omnigraph/docs/dev/writes.md
devin-ai-integration[bot] 2c578a60b2
(feat) convert engine call sites to &dyn TableStorage; demote legacy TableStore methods to pub(crate) (#86)
* MR-854: convert engine call sites to &dyn TableStorage; demote legacy methods

Phase 1b: every db.table_store.X(...) call site converts to
db.storage().X(...), reaching the storage layer through the sealed
TableStorage trait (returns &dyn TableStorage). Opaque SnapshotHandle
and StagedHandle replace bare lance::Dataset and Transaction in the
threaded values.

Phase 9: the inherent inline-commit methods on TableStore
(append_batch, merge_insert_batch{,es}, overwrite_batch,
create_btree_index, create_inverted_index) demote from pub to
pub(crate). Their only remaining direct users are table_store.rs
itself and the bulk loader's LoadMode::{Append, Overwrite, Merge}
concurrent fast-paths in loader::write_batch_to_dataset (no
two-phase shape in Lance 4.0.0 — closes after lance#6658 and #6666).

Docs:
- invariants.md \u00a7VI.23: drop "at the writer-trait surface"
  qualifier; staged primitives are now the only engine surface.
- runs.md: residual matrix shrinks to delete_where and
  create_vector_index (the two upstream-blocked residuals).
- forbidden_apis.rs: replace transitional language with the
  current allow-list shape (table_store.rs + loader concurrent
  fast-path only).

Files touched:
- changes/mod.rs, db/omnigraph.rs (+export/optimize/schema_apply/
  table_ops.rs), exec/{merge,mod,mutation,staging}.rs,
  loader/mod.rs, storage_layer.rs, table_store.rs,
  tests/forbidden_apis.rs, docs/{invariants,runs}.md.

Co-Authored-By: Ragnor Comerford <ragnor.comerford@gmail.com>

* MR-854: replace test-only inline-commit append callers with local Lance helpers

After demoting TableStore::append_batch from pub to pub(crate), the
integration tests in tests/recovery.rs and tests/staged_writes.rs
that previously called store.append_batch(...) directly to simulate
HEAD-ahead-of-manifest drift can no longer access the inherent
method. Replace those calls with small in-test helpers that do a raw
Dataset::append (the same body the inherent method runs).

- tests/helpers/mod.rs gains lance_append_inline (shared helper).
- tests/staged_writes.rs gets a file-local lance_append_inline_local
  (staged_writes.rs does not import helpers::).
- tests/recovery.rs drops the unused TableStore import in the one
  function whose store binding became unused after the conversion.

Co-Authored-By: Ragnor Comerford <ragnor.comerford@gmail.com>

* MR-854: retrigger CI for flaky Test Workspace job

Co-Authored-By: Ragnor Comerford <ragnor.comerford@gmail.com>

* MR-854: convert remaining table_store call sites in export.rs / read_blob

Two leftover `self.table_store.X` / `db.table_store.X` call sites were
missed in the initial sweep — flagged by Devin Review on PR #86. Both
now go through the trait surface:

- `entity_from_snapshot` (db/omnigraph/export.rs): switch from
  `db.table_store.open_snapshot_table` + `db.table_store.scan` to
  `db.storage().open_snapshot_at_table` + `db.storage().scan`.

- `read_blob` (db/omnigraph.rs): replace
  `snapshot.open(table_key)` + `self.table_store.first_row_id_for_filter`
  with `self.storage().open_snapshot_at_table` +
  `self.storage().first_row_id_for_filter`. The follow-up
  `take_blobs` call still needs an `Arc<Dataset>` (it's a Lance blob
  accessor not surfaced through the trait), so we hand off via
  `SnapshotHandle::into_arc()` with a comment.

After this commit, no engine code outside `table_store.rs` reaches the
inherent `TableStore` API — the docs/runs.md and docs/invariants.md
claim is now uniformly true.

Co-Authored-By: Ragnor Comerford <ragnor.comerford@gmail.com>

* MR-854: post-rebase doc fixes (Lance 6.0.1, MR-A framing, into_dataset note)

Reviewer feedback on the rebased PR:

* docs/dev/writes.md residuals matrix: drop demoted methods from the trait-surface table (now `pub(crate)`); keep only the two genuine trait-surface residuals (`delete_where`, `create_vector_index`); reframe under MR-A (Lance v7.x bump) per docs/dev/lance.md.

