side. Three new positive tests pin the optional-payload arms of IsOptionalCreateField in isolation: a junction whose only payload column is nullable (user_tags.note), one whose only payload column carries a storage default (user_tags.created_at), and one whose required payload column is satisfied solely by an execution onCreate default (user_roles.level — the type-level counterpart of the insertJunctionLink defaults fix). Verified non-vacuous by temporarily removing the storage-default arm from IsOptionalCreateField, which turns the storage-default pin red. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

…ntentional divergence Duplicate M:N connect deliberately errors with a wrapped unique-violation domain error where Prisma-classic implicit-M:N connect is idempotent (ON CONFLICT DO NOTHING). Per round-3 review, record the divergence in the slice 03 decision surface: a note on the connect design bullet plus an edge-case row carrying the rationale (portable write path, surface caller bugs) and the revisit condition. Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

tensordreams force-pushed the tml-2786-slice-2-filter branch from 92d1741 to 25a3c7e Compare June 2, 2026 12:40

tensordreams force-pushed the tml-2787-slice-3-write branch from 7cbb4a0 to de78282 Compare June 2, 2026 12:40

tensordreams force-pushed the tml-2786-slice-2-filter branch from 25a3c7e to 06fc270 Compare June 2, 2026 13:07

tensordreams force-pushed the tml-2787-slice-3-write branch from de78282 to 3e1c908 Compare June 2, 2026 13:07

tensordreams marked this pull request as ready for review June 2, 2026 13:08

tensordreams requested a review from a team as a code owner June 2, 2026 13:08

tensordreams changed the title ~~TML-2787: M:N slice 3 — nested writes through the junction (DRAFT — type-level disable pending decision)~~ TML-2787: M:N slice 3 — nested writes through the junction Jun 2, 2026

tensordreams mentioned this pull request Jun 2, 2026

TML-2790: M:N SQLite demo examples + project plan expansion (slices 4–6) #697

Merged

tensordreams force-pushed the tml-2786-slice-2-filter branch from 06fc270 to bc2d4b9 Compare June 3, 2026 08:51

tensordreams force-pushed the tml-2787-slice-3-write branch 2 times, most recently from 308b48d to bb3e246 Compare June 3, 2026 11:32

tensordreams force-pushed the tml-2786-slice-2-filter branch 2 times, most recently from 3982e13 to b1dadcf Compare June 4, 2026 15:12

tensordreams force-pushed the tml-2787-slice-3-write branch from bb3e246 to 47cf59e Compare June 4, 2026 15:12

tensordreams force-pushed the tml-2786-slice-2-filter branch from b1dadcf to df900cd Compare June 4, 2026 15:41

tensordreams force-pushed the tml-2787-slice-3-write branch from 47cf59e to b04a59c Compare June 4, 2026 15:41

tensordreams force-pushed the tml-2786-slice-2-filter branch from df900cd to ee44053 Compare June 5, 2026 12:44

tensordreams force-pushed the tml-2787-slice-3-write branch from b04a59c to 04522c0 Compare June 5, 2026 12:44

tensordreams force-pushed the tml-2786-slice-2-filter branch from ee44053 to 65d4fe3 Compare June 5, 2026 14:21

tensordreams force-pushed the tml-2787-slice-3-write branch from 04522c0 to 4d8dad2 Compare June 5, 2026 14:21

tensordreams force-pushed the tml-2786-slice-2-filter branch from 65d4fe3 to bd47c00 Compare June 5, 2026 15:02

tensordreams force-pushed the tml-2787-slice-3-write branch from 4d8dad2 to 16dc690 Compare June 5, 2026 15:02

tensordreams mentioned this pull request Jun 5, 2026

TML-2790: M:N SQLite demo examples + project plan expansion (slices 4–6) #742

Open

tensordreams force-pushed the tml-2787-slice-3-write branch from 16dc690 to f3a7a5c Compare June 8, 2026 10:19

tensordreams force-pushed the tml-2786-slice-2-filter branch from bd47c00 to 40b8cbe Compare June 8, 2026 10:19

tensordreams force-pushed the tml-2787-slice-3-write branch from f3a7a5c to b772b54 Compare June 8, 2026 11:19

tensordreams added 23 commits June 12, 2026 19:51

docs(upgrade): note the M:N nested-write path in 0.13→0.14 instructions

1de6e60

Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

fix(sql-orm-client): harden junction nested writes

cd21f38

Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

fix(sql-orm-client): refine junction relation write gates

7a684e5

Signed-off-by: Alexey Orlenko's AI Agent <robot@aqrln.net>

tensordreams force-pushed the tml-2787-slice-3-write branch from 3440e5d to 1f24648 Compare June 12, 2026 17:56

This was referenced Jun 12, 2026

TML-2794: M:N slice 5 — PSL authors many-to-many (junction → N:M + through) #819

Open

TML-2795: M:N slice 6 — PG demo M:N examples + PSL-only contract source #820

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TML-2787: M:N slice 3 — nested writes through the junction#683

TML-2787: M:N slice 3 — nested writes through the junction#683
tensordreams wants to merge 23 commits into
mainfrom
tml-2787-slice-3-write

tensordreams commented Jun 1, 2026 •

edited by aqrln

Loading

Uh oh!

coderabbitai Bot commented Jun 1, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Jun 1, 2026 •

edited

Loading

Uh oh!

pkg-pr-new Bot commented Jun 1, 2026 •

edited

Loading

Uh oh!

tensordreams commented Jun 2, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

tensordreams commented Jun 1, 2026 • edited by aqrln Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Overview

Changes (5 commits)

Integration tests (per the project standard)

AC status

⚠ Open decision (blocks marking this slice done)

Notes

Summary by CodeRabbit

Release Notes

Uh oh!

coderabbitai Bot commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

size-limit report 📦

Uh oh!

pkg-pr-new Bot commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tensordreams commented Jun 2, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

tensordreams commented Jun 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

tensordreams commented Jun 1, 2026 •

edited by aqrln

Loading

coderabbitai Bot commented Jun 1, 2026 •

edited

Loading

github-actions Bot commented Jun 1, 2026 •

edited

Loading

pkg-pr-new Bot commented Jun 1, 2026 •

edited

Loading