Files
qdrant/lib/shard/src/update.rs
Andrey Vasnetsov ba3472ea49 feat(id_tracker): deferred-aware mutable id tracker (#9249)
* feat(debug): QDRANT_APPEND_ONLY_MUTATIONS env override

Debug-only escape hatch so newly built segments default to append-only
mutation routing when QDRANT_APPEND_ONLY_MUTATIONS=1 (or true/yes) is
set in the environment. Lets us run the existing test suites against
the append-only path without wiring a collection-level config knob
first.

Release builds compile this out — the function is a const false.
Logs a single warn-level message the first time the override fires.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* fix(append-only): skip deleted vectors on snapshot, always write payload

Two correctness fixes to clone_and_mutate_point uncovered by running
the openapi suite with QDRANT_APPEND_ONLY_MUTATIONS=1:

1. The snapshot loop read every named vector via get_vector_opt, which
   returns slot bytes even when the per-vector deletion bit is set.
   For sparse-only points (or any point that had update_vector(_, None)
   applied) this materialised the default-zero vector as if it were
   real data, wrote it at new_id, and never re-tombstoned the slot —
   so dense search started scoring phantom vectors. Now we check
   is_deleted_vector(old_id) and skip the read, letting the writer
   loop emit update_vector(new_id, None) and re-mark the slot deleted.

2. The payload write was skipped when the snapshot ended up empty.
   That dropped two side effects the field indexes rely on:
   payload_storage.overwrite(new_id, empty), and the remove_point
   fan-out across configured field indexes that bumps each index's
   total_point_count to cover new_id. Without the bump the null index
   doesn't see new_id, so is_empty / is_null filters lose the point
   even though its mapping is live in the id tracker. The skip was an
   optimisation, not a contract; remove it so the field indexes get
   the same registration they'd get from the standard clear_payload
   path.

Also collapses the debug env override to an inline cfg!()-gated
check at the struct literal — the helper with one-time logging and
multi-value matching was disproportionate for a debug-only escape
hatch.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* refactor(id_tracker): split active vs deferred maps in PointMappings (PR A)

First step of moving the mutable id tracker to a "two-track" model so
append-only mutations into a deferred segment can keep both the
visible (active) version and the latest mutated (deferred) version of
a point at the same time.

This PR is purely structural. On-disk format is unchanged: the loader
still produces a single combined map, and `PointMappings::new`
partitions it at construction time based on `deferred_internal_id`.
Every observable behaviour at the existing API surface is preserved.

Concretely:

- New fields: `external_to_internal_num_deferred`,
  `external_to_internal_uuid_deferred`, and a `shadowed: BitVec` for
  future use by PR B (lazy-grown, default-false).
- `internal_id(ext)` checks active first, falls through to deferred —
  matches the pre-split "any matching id" contract for ext ids whose
  internal id sat above the cutoff.
- `set_link(ext, new_id)` now routes by cutoff: writes below the
  cutoff land in active, writes at or above land in deferred. Any
  prior head in the other track is tombstoned, so each ext still owns
  exactly one slot — same observable result as the pre-split
  single-map insert. PR B replaces the cross-track tombstone with a
  shadow-bit flip; PR A keeps current semantics on purpose.
- `drop(ext)` clears entries from both tracks and tombstones each
  one, again matching the prior single-map behaviour.
- `iter_external` and `iter_from` merge the active and deferred
  BTreeMap views into one sorted-by-key stream (dedup'd in case an
  ext exists in both tracks).
- `available_point_count` counts distinct external ids across both
  tracks — preserves the prior observable count for segments where
  some entries used to sit above the cutoff in the single map.

No write-path or read-path behaviour change. Reads still filter
`internal_id >= cutoff` exactly as before; mutations still tombstone
prior heads. The shadow bit and the deferred-aware lookup wiring
land in PR B and PR C.

All existing id_tracker tests pass against the split layout.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* feat(id_tracker): shadow active head on deferred writes (PR B)

Step 2 of the deferred-aware mutable id tracker. Replaces PR A's
"tombstone the other track" cross-track cleanup with shadow-bit logic
so a deferred mutation no longer hides the visible active version.

Behaviour by case:

- Active write (`internal_id < cutoff`, or no cutoff): unchanged
  semantically. Any prior deferred head for the same ext is dropped
  and tombstoned (the new active is the single visible head). Same
  observable result as PR A.
- Deferred write (`internal_id >= cutoff`): the prior deferred head
  (if any) is dropped and tombstoned. The active head, if it exists,
  is **shadowed** — its bit is set in `shadowed: BitVec` but its slot
  stays alive in the active map. Read paths in `Exclude` mode
  continue to return that active version; PR C will teach
  `IncludeAll` paths (the optimiser) to skip shadowed actives and
  prefer the deferred head.

Also adds `is_shadowed(internal_id)` and `shadowed_bitslice()`
accessors (the latter for PR C's filter pipeline). New unit tests
cover the routing matrix:

- no cutoff — active replacement, no shadow,
- below-cutoff replacement — active path, no shadow,
- deferred-on-top-of-active — shadow set, active retained,
- two deferred writes — prior deferred tombstoned, shadow persists,
- fresh insert above cutoff — no shadow,
- `drop(ext)` clears both tracks plus the shadow bit.

WAL replay reuses the same `set_link`, so deferred routing on
replay falls out of this change with no extra wiring.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* feat(id_tracker): IncludeAll skips shadowed actives, deferred-aware lookups (PR C)

Step 3 of the deferred-aware id tracker stack. PR B introduced the
shadow bit on `set_link`; this PR teaches the read paths to honour it
so the optimiser (and any other `DeferredBehavior::IncludeAll`
consumer) sees each external id exactly once, with the deferred head
winning over the now-shadowed active.