* tests/forbidden_apis.rs: update transitional allow-list header to (a) drop the truncate_table mislabel (truncate_table is a Lance Dataset method, not a TableStore method — overwrite_batch's internal call), (b) reframe trait-surface residuals under MR-A / Lance #6666.

* crates/omnigraph/src/storage_layer.rs::SnapshotHandle::{into_arc, into_dataset}: add single-ref invariant doc — both consume Arc via try_unwrap-or-clone; sibling SnapshotHandle clones across an await point force a deep Dataset clone.

* Replace lance-4.0.0 version refs with lance-6.0.1 in active source/test/dev-doc comments (storage_layer.rs, table_store.rs, table_ops.rs, schema_apply.rs, merge.rs, recovery.rs, staged_writes.rs, consistency.rs, docs/dev/execution.md, docs/user/query-language.md). Historical refs in docs/releases/v0.4.1.md and the canonical "Lance 4.0.0 → 6.0.1 migration" line in docs/dev/lance.md left intact.

No engine code changes.

* MR-854: update docs/dev/invariants.md Storage trait row + gap entry

Reviewer feedback: the docs reorg landed; the invariant row now lives in
docs/dev/invariants.md with stable headings (no more numbered §VI.23).

Update two pieces to reflect MR-854 completion:

* Status table 'Storage trait' row: was 'full call-site migration ... incomplete';
  now 'engine call sites all route through db.storage() (MR-854); inline-commit
  inherent methods are pub(crate)-demoted; capability/stat surfaces are roadmap'.

* 'Known Gaps' 'Storage abstraction' entry: was 'older inherent TableStore call
  sites and inline residuals remain'; now names the closed scope (MR-854 — call
  sites migrated, methods demoted, loader fast-paths) and the remaining
  trait-surface residuals under MR-A (Lance v7.x bump) and Lance #6666.

Cross-links to docs/dev/lance.md and docs/dev/writes.md so the framing stays
co-located with the canonical Lance surface tracking.

* MR-854: remove dead inline-commit methods from the storage surface

The loader concurrent fast-path (write_batch_to_dataset) is only reached
for LoadMode::Overwrite — Append/Merge route through MutationStaging — so
its Append/Merge arms were unreachable. Collapse it to overwrite-only and
drop the now-unused mode params, which removes the only callers of:

- TableStorage::append_batch + TableStorage::merge_insert_batches (trait)
- TableStore::merge_insert_batch + merge_insert_batches (inherent)

create_btree_index / create_inverted_index had zero callers anywhere
(scalar index builds use the stage_* primitives). Remove both from the
trait and the inherent impl.

Inherent append_batch stays pub(crate): overwrite_batch and recovery
tests use it. Migrate the one trait-append_batch test caller
(seed_person_row) to stage_append + commit_staged. The merge_insert
FirstSeen-workaround rationale moves from the deleted merge_insert_batch
into stage_merge_insert (now the sole merge path). No behavior change.

Also corrects the inaccurate loader residual comment (the prior text
blamed Lance #6658/#6666, which are the delete and vector-index issues,
for keeping overwrite inline; a stage_overwrite primitive already exists
and schema_apply uses it).

* MR-854: seal db.storage() to staged-only; move residuals to InlineCommitResidual

Split the three remaining inline-commit writes (overwrite_batch,
delete_where, create_vector_index) off the TableStorage trait onto a new
sealed InlineCommitResidual trait, reachable only via the explicit
Omnigraph::storage_inline_residual() accessor. db.storage() now exposes
only staged primitives + reads, so engine code cannot couple a write
with a Lance HEAD advance through the default surface — MR-793 acceptance
§1 ("no public method commits as a side effect of writing") now holds by
construction, not by review + naming.

Call sites moved to storage_inline_residual(): loader overwrite
fast-path, the three mutation delete_where paths, the branch-merge
delete, and the vector-index build. Impl bodies are unchanged (same
delegation to the pub(crate) inherent methods); this is a pure surface
reshape with no behavior change.

The residual trait holds two genuinely upstream-blocked methods
(delete_where -> Lance #6658/v7.x, create_vector_index -> Lance #6666)
plus overwrite_batch, kept for the loader's cross-table bulk-overwrite
concurrency until its staged migration lands (tracked follow-up).

* MR-854 docs: describe the staged-only seal; fix stale Lance index URLs

- writes.md / invariants.md / AGENTS.md: the inline-commit residuals now
  live on InlineCommitResidual behind db.storage_inline_residual(), so
  acceptance §1 holds by construction rather than 'option (b)' per-method
  enumeration. Drop the inaccurate 'until Lance exposes
  Operation::Overwrite { fragments }' claim (that op exists; stage_overwrite
  already builds it) and reframe overwrite_batch as a removable legacy
  residual gated on the loader's bulk-overwrite concurrency.
- forbidden_apis.rs: rewrite the allow-list doc for the split surface.
- lance.md: the index spec pages moved from /format/table/index/ to
  /format/index/ in Lance 6.x (the old paths 404). Fix all 13 URLs.

* MR-854: fix stale lance-4.0.0 comment refs flagged in review

Addresses greptile (exec/merge.rs) and aaltshuler's stale-version blocker:
update lance-4.0.0 -> 6.0.1 in the comment/doc refs within this PR's
footprint (exec/merge.rs, exec/mutation.rs, docs/dev/writes.md). Also
corrects exec/merge.rs to cite lance#6666 (not #6658) for
build_index_metadata_from_segments — that is the vector-index segment-commit
API; #6658 is the two-phase delete. (Pre-existing 4.0.0 refs in untouched
files like architecture.md/storage.md are main's incomplete migration
cleanup, left out of scope.)

* fix(storage): stage loader overwrites

* fix(storage): stage empty schema rewrites

---------

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Ragnor Comerford <ragnor.comerford@gmail.com>
Co-authored-by: Ragnor Comerford <hello@ragnor.co>
2026-06-09 23:03:08 +02:00