Concretely:

- `PointMappings::internal_id_with_behavior(ext, behavior)`:
  * `Exclude` returns the active head only — `None` for a
    deferred-only ext, so query paths never see a deferred mutation;
  * `IncludeAll` prefers the deferred head and falls back to active
    for points that never crossed the cutoff. Yields at most one
    internal id per ext.
- `IdTrackerRead::internal_id_with_behavior` mirrors the new method
  with a default impl that delegates to `internal_id` for trackers
  that don't carry deferred mutations.
- `PointMappingsRefEnum::iter_internal_with_behavior(IncludeAll)`
  now filters shadowed actives via the new
  `PointMappings::shadowed_bitslice()` accessor.
- `PointMappingsRefEnum::filter_deferred_and_deleted(IncludeAll)`
  also filters shadowed actives — same single-yield-per-external
  guarantee for external iterator sources like field-index outputs.
- `IdTrackerRead::resolve_external_ids` switches to the deferred-aware
  lookup. No more post-lookup `id >= cutoff` filter — the behaviour
  enum lookup gets it right at the source.

New unit tests cover the two new entry points:

- IncludeAll prefers the deferred head when an active is shadowed;
- Exclude returns None for deferred-only ext ids;
- `filter_deferred_and_deleted` over a mixed candidate list yields
  the expected per-mode result (Exclude: actives below cutoff;
  IncludeAll: every visible head, no shadowed actives).

No queries observable behaviour change today — production Exclude
paths still resolve via `internal_id` and the active head. The
optimiser will start using `IncludeAll` (and reap the dedup) in
follow-up work.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* feat(id_tracker): expose shadowed_point_count in SegmentInfo (PR D)

Final step of the deferred-aware id tracker stack. Adds a single new
counter — number of active heads currently shadowed by a deferred
mutation — and plumbs it through the id-tracker trait, the segment
read-view helper, and `SegmentInfo` for telemetry.

Preserves existing semantics:

- `available_point_count` is unchanged. Each distinct external id
  still counts once, regardless of which track holds its head.
- `deferred_point_count`, `deferred_internal_id`,
  `num_deleted_deferred_points` keep their current values.
- Non-appendable trackers default `shadowed_point_count()` to `0`,
  so the new `SegmentInfo.num_shadowed_points` is `None` for them.

Concrete changes:

- `PointMappings::shadowed_count()` — popcount of the shadowed
  bitslice.
- `IdTrackerRead::shadowed_point_count()` trait method with a `0`
  default; wired through `MutableIdTracker`, `InMemoryIdTracker`,
  the mutable read-only tracker, and both enum dispatchers.
- `SegmentReadView::shadowed_point_count()` helper.
- New `SegmentInfo.num_shadowed_points: Option<usize>`, populated
  with `Some(_)` for appendable segments and proxied through
  `ProxySegment` from the wrapped segment's value.

New unit test covers the counter lifecycle:

- active-only writes don't grow it,
- a deferred write over an active adds one shadow,
- a second deferred write supersedes the prior deferred head but
  the shadow stays put (still one active being shadowed),
- `drop(ext)` clears the shadow bit,
- a fresh deferred insert with no active prior doesn't add a shadow.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* refactor(id_tracker): rename DeferredBehavior, push behavior down to PointMappings

Three connected pieces of polish on top of the deferred-aware id
tracker stack:

1. Rename `DeferredBehavior::Exclude` → `VisibleOnly` and
   `IncludeAll` → `WithDeferred`. The pre-PR-C semantics of
   "include/exclude the deferred cutoff" became misleading once
   `IncludeAll` started skipping shadowed actives — it doesn't
   "include all" anymore, it yields one slot per external (deferred
   head preferred, active fallback). New names describe what each
   variant returns instead of how it relates to the cutoff. The
   helper method `include_all_points` becomes
   `with_deferred_points`. Updates ~90 call sites across the
   workspace; behaviour is unchanged.

2. Push `iter_internal_with_behavior` down from
   `PointMappingsRefEnum` into `PointMappings`. The per-mode
   logic (cutoff `take_while`, shadowed `filter`) now lives next
   to the data it consults; the enum layer becomes a two-arm
   `Either` dispatcher. `CompressedPointMappings` short-circuits
   to `iter_internal()` since compressed mappings can't carry
   deferred mutations.

3. Add a short docstring on `internal_to_external` describing the
   two-track model: no active-vs-deferred bias, shadowed pairs
   occupy two slots with the same value, and reads must gate on
   `deleted` because `set_link`'s same-track replacement leaves a
   real-looking stale ext id in place.

Returning `impl Iterator` instead of `Box<dyn Iterator>` for
`iter_internal`, `iter_internal_excluding`, `iter_internal_visible`,
and `iter_internal_with_behavior` removes the double-Box at the
enum boundary. The original branching structure is preserved with
`itertools::Either` instead of restructured into one big filter
chain.

All existing `id_tracker` tests still pass (47 total).
`cargo check --all-targets` + `cargo clippy --all-targets` clean.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* refactor(id_tracker): drop shadowed API surface, unbox iter_from/iter_random

Cleanup pass on top of the deferred-aware id tracker stack.

API surface:
- Drop SegmentInfo.num_shadowed_points + its trait method, read-view
  helper, tracker impls and ProxySegment passthrough. Cross-segment
  shadow already exists in the appendable update flow (the source
  segment keeps its copy when the new write lands above the cutoff,
  see SegmentHolder::apply_points_with_conditional_move) and is
  handled at the aggregation layer, not reported per-tracker.
- Keep PointMappings::shadowed_count() as a private popcount helper
  guarded by expect(dead_code) so a future caller can opt back into
  the dedup'd count without re-plumbing the trait.

available_point_count:
- Replace the cross-track dedup (active + deferred-only via
  contains_key filter) with a straight 4-term sum. A shadowed ext
  contributes two slots, matching the two non-tombstoned internal
  ids it actually occupies. Old dedup broke the invariant that
  deleted_point_count == deleted_bitslice.count_ones(): for each
  shadow it overcounted deletions by one without any tombstone
  actually being set.

DeferredBehavior pushdown:
- iter_random_with_behavior: caps the sampling range at the
  deferred threshold in VisibleOnly mode (no wasted samples above
  cutoff), filters shadowed actives via the bit in WithDeferred.
- iter_from_with_behavior: VisibleOnly walks the active maps only
  (no merge with deferred); WithDeferred delegates to iter_from's
  existing merge.
- scroll.rs read_by_id_stream / filtered_read_by_id_stream collapse
  their manual if-deferred-behavior branching into a single
  iter_from_with_behavior call.
- Old iter_random (no behavior) at PointMappings + ref enum was
  unused after the migration, deleted.