293 lines
17 KiB
Markdown

# Direct-Publish Write Path
> History: the Run state machine and `__run__<id>` staging branches were
> removed in MR-771 (shipped v0.4.0). Writes now go directly to the target
> table; this document specifies that direct-publish path.
`mutate_as` and `load` write **directly to the target table**
and call `ManifestBatchPublisher::publish` once at the end with
`expected_table_versions` (the per-table manifest versions captured before
the first write). Cross-table OCC is enforced inside the publisher; the
publisher's row-level CAS on `__manifest` is the single fence.
## What this means in practice
- No `RunRecord`, no `_graph_runs.lance`, no `_graph_run_actors.lance`.
- No `omnigraph run *` CLI subcommands and no `/runs/*` HTTP endpoints.
- No `__run__<id>` staging branches; `__run__*` is no longer a reserved
name. The branch-name guard was removed in MR-770, and any stale
`__run__*` branch on an upgraded graph is swept off `__manifest` by the
v2→v3 internal-schema migration on first read-write open. (The inert
`_graph_runs.lance` bytes remain until a `delete_prefix` primitive lands.)
- Cancelled mutation futures leave **no graph-level state** — only orphaned
Lance fragments, which the existing `omnigraph cleanup` pipe reclaims.
## Read-your-writes within a multi-statement mutation
A `.gq` query with multiple ops (e.g. `insert Person … insert Knows …`)
must observe earlier ops' writes when validating later ops (referential
integrity, edge cardinality). After MR-794 step 2+ this is implemented
via an in-memory `MutationStaging` accumulator in
[`crates/omnigraph/src/exec/staging.rs`](../../crates/omnigraph/src/exec/staging.rs),
shared by both `mutate_as` and the bulk loader:
- On the first touch of each table, the pre-write manifest version is
captured into `expected_versions[table_key]` (the publisher's CAS
fence at end-of-query).
- Each insert/update op pushes a `RecordBatch` into the per-table
pending accumulator. Lance HEAD does **not** advance during op
execution.
- Read sites (validation, predicate matching for `update`) consume
`TableStore::scan_with_pending`, which scans committed via Lance
and applies the same SQL filter to the pending batches via DataFusion
`MemTable`. Same-query writes are visible to subsequent reads.
- At end-of-query, `MutationStaging::finalize` issues exactly one
`stage_*` + `commit_staged` per touched table (concatenating
accumulated batches; merge-mode dedupes by `id`, last-write-wins),
and the publisher publishes the manifest atomically across all
touched sub-tables. Cross-table conflicts surface as
`ManifestConflictDetails::ExpectedVersionMismatch`.
- **Deletes still inline-commit.** Lance's `Dataset::delete` is not
exposed as a two-phase op in 6.0.1; deletes go through `delete_where`
immediately and record their post-write state in
`MutationStaging.inline_committed`. The parse-time D₂ rule (below)
prevents inserts/updates from coexisting with deletes in one query,
so the inline path is safe for delete-only mutations.
This upholds the manifest-atomic mutation and read-your-writes invariants
tracked in [docs/dev/invariants.md](invariants.md).
### D₂ — parse-time mixed-mode rejection
A single mutation query is either insert/update-only or delete-only.
Mixed → rejected at parse time with a clear error directing the user to
split the query. Reason: mixing creates ordering hazards
(insert→delete on the same row would silently no-op because the staged
insert isn't visible to delete; cascading deletes of just-inserted
edges break referential integrity). Until Lance exposes a two-phase
delete API, the parse-time rejection keeps both paths atomic and
correct. Tracked: MR-793, plus a Lance-upstream ticket.
### MR-793 status (storage trait two-phase invariant) — partial
MR-793 hoists the staged-write pattern into a `TableStorage` trait
surface with sealed-trait enforcement and opaque `SnapshotHandle` /
`StagedHandle` types — see `crates/omnigraph/src/storage_layer.rs`.
The trait is the canonical surface for new engine code; existing call
sites still use the inherent `TableStore` methods (mechanical migration
deferred to a follow-up cycle — tracked).
Three writers have been migrated onto staged primitives:
* **`ensure_indices`** (`db/omnigraph/table_ops.rs::build_indices_on_dataset_for_catalog`)
— scalar indices (BTree, Inverted) now use `stage_create_*_index` +
`commit_staged`. Vector indices stay inline (residual — Lance
`build_index_metadata_from_segments` is `pub(crate)` in 6.0.1;
companion ticket to lance-format/lance#6658 needed).
* **`branch_merge::publish_rewritten_merge_table`**
(`exec/merge.rs`) — merge_insert now uses `stage_merge_insert` +
`commit_staged`. Deletes stay inline (Lance #6658 residual).
* **`schema_apply` rewritten_tables** (`db/omnigraph/schema_apply.rs`)
— rewrites use `stage_overwrite` + `commit_staged`, including empty-table
rewrites via a zero-fragment Lance `Operation::Overwrite`.
A defense-in-depth integration test (`tests/forbidden_apis.rs`) walks
engine source and fails if non-allow-listed code calls Lance's
inline-commit APIs directly. The trait surface itself is the primary
enforcement (sealed + only-callable-via-trait once call sites land);
the grep test catches type-system bypass attempts.
The "finalize → publisher residual" described below applies equally to
the migrated writers — Lance has no multi-dataset atomic commit
primitive, so the per-table commit_staged → manifest publish gap is
the same drift class. Closing it requires either upstream Lance
multi-dataset commit OR the omnigraph-side recovery-on-open reconciler
described in `.context/mr-793-design.md` §15 (deferred to MR-795).
### Inline-commit residuals live on `InlineCommitResidual`, not `db.storage()` (MR-793 acceptance §1, by construction)
MR-793's acceptance criterion §1 ("`TableStore` (or successor) public API has no method that performs a manifest commit as a side effect of writing") holds **by construction** after MR-854. `db.storage()` (`&dyn TableStorage`) exposes only staged primitives + reads; the inline-commit writes Lance cannot yet stage live on a separate `InlineCommitResidual` trait reached via `Omnigraph::storage_inline_residual()`. A new engine writer cannot couple a write with a Lance HEAD advance through the default surface — it would have to name the residual accessor explicitly. The dead legacy methods (trait `append_batch` / `merge_insert_batches`, inherent `merge_insert_batch{,es}`, `create_{btree,inverted}_index`) were removed; appends/merges and scalar index builds all use the `stage_*` primitives.
Two methods remain on `InlineCommitResidual`, each named honestly at its call site:
| Residual method | Inline-commit reason | Closes when |
|---|---|---|
| `delete_where` | `DeleteBuilder::execute_uncommitted` is not in Lance v6.0.1 (closed upstream as [#6658](https://github.com/lance-format/lance/issues/6658) but first ships in `v7.0.0-beta.10`); see [docs/dev/lance.md](lance.md) | MR-A: Lance v7.x bump migrates `delete_where` to staged, retires the parse-time D₂ mutation rule, and extends recovery sidecar coverage |
| `create_vector_index` | Vector indices take Lance's "segment commit path"; `build_index_metadata_from_segments` is `pub(crate)` (Lance [#6666](https://github.com/lance-format/lance/issues/6666) still open) | Lance #6666 lands and `stage_create_vector_index` joins the staged surface |
The `tests/forbidden_apis.rs` guard still catches direct `lance::*` inline-commit misuse outside the storage layer; the trait split makes the staged-only default a type-system guarantee on top of it.
### `LoadMode::Overwrite` uses staged Lance `Overwrite`
The bulk loader's Append, Merge, and Overwrite modes all use the
staged-write path described above. `LoadMode::Overwrite` accumulates
replacement batches in memory, validates node/edge constraints, referential
integrity, and edge cardinality before any Lance HEAD movement, stages
each touched table with Lance `Operation::Overwrite`, then runs
`commit_staged` under the normal `SidecarKind::Load` recovery sidecar
before publishing `__manifest`. `OMNIGRAPH_LOAD_CONCURRENCY` applies to the
fragment-writing stage only; the commit and manifest publish still run
under the per-table write queues. Empty-table overwrite is represented as
a valid zero-fragment Lance `Overwrite` transaction, not as
truncate-then-append.
### Open-time recovery sweep
The staged-write rewire eliminates one drift class **by construction at
the writer layer**: an op that fails before pushing to the in-memory
accumulator (validation errors, missing endpoints, parse-time D₂
rejection) leaves Lance HEAD untouched on every staged table. This is
the case the `partial_failure_leaves_target_queryable_and_unblocks_next_mutation`
test pins.
A second, narrower drift class — the **finalize → publisher window**
is closed across one open cycle by the open-time recovery sweep:
`MutationStaging::finalize` runs `stage_*` + `commit_staged` per touched
table sequentially, then the publisher commits the manifest. Lance has
no multi-dataset atomic commit, so the per-table `commit_staged` calls
are independent operations: if commit_staged on table N+1 fails *after*
commit_staged on tables 1..N succeeded, or if the publisher's CAS
pre-check rejects *after* every commit_staged succeeded, tables 1..N
are left at `Lance HEAD = manifest_pinned + 1`.
**Recovery protocol** (lifecycle of every staged-write writer —
`MutationStaging::finalize`, `schema_apply::apply_schema_with_lock`,
`branch_merge_on_current_target`, `ensure_indices_for_branch`,
`optimize_all_tables`):
1. **Phase A**: writer writes a sidecar JSON to
`__recovery/{ulid}.json` BEFORE its first HEAD-advancing commit
(`commit_staged`, or `compact_files` for `optimize_all_tables`,
which advances the Lance HEAD via a reserve-fragments + rewrite
commit rather than a staged write). The
sidecar names every `(table_key, table_path, expected_version,
post_commit_pin)` it intends to commit + the writer kind +
actor_id.
2. **Phase B**: writer's per-table `commit_staged` loop runs.
3. **Phase C**: publisher commits the manifest.
4. **Phase D**: writer deletes the sidecar.
> **Phase letter convention.** Throughout the recovery code, log
> messages, failpoint names (e.g. `branch_merge.post_phase_b_pre_manifest_commit`),
> and the per-writer integration tests, "Phase A/B/C/D" refers
> exclusively to the four-step lifecycle above. The per-table
> staged-write contract (`stage_*` then `commit_staged`, two steps)
> is referred to by those API verbs — never by phase letters — so a
> reader of `recovery.rs`, `failpoints.rs`, or this document only
> encounters phase letters in the per-writer context.
A failure between Phase A and Phase D leaves the sidecar on disk. The
next `Omnigraph::open` (gated on `OpenMode::ReadWrite`) runs the
recovery sweep in `crates/omnigraph/src/db/manifest/recovery.rs`:
- For each sidecar in `__recovery/`, compare every named table's
Lance HEAD to the manifest pin. Classify per the all-or-nothing
decision tree (RolledPastExpected / NoMovement / UnexpectedAtP1 /
UnexpectedMultistep / InvariantViolation).
- If any table is `InvariantViolation` (Lance HEAD < manifest pinned
should be impossible), **abort** with a loud error and leave the
sidecar on disk for operator review.
- Otherwise, if every table is `RolledPastExpected`, **roll forward**:
a single `ManifestBatchPublisher::publish` call extends every pin
atomically. `SchemaApply` sidecars are eligible only when schema-state
recovery promoted the matching staging files in the same recovery pass;
otherwise full open-time recovery rolls them back and refresh-time
recovery leaves them for the next read-write open.
- Otherwise **roll back**: per-table `Dataset::restore` to the
manifest-pinned table version, then a single `ManifestBatchPublisher::publish`
of the restored HEAD symmetric with roll-forward, so `manifest == HEAD`
after recovery (no residual drift). This convergence is what lets a
failed-then-retried schema apply succeed instead of failing one version higher
each iteration. The audit row's `to_version` records the logical
rolled-back-to version (`manifest_pinned`); the manifest is published at the
restore commit (`manifest_pinned + 1`, same content).
- After a successful roll-forward or roll-back, an audit row is
recorded `_graph_commits.lance` carries
a commit tagged `actor_id = "omnigraph:recovery"`, and a sibling
`_graph_commit_recoveries.lance` row carries `recovery_kind`,
`recovery_for_actor` (the original sidecar's actor), `operation_id`,
per-table outcomes. Operators run `omnigraph commit list --filter
actor=omnigraph:recovery` to find recoveries.
- Sidecar deleted as the final step.
Triggers for the residual: transient Lance write errors during finalize
(object-store retry budget exhaustion, disk full); persistent publisher
contention exceeding `PUBLISHER_RETRY_BUDGET = 5` retries.
**Long-running servers**: `Omnigraph::refresh` runs roll-forward-only
recovery in-process the common Phase B Phase C residual closes
without a restart. The next mutation on the same handle (after refresh)
no longer surfaces `ExpectedVersionMismatch` for the failed table.
Sidecars that would require a `Dataset::restore` (mixed / unexpected
state) are deferred to the next `OpenMode::ReadWrite` open: restore is
unsafe under concurrency because Lance's `check_restore_txn` accepts
the restore against in-flight Append/Update/Delete commits and
silently orphans them (pinned by
`tests/staged_writes.rs::lance_restore_loses_to_concurrent_append_via_orphaning`).
Continuous in-process recovery for the rollback path is the goal of a
future background reconciler with per-(table, branch) writer-queue
acquisition.
The publisher-CAS contract is unchanged: a *concurrent writer* that
advances any of our touched tables between snapshot capture and
publisher commit produces exactly one winner. The residual above is
about *our* abandoned commits in the failure path, not about
concurrency races.
## Conflict shape
Concurrent writers to the same `(table, branch)` produce exactly one
success and one failure. The losing writer's error is
`OmniError::Manifest` with kind `Conflict` and details
`ManifestConflictDetails::ExpectedVersionMismatch { table_key, expected,
actual }`. The HTTP server maps this to **409 Conflict** with body
`{"error": "...", "code": "conflict", "manifest_conflict": { "table_key":
"...", "expected": N, "actual": M }}` see [docs/user/server.md](../user/server.md).
## Audit
`actor_id` lands in `_graph_commits.lance` via `record_graph_commit` (no
intermediate run record). Audit history is queried via `omnigraph commit
list`.
## Migration code
`db/manifest/migrations.rs` carries the v2v3 internal-schema step (MR-770):
a one-time sweep that deletes legacy `__run__*` staging branches off
`__manifest`. It runs in `Omnigraph::open(ReadWrite)` (via
`manifest::migrate_on_open`, before the coordinator reads branch state) and
again on the publisher's write path; both are idempotent once the stamp is at
v3. Deleting the inert `_graph_runs.lance` / `_graph_run_actors.lance` dataset
*bytes* is still deferred it needs a `StorageAdapter::delete_prefix`
primitive but those bytes are invisible to graph-level state.
## Mid-query partial failure: closed by MR-794
The pre-MR-794 design had a known limitation: a multi-statement `.gq`
mutation where op-N inline-committed a Lance fragment and op-N+1 then
failed left the touched table at `Lance HEAD = manifest_version + 1`,
blocking the next mutation with `ExpectedVersionMismatch`.
MR-794 (step 1 + step 2+) closed this for inserts/updates **by
construction at the writer layer**: insert and update batches accumulate
in memory; no Lance HEAD advance happens during op execution; one
`stage_*` + `commit_staged` per touched table runs at end-of-query, and
only after every op succeeded. A failed op leaves Lance HEAD untouched
on the staged tables, so the next mutation proceeds normally with no
drift to reconcile.
The cancellation case (future drop mid-mutation) inherits the same
guarantee the in-memory accumulator evaporates with the dropped task
and no Lance write was ever issued.
For delete-touching mutations the legacy inline-commit shape is
preserved (Lance has no public two-phase delete in 6.0.1) the same
narrow window remains. The parse-time D rule prevents inserts/updates
from coexisting with deletes in one query, so a pure-delete failure
cannot drift any staged-table state. If a delete-only multi-table
mutation fails mid-cascade, the same workaround as before applies
(retry; rely on `omnigraph cleanup` once a later successful commit
moves HEAD past the orphan version). Closing this requires Lance to
expose `DeleteJob::execute_uncommitted`; tracked in MR-793 and a
Lance-upstream ticket.