Unboxing iter_from / iter_random / iter_from_with_behavior:
- PointMappings::iter_from returns impl Iterator + '_ via Either
  inside the merged_num/merged_uuid closures (BTreeMap::iter vs
  range), Either at the outer match (num+uuid chain vs uuid-only).
- PointMappings::iter_from_with_behavior unboxed with a triple
  Either (behavior, external-id arm, closure start).
- CompressedPointMappings::iter_from unboxed (Either over None/Some).
- PointMappingsRefEnum::{iter_from, iter_from_with_behavior,
  iter_from_visible} all return impl Iterator + 'a via Either on
  the Plain/Compressed dispatch.

Other:
- Update outdated PR-A/PR-B comment on PointMappings::drop.
- Make the max_internal match in iter_random_with_behavior
  exhaustive (VisibleOnly+None | WithDeferred+_ instead of `_`).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* refactor(id_tracker): unbox iter_external

Return impl Iterator from PointMappings::iter_external,
CompressedPointMappings::iter_external and the PointMappingsRefEnum
wrapper, matching the style of the other iter_* helpers. The wrapper
dispatches via Either. The remaining Box::new at Segment::iter_points
stays because self_cell's BoxedPointIdIterator alias needs a sized
type.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(id_tracker): count set_link tombstones in deferred_deleted_count

PointMappings::set_link tombstones the prior slot at two sites — when
an active write supersedes a deferred head (cross-track) and when any
write replaces a same-track head — using a bare deleted.set(old, true).
Neither path updated deferred_deleted_count, so tombstones above the
cutoff weren't reflected in the counter.

The append-only flow exposes this constantly: every set_full_payload
after upsert_point routes through clone_and_mutate_point, which
re-issues set_link with a fresh internal id and tombstones the prior
one. Most tombstones land above the cutoff, so deferred_point_count
(total - cutoff - deferred_deleted_count) over-reports by the missing
count. The openapi test_deferred_points integration test caught this
as `num_points - num_deferred_points = -1800` across two segments.

Extracted the tombstone bookkeeping into PointMappings::tombstone_slot
and routed both set_link sites + drop's loop through it. Behaviour on
drop is unchanged; set_link now bumps the counter on the
live → tombstoned transition for slots at or above the cutoff.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* style(id_tracker): collapse set_link tombstone if-let chains

Clippy's collapsible_if on the two if-let blocks added by the
deferred_deleted_count fix.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* Update lib/segment/src/id_tracker/point_mappings.rs

Co-authored-by: Tim Visée <tim+github@visee.me>

* refactor(id_tracker): size shadowed BitVec once instead of growing lazily

Collect the shadowed active ids up front and allocate the BitVec to the
highest offset in a single resize, avoiding repeated reallocations while
marking shadows. Addresses review feedback.

Co-authored-by: Cursor <cursoragent@cursor.com>

* revert: restore lazily-grown shadowed BitVec

Roll back the up-front sizing of the shadowed BitVec; last_entry doesn't
fit here and swapping one allocation for another isn't worthwhile.

Co-authored-by: Cursor <cursoragent@cursor.com>

* fix(id_tracker): prefer deferred head in iter_from merge

When an external id has both an active and a deferred head, iter_from's
merge collapsed the pair to the active (stale) offset. That contradicts
the WithDeferred contract used everywhere else: internal_id_with_behavior
and iter_random_with_behavior both surface the deferred head (the latest
mutation) over the shadowed active. Consumers that use the returned
internal id (payload-filter checks, the optimizer's version merge, HNSW
old->new mapping) therefore saw the stale copy.

Flip the Both arm to take the deferred operand and document the rule.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(id_tracker): preserve active+deferred heads when loading mappings

read_mappings replayed the persisted change log into a single flat
external -> internal map per id-type, then split it by the deferred
cutoff in PointMappings::new. A flat map holds one head per external id,
so it collapsed the active+deferred coexistence case (an external id
linked first to an active slot, then to a deferred one via sequential
set_link) down to the last write — silently dropping the other head,
orphaning its slot, and leaving the shadowed bit unset. The split in
new() could not recover what was already lost before it.

Replay the log through the canonical set_link/drop mutators on a
PointMappings seeded with the cutoff instead. The log is the sequence of
set_link/drop calls that produced the live in-memory state, so this
reconstructs that state exactly — both heads, the shadowed bit, and
deferred_deleted_count — with no logic duplication. Drops the
debug_assert-guarded corruption-recovery branch (subsumed by set_link's
re-link handling) and the now-unused Uuid/PointIdType imports.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* refactor(id_tracker): make deferred behavior explicit at point resolution

Replace the ambiguous "any matching head" point resolution with an
explicit DeferredBehavior at every resolution boundary, so callers no
longer rely on a hidden active-vs-deferred policy.

- Remove PointMappings::internal_id and the bare IdTrackerRead::internal_id
  (the active-first-else-deferred hybrid). internal_id_with_behavior is now
  the single required resolution method; immutable/compressed trackers
  implement it by ignoring the behavior (they never carry deferred heads).
- Migrate every caller to an explicit behavior, audit-driven:
  - writes (upsert/delete/payload/vectors), point_version, point_is_deferred,
    get_internal_id, drop, consistency + builder dedup -> WithDeferred (the
    latest/live head);
  - single-point payload/vector retrieval and formula rescore -> VisibleOnly;
  - HasId/CustomIdChecker/cardinality resolution -> the request's behavior,
    threaded through the filter chain from iter_filtered_points (other entry
    points default to VisibleOnly).
- lookup_internal_id takes an explicit DeferredBehavior instead of assuming
  VisibleOnly internally.
- has_point takes an explicit DeferredBehavior (drop the has_point_with_behavior
  wrapper). Thread it through read_points/_read_points/read_points_locked so
  retrieve_blocking passes its request behavior to the existence filter; all
  other existence/dedup callers pass WithDeferred (unchanged behavior).
- set_link now detaches a stale live occupant of a reused internal id, keeping
  the forward and reverse maps consistent when recovering a corrupted log.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* test(id_tracker): failing tests for two correctness findings in #9249 (#9311)

* test(id_tracker): failing tests for two review findings

Two intentionally-failing tests pinning correctness gaps in the
deferred-aware id tracker (#9249). Both fail on the final assertion;
the earlier assertions establish the expected/consistent behavior.

1. iter_from_with_behavior(WithDeferred) resolves an active/deferred
   shadow collision to the stale ACTIVE internal id, while its siblings
   internal_id_with_behavior and iter_internal_with_behavior correctly
   surface the DEFERRED (latest) head. Consumers that use the yielded
   internal id (optimizer merge via for_each_unique_point,
   filtered_read_by_id_stream) therefore observe the pre-mutation
   version.
     left: [(NumId(7), 2)]   right: [(NumId(7), 9)]

2. The PR-B shadow/visible invariant is not durable: the on-disk single
   combined map cannot represent a shadowed ext, so a plain mappings
   flush + reload collapses the shadow to deferred-only and the visible
   (active) head is lost (VisibleOnly resolves None where it resolved
   Some(2) live). Restoration then depends entirely on WAL replay, i.e.
   on flush-vs-WAL-truncate ordering.
     left: None   right: Some(2)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* test(id_tracker): strengthen shadow tests + merge-primitive proof

Follow-up to the two failing tests, addressing self-review:

- Add for_each_unique_point_keeps_deferred_head_for_shadowed_point: the
  optimizer merge primitive (used by segment_builder::update_from) yields
  the stale active copy (internal 2, version 5) for a shadowed point and
  drops the deferred latest (internal 9, version 8). This directly
  exercises the data-loss consequence of finding #1 at the merge layer.
    left: [(NumId(7), 2, 5)]   right: [(NumId(7), 9, 8)]

- Tighten shadow_visible_head_survives_mapping_flush_reload: pin the exact
  reload failure mode. After flush+reload the mapping collapses to
  deferred-only (internal_id == Some(9)) and the active slot survives as a
  live orphan in the inverse map (external_id(2) == Some(7), not deleted) —
  a torn state where a VisibleOnly scroll still surfaces the stale copy
  while by-id VisibleOnly resolution breaks. Reframe as the live-vs-reload
  divergence the PR introduces (Some(2) live -> None reload; dev is
  consistently None).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix test

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: generall <andrey@vasnetsov.com>

---------

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Tim Visée <tim+github@visee.me>
Co-authored-by: root <111755117+qdrant-cloud-bot@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Arnaud Gourlay <arnaud.gourlay@gmail.com>
2026-06-08 15:03:03 +02:00

1624 lines
60 KiB
Rust
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
//! A collection of functions for updating points and payloads stored in segments
use std::sync::atomic::AtomicBool;
use ahash::{AHashMap, AHashSet};
use common::counter::hardware_counter::HardwareCounterCell;
use common::types::DeferredBehavior;
use parking_lot::RwLockWriteGuard;
use segment::common::operation_error::{OperationError, OperationResult};
use segment::data_types::build_index_result::BuildFieldIndexResult;
use segment::data_types::named_vectors::NamedVectors;
use segment::entry::entry_point::SegmentEntry;
use segment::json_path::JsonPath;
use segment::types::{
Condition, Filter, Payload, PayloadFieldSchema, PayloadKeyType, PayloadKeyTypeRef, PointIdType,
SeqNumberType, VectorNameBuf, WithPayload, WithVector,
};
use crate::operations::payload_ops::PayloadOps;
use crate::operations::point_ops::{
ConditionalInsertOperationInternal, PointOperations, PointStructPersisted, UpdateMode,
};
use crate::operations::vector_ops::{PointVectorsPersisted, UpdateVectorsOp, VectorOperations};
use crate::operations::{
CreateVectorName, DeleteVectorName, FieldIndexOperations, VectorNameOperations,
};
use crate::segment_holder::{SegmentHolder, SegmentId};
pub fn process_point_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
point_operation: PointOperations,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
match point_operation {
PointOperations::UpsertPoints(operation) => {
let points = operation.into_point_vec();
let res = upsert_points(segments, op_num, points.iter(), hw_counter)?;
Ok(res)
}
PointOperations::UpsertPointsConditional(operation) => {
conditional_upsert(segments, op_num, operation, hw_counter)
}
PointOperations::DeletePoints { ids } => delete_points(segments, op_num, &ids, hw_counter),
PointOperations::DeletePointsByFilter(filter) => {
delete_points_by_filter(segments, op_num, &filter, hw_counter)
}
PointOperations::SyncPoints(operation) => {
let (deleted, new, updated) = sync_points(
segments,
op_num,
operation.from_id,
operation.to_id,
&operation.points,
hw_counter,
)?;
Ok(deleted + new + updated)
}
}
}
#[cfg(feature = "staging")]
pub fn process_staging_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
operation: crate::operations::staging::StagingOperations,
) -> OperationResult<usize> {
match operation {
crate::operations::staging::StagingOperations::Delay(delay_op) => {
delay_op.execute();
}
}
// This operation doesn't directly affect segment/point versions, so we bump it here
segments.bump_max_segment_version_overwrite(op_num);
Ok(0)
}
pub fn process_vector_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
vector_operation: VectorOperations,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
match vector_operation {
VectorOperations::UpdateVectors(update_vectors) => {
update_vectors_conditional(segments, op_num, update_vectors, hw_counter)
}
VectorOperations::DeleteVectors(ids, vector_names) => {
delete_vectors(segments, op_num, &ids.points, &vector_names, hw_counter)
}
VectorOperations::DeleteVectorsByFilter(filter, vector_names) => {
delete_vectors_by_filter(segments, op_num, &filter, &vector_names, hw_counter)
}
}
}
pub fn process_payload_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
payload_operation: PayloadOps,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
match payload_operation {
PayloadOps::SetPayload(sp) => {
let payload: Payload = sp.payload;
if let Some(points) = sp.points {
set_payload(segments, op_num, &payload, &points, &sp.key, hw_counter)
} else if let Some(filter) = sp.filter {
set_payload_by_filter(segments, op_num, &payload, &filter, &sp.key, hw_counter)
} else {
// TODO: BadRequest (prev) vs BadInput (current)!?
Err(OperationError::validation_error(
"No points or filter specified",
))
}
}
PayloadOps::DeletePayload(dp) => {
if let Some(points) = dp.points {
delete_payload(segments, op_num, &points, &dp.keys, hw_counter)
} else if let Some(filter) = dp.filter {
delete_payload_by_filter(segments, op_num, &filter, &dp.keys, hw_counter)
} else {
// TODO: BadRequest (prev) vs BadInput (current)!?
Err(OperationError::validation_error(
"No points or filter specified",
))
}
}
PayloadOps::ClearPayload { ref points, .. } => {
clear_payload(segments, op_num, points, hw_counter)
}
PayloadOps::ClearPayloadByFilter(ref filter) => {
clear_payload_by_filter(segments, op_num, filter, hw_counter)
}
PayloadOps::OverwritePayload(sp) => {
let payload: Payload = sp.payload;
if let Some(points) = sp.points {
overwrite_payload(segments, op_num, &payload, &points, hw_counter)
} else if let Some(filter) = sp.filter {
overwrite_payload_by_filter(segments, op_num, &payload, &filter, hw_counter)
} else {
// TODO: BadRequest (prev) vs BadInput (current)!?
Err(OperationError::validation_error(
"No points or filter specified",
))
}
}
}
}
pub fn process_field_index_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
field_index_operation: &FieldIndexOperations,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
match field_index_operation {
FieldIndexOperations::CreateIndex(index_data) => create_field_index(
segments,
op_num,
&index_data.field_name,
index_data.field_schema.as_ref(),
hw_counter,
),
FieldIndexOperations::DeleteIndex(field_name) => {
delete_field_index(segments, op_num, field_name)
}
}
}
/// Do not insert more than this number of points in a single update operation chunk
/// This is needed to avoid locking segments for too long, so that
/// parallel read operations are not starved.
const UPDATE_OP_CHUNK_SIZE: usize = 32;
/// Checks point id in each segment, update point if found.
/// All not found points are inserted into random segment.
/// Returns: number of updated points.
pub fn upsert_points<'a, T>(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: T,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize>
where
T: IntoIterator<Item = &'a PointStructPersisted>,
{
let points_map: AHashMap<PointIdType, _> = points.into_iter().map(|p| (p.id, p)).collect();
let ids: Vec<PointIdType> = points_map.keys().copied().collect();
let mut res = 0;
for ids_chunk in ids.chunks(UPDATE_OP_CHUNK_SIZE) {
// Update points in writable segments
let updated_points = segments.apply_points_with_conditional_move(
op_num,
ids_chunk,
|id, write_segment| {
let point = points_map[&id];
upsert_with_payload(
write_segment,
op_num,
id,
point.get_vectors(),
point.payload.as_ref(),
hw_counter,
)
},
|id, vectors, old_payload| {
let point = points_map[&id];
for (name, vec) in point.get_vectors() {
vectors.insert(name.into(), vec.to_owned());
}
if let Some(payload) = &point.payload {
*old_payload = payload.clone();
}
},
hw_counter,
)?;
res += updated_points.len();
// Insert new points, which was not updated or existed
let new_point_ids = ids_chunk
.iter()
.copied()
.filter(|x| !updated_points.contains(x));
{
let default_write_segment =
segments.smallest_appendable_segment().ok_or_else(|| {
OperationError::service_error(
"No appendable segments exist, expected at least one",
)
})?;
let segment_arc = default_write_segment.get();
let mut write_segment = segment_arc.write();
for point_id in new_point_ids {
let point = points_map[&point_id];
res += usize::from(upsert_with_payload(
&mut write_segment,
op_num,
point_id,
point.get_vectors(),
point.payload.as_ref(),
hw_counter,
)?);
}
RwLockWriteGuard::unlock_fair(write_segment);
};
}
Ok(res)
}
pub fn conditional_upsert(
segments: &SegmentHolder,
op_num: SeqNumberType,
operation: ConditionalInsertOperationInternal,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
// Find points, which do exist, but don't match the condition.
// Exclude those points from the upsert operation.
let ConditionalInsertOperationInternal {
mut points_op,
condition,
update_mode,
} = operation;
let point_ids = points_op.point_ids();
let update_mode = update_mode.unwrap_or_default();
match update_mode {
UpdateMode::Upsert => {
// Default behavior: insert new points, update existing points that match the condition
let points_to_exclude =
select_excluded_by_filter_ids(segments, point_ids, condition, hw_counter)?;
points_op.retain_point_ids(|idx| !points_to_exclude.contains(idx));
}
UpdateMode::InsertOnly => {
// Only insert new points, skip all existing points entirely
let existing_points = segments.select_existing_points(point_ids);
points_op.retain_point_ids(|idx| !existing_points.contains(idx));
}
UpdateMode::UpdateOnly => {
// Only update existing points that match the condition, don't insert new points
let points_to_exclude =
select_excluded_by_filter_ids(segments, point_ids.clone(), condition, hw_counter)?;
let existing_points = segments.select_existing_points(point_ids);
// Keep only points that exist AND are not excluded by the condition
points_op.retain_point_ids(|idx| {
existing_points.contains(idx) && !points_to_exclude.contains(idx)
});
}
}
let points = points_op.into_point_vec();
let upserted_points = upsert_points(segments, op_num, points.iter(), hw_counter)?;
if upserted_points == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(upserted_points)
}
/// Upsert to a point ID with the specified vectors and payload in the given segment.
///
/// If the payload is None, the existing payload will be cleared.
///
/// Returns
/// - Ok(true) if the operation was successful and point replaced existing value
/// - Ok(false) if the operation was successful and point was inserted
/// - Err if the operation failed
fn upsert_with_payload(
segment: &mut RwLockWriteGuard<dyn SegmentEntry>,
op_num: SeqNumberType,
point_id: PointIdType,
vectors: NamedVectors,
payload: Option<&Payload>,
hw_counter: &HardwareCounterCell,
) -> OperationResult<bool> {
let mut res = segment.upsert_point(op_num, point_id, vectors, hw_counter)?;
if let Some(full_payload) = payload {
res &= segment.set_full_payload(op_num, point_id, full_payload, hw_counter)?;
} else {
res &= segment.clear_payload(op_num, point_id, hw_counter)?;
}
debug_assert!(
segment.has_point(point_id, DeferredBehavior::WithDeferred),
"the point {point_id} should be present immediately after the upsert"
);
Ok(res)
}
/// Max amount of points to delete in a batched deletion iteration
const DELETION_BATCH_SIZE: usize = 512;
/// Tries to delete points from all segments, returns number of actually deleted points.
///
/// Iterates all segments directly (rather than going through `apply_points`) to ensure
/// that ALL copies of a point are deleted, including old non-deferred copies in optimized
/// segments when the latest version is deferred in an appendable segment.
pub fn delete_points(
segments: &SegmentHolder,
op_num: SeqNumberType,
ids: &[PointIdType],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_deleted_points = 0;
for batch in ids.chunks(DELETION_BATCH_SIZE) {
for (_segment_id, segment) in segments.iter() {
let segment_arc = segment.get();
let mut write_segment = segment_arc.write();
for &id in batch {
if write_segment.delete_point(op_num, id, hw_counter)? {
total_deleted_points += 1;
}
}
}
}
Ok(total_deleted_points)
}
/// Deferred points corner case for filtered operations.
///
/// When a point has multiple copies across segments, the old non-deferred copy may match
/// a filter while the newest deferred copy does not. In this case, the operation should NOT
/// be applied to the point — the old copy will be cleaned up during optimization deduplication.
///
/// Given per-segment filter match results, this function returns the set of point IDs that
/// should be excluded from the operation because a newer deferred version exists that wasn't
/// matched by the filter.
fn deferred_points_to_exclude_by_filter(
segments: &SegmentHolder,
per_segment_points: &AHashMap<SegmentId, Vec<PointIdType>>,
) -> AHashSet<PointIdType> {
// Find the maximum version for each point across segments where the filter matched.
let mut max_versions: AHashMap<PointIdType, Option<SeqNumberType>> = Default::default();
for (segment_id, point_ids) in per_segment_points {
let segment = segments.get(*segment_id).unwrap().get().read();
for point_id in point_ids {
let version = segment.point_version(*point_id);
let entry = max_versions.entry(*point_id).or_insert(None);
*entry = std::cmp::max(*entry, version);
}
}
// Check if any deferred point has a newer version than the max matched version.
// Such a point was not matched by the filter (its deferred version has different data),
// so the operation should not be applied.
let mut to_exclude = AHashSet::new();
for (_segment_id, segment) in segments.iter() {
let segment = segment.get().read();
if !segment.has_deferred_points() {
continue;
}
for (point_id, max_version) in &max_versions {
if segment.has_point(*point_id, DeferredBehavior::WithDeferred)
&& segment.point_version(*point_id) > *max_version
&& segment.point_is_deferred(*point_id)
{
to_exclude.insert(*point_id);
}
}
}
to_exclude
}
/// Deletes points from all segments matching the given filter
pub fn delete_points_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
filter: &Filter,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_deleted = 0;
// we dont want to cancel this filtered read
let is_stopped = AtomicBool::new(false);
let mut has_deferred = false;
let mut points_to_delete: AHashMap<_, _> = segments
.iter()
.map(|(segment_id, segment)| {
let segment = segment.get().read();
let point_ids = segment.read_filtered(
None,
None,
Some(filter),
&is_stopped,
hw_counter,
// Include also deferred points.
DeferredBehavior::WithDeferred,
)?;
has_deferred |= segment.has_deferred_points();
Ok((segment_id, point_ids))
})
.collect::<OperationResult<_>>()?;
// Deferred points corner case.
// If the latest version of a point is deferred and does not match the filter,
// we need to skip deletion for all copies and let deduplication during optimization delete old points.
if has_deferred {
let points_to_keep = deferred_points_to_exclude_by_filter(segments, &points_to_delete);
// Expand per-segment lists to include all segments that have each matched point,
// so that ALL copies get deleted (not just the segment where the filter matched).
let all_matched_points: AHashSet<PointIdType> = points_to_delete
.values()
.flat_map(|v| v.iter().copied())
.collect();
for (segment_id, segment) in segments.iter() {
let segment = segment.get().read();
let present: Vec<_> = all_matched_points
.iter()
.copied()
.filter(|point_id| {
segment.has_point(*point_id, DeferredBehavior::WithDeferred)
&& !points_to_keep.contains(point_id)
})
.collect();
points_to_delete.insert(segment_id, present);
}
}
segments.apply_segments_batched(|s, segment_id| {
let Some(curr_points) = points_to_delete.get_mut(&segment_id) else {
return Ok(false);
};
if curr_points.is_empty() {
return Ok(false);
}
let mut deleted_in_batch = 0;
while let Some(point_id) = curr_points.pop() {
if s.delete_point(op_num, point_id, hw_counter)? {
total_deleted += 1;
deleted_in_batch += 1;
}
if deleted_in_batch >= DELETION_BATCH_SIZE {
break;
}
}
Ok(true)
})?;
if total_deleted == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(total_deleted)
}
/// Sync points within a given [from_id; to_id) range.
///
/// 1. Retrieve existing points for a range
/// 2. Remove points, which are not present in the sync operation
/// 3. Retrieve overlapping points, detect which one of them are changed
/// 4. Select new points
/// 5. Upsert points which differ from the stored ones
///
/// Returns:
/// (number of deleted points, number of new points, number of updated points)
pub fn sync_points(
segments: &SegmentHolder,
op_num: SeqNumberType,
from_id: Option<PointIdType>,
to_id: Option<PointIdType>,
points: &[PointStructPersisted],
hw_counter: &HardwareCounterCell,
) -> OperationResult<(usize, usize, usize)> {
let id_to_point: AHashMap<PointIdType, _> = points.iter().map(|p| (p.id, p)).collect();
let sync_points: AHashSet<_> = points.iter().map(|p| p.id).collect();
// 1. Retrieve existing points for a range
let stored_point_ids: AHashSet<_> = segments
.iter()
.flat_map(|(_, segment)| segment.get().read().read_range(from_id, to_id))
.collect();
// 2. Remove points, which are not present in the sync operation
let points_to_remove: Vec<_> = stored_point_ids.difference(&sync_points).copied().collect();
let deleted = delete_points(segments, op_num, points_to_remove.as_slice(), hw_counter)?;
// 3. Retrieve overlapping points, detect which one of them are changed
let existing_point_ids: Vec<_> = stored_point_ids
.intersection(&sync_points)
.copied()
.collect();
let mut points_to_update: Vec<_> = Vec::new();
// we dont want to cancel this filtered read
let is_stopped = AtomicBool::new(false);
let _num_updated = segments.read_points(
existing_point_ids.as_slice(),
&is_stopped,
DeferredBehavior::WithDeferred,
|ids, segment| {
let with_vector = WithVector::Bool(true);
let with_payload = WithPayload::from(true);
// Since we retrieve points, which we already know exist, we expect all of them to be found
let stored_records = segment.retrieve(
ids,
&with_payload,
&with_vector,
hw_counter,
&is_stopped,
DeferredBehavior::WithDeferred,
)?;
let mut updated = 0;
for (id, stored_record) in stored_records {
let point = id_to_point.get(&id).unwrap();
if !point.is_equal_to(&stored_record) {
points_to_update.push(*point);
updated += 1;
}
}
Ok(updated)
},
)?;
// 4. Select new points
let num_updated = points_to_update.len();
let mut num_new = 0;
sync_points.difference(&stored_point_ids).for_each(|id| {
num_new += 1;
points_to_update.push(*id_to_point.get(id).unwrap());
});
// 5. Upsert points which differ from the stored ones
let num_replaced = upsert_points(segments, op_num, points_to_update, hw_counter)?;
debug_assert!(
num_replaced <= num_updated,
"number of replaced points cannot be greater than points to update ({num_replaced} <= {num_updated})",
);
Ok((deleted, num_new, num_updated))
}
/// Batch size when modifying vector
const VECTOR_OP_BATCH_SIZE: usize = 32;
pub fn update_vectors_conditional(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: UpdateVectorsOp,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let UpdateVectorsOp {
mut points,
update_filter,
} = points;
let Some(filter_condition) = update_filter else {
return update_vectors(segments, op_num, points, hw_counter);
};
let point_ids: Vec<_> = points.iter().map(|point| point.id).collect();
let points_to_exclude =
select_excluded_by_filter_ids(segments, point_ids, filter_condition, hw_counter)?;
points.retain(|p| !points_to_exclude.contains(&p.id));
update_vectors(segments, op_num, points, hw_counter)
}
/// Update the specified named vectors of a point, keeping unspecified vectors intact.
fn update_vectors(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: Vec<PointVectorsPersisted>,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
// Build a map of vectors to update per point, merge updates on same point ID
let mut points_map: AHashMap<PointIdType, NamedVectors> = AHashMap::new();
for point in points {
let PointVectorsPersisted { id, vector } = point;
let named_vector = NamedVectors::from(vector);
let entry = points_map.entry(id).or_default();
entry.merge(named_vector);
}
let ids: Vec<PointIdType> = points_map.keys().copied().collect();
let mut total_updated_points = 0;
for batch in ids.chunks(VECTOR_OP_BATCH_SIZE) {
let updated_points = segments.apply_points_with_conditional_move(
op_num,
batch,
|id, write_segment| {
let vectors = points_map[&id].clone();
write_segment.update_vectors(op_num, id, vectors, hw_counter)
},
|id, owned_vectors, _| {
for (vector_name, vector_ref) in points_map[&id].iter() {
owned_vectors.insert(vector_name.to_owned(), vector_ref.to_owned());
}
},
hw_counter,
)?;
check_unprocessed_points(batch, &updated_points)?;
total_updated_points += updated_points.len();
}
Ok(total_updated_points)
}
/// Delete the given named vectors for the given points, keeping other vectors intact.
pub fn delete_vectors(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: &[PointIdType],
vector_names: &[VectorNameBuf],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_deleted_points = 0;
for batch in points.chunks(VECTOR_OP_BATCH_SIZE) {
let modified_points = segments.apply_points_with_conditional_move(
op_num,
batch,
|id, write_segment| {
let mut res = true;
for name in vector_names {
res &= write_segment.delete_vector(op_num, id, name)?;
}
Ok(res)
},
|_, owned_vectors, _| {
for name in vector_names {
owned_vectors.remove_ref(name);
}
},
hw_counter,
)?;
check_unprocessed_points(batch, &modified_points)?;
total_deleted_points += modified_points.len();
}
Ok(total_deleted_points)
}
/// Delete the given named vectors for points matching the given filter, keeping other vectors intact.
pub fn delete_vectors_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
filter: &Filter,
vector_names: &[VectorNameBuf],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let affected_points = points_by_filter(segments, filter, hw_counter)?;
let vectors_deleted =
delete_vectors(segments, op_num, &affected_points, vector_names, hw_counter)?;
if vectors_deleted == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(vectors_deleted)
}
/// Batch size when modifying payload
const PAYLOAD_OP_BATCH_SIZE: usize = 32;
pub fn set_payload(
segments: &SegmentHolder,
op_num: SeqNumberType,
payload: &Payload,
points: &[PointIdType],
key: &Option<JsonPath>,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_updated_points = 0;
for chunk in points.chunks(PAYLOAD_OP_BATCH_SIZE) {
let updated_points = segments.apply_points_with_conditional_move(
op_num,
chunk,
|id, write_segment| write_segment.set_payload(op_num, id, payload, key, hw_counter),
|_, _, old_payload| match key {
Some(key) => old_payload.merge_by_key(payload, key),
None => old_payload.merge(payload),
},
hw_counter,
)?;
check_unprocessed_points(chunk, &updated_points)?;
total_updated_points += updated_points.len();
}
Ok(total_updated_points)
}
pub fn set_payload_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
payload: &Payload,
filter: &Filter,
key: &Option<JsonPath>,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let affected_points = points_by_filter(segments, filter, hw_counter)?;
let points_updated = set_payload(segments, op_num, payload, &affected_points, key, hw_counter)?;
if points_updated == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(points_updated)
}
pub fn delete_payload(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: &[PointIdType],
keys: &[PayloadKeyType],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_deleted_points = 0;
for batch in points.chunks(PAYLOAD_OP_BATCH_SIZE) {
let updated_points = segments.apply_points_with_conditional_move(
op_num,
batch,
|id, write_segment| {
let mut res = true;
for key in keys {
res &= write_segment.delete_payload(op_num, id, key, hw_counter)?;
}
Ok(res)
},
|_, _, payload| {
for key in keys {
payload.remove(key);
}
},
hw_counter,
)?;
check_unprocessed_points(batch, &updated_points)?;
total_deleted_points += updated_points.len();
}
Ok(total_deleted_points)
}
pub fn delete_payload_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
filter: &Filter,
keys: &[PayloadKeyType],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let affected_points = points_by_filter(segments, filter, hw_counter)?;
let points_updated = delete_payload(segments, op_num, &affected_points, keys, hw_counter)?;
if points_updated == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(points_updated)
}
pub fn clear_payload(
segments: &SegmentHolder,
op_num: SeqNumberType,
points: &[PointIdType],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_updated_points = 0;
for batch in points.chunks(PAYLOAD_OP_BATCH_SIZE) {
let updated_points = segments.apply_points_with_conditional_move(
op_num,
batch,
|id, write_segment| write_segment.clear_payload(op_num, id, hw_counter),
|_, _, payload| payload.0.clear(),
hw_counter,
)?;
check_unprocessed_points(batch, &updated_points)?;
total_updated_points += updated_points.len();
}
Ok(total_updated_points)
}
/// Clear Payloads from all segments matching the given filter
pub fn clear_payload_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
filter: &Filter,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let points_to_clear = points_by_filter(segments, filter, hw_counter)?;
let points_cleared = clear_payload(segments, op_num, &points_to_clear, hw_counter)?;
if points_cleared == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(points_cleared)
}
pub fn overwrite_payload(
segments: &SegmentHolder,
op_num: SeqNumberType,
payload: &Payload,
points: &[PointIdType],
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let mut total_updated_points = 0;
for batch in points.chunks(PAYLOAD_OP_BATCH_SIZE) {
let updated_points = segments.apply_points_with_conditional_move(
op_num,
batch,
|id, write_segment| write_segment.set_full_payload(op_num, id, payload, hw_counter),
|_, _, old_payload| {
*old_payload = payload.clone();
},
hw_counter,
)?;
total_updated_points += updated_points.len();
check_unprocessed_points(batch, &updated_points)?;
}
Ok(total_updated_points)
}
pub fn overwrite_payload_by_filter(
segments: &SegmentHolder,
op_num: SeqNumberType,
payload: &Payload,
filter: &Filter,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let affected_points = points_by_filter(segments, filter, hw_counter)?;
let points_updated =
overwrite_payload(segments, op_num, payload, &affected_points, hw_counter)?;
if points_updated == 0 {
// In case we didn't hit any points, we suggest this op_num to the segment-holder to make WAL acknowledge this operation.
// If we don't do this, startup might take up a lot of time in some scenarios because of recovering these no-op operations.
segments.bump_max_segment_version_overwrite(op_num);
}
Ok(points_updated)
}
pub fn create_field_index(
segments: &SegmentHolder,
op_num: SeqNumberType,
field_name: PayloadKeyTypeRef,
field_schema: Option<&PayloadFieldSchema>,
hw_counter: &HardwareCounterCell,
) -> OperationResult<usize> {
let Some(field_schema) = field_schema else {
return Err(OperationError::TypeInferenceError {
field_name: field_name.to_owned(),
});
};
segments.apply_segments(|write_segment| {
write_segment.with_upgraded(|segment| {
segment.delete_field_index_if_incompatible(op_num, field_name, field_schema)
})?;
let (schema, indexes) =
match write_segment.build_field_index(op_num, field_name, field_schema, hw_counter)? {
BuildFieldIndexResult::SkippedByVersion => {
return Ok(false);
}
BuildFieldIndexResult::AlreadyExists => {
return Ok(false);
}
BuildFieldIndexResult::IncompatibleSchema => {
// This is a service error, as we should have just removed the old index
// So it should not be possible to get this error
return Err(OperationError::service_error(format!(
"Incompatible schema for field index on field {field_name}",
)));
}
BuildFieldIndexResult::Built { schema, indexes } => (schema, indexes),
};
write_segment.with_upgraded(|segment| {
segment.apply_field_index(op_num, field_name.to_owned(), schema, indexes)
})
})
}
pub fn delete_field_index(
segments: &SegmentHolder,
op_num: SeqNumberType,
field_name: PayloadKeyTypeRef,
) -> OperationResult<usize> {
segments.apply_segments(|write_segment| {
write_segment.with_upgraded(|segment| segment.delete_field_index(op_num, field_name))
})
}
pub fn process_vector_name_operation(
segments: &SegmentHolder,
op_num: SeqNumberType,
vector_name_operation: &VectorNameOperations,
) -> OperationResult<usize> {
match vector_name_operation {
VectorNameOperations::CreateVectorName(create_data) => {
let CreateVectorName {
vector_name,
config,
} = create_data;
segments.apply_segments(|write_segment| {
write_segment.with_upgraded(|segment| {
segment.create_vector_name(op_num, vector_name, config)
})
})
}
VectorNameOperations::DeleteVectorName(delete_data) => {
let DeleteVectorName { vector_name } = delete_data;
segments.apply_segments(|write_segment| {
write_segment
.with_upgraded(|segment| segment.delete_vector_name(op_num, vector_name))
})
}
}
}
fn select_excluded_by_filter_ids(
segments: &SegmentHolder,
point_ids: impl IntoIterator<Item = PointIdType>,
filter: Filter,
hw_counter: &HardwareCounterCell,
) -> OperationResult<AHashSet<PointIdType>> {
// Filter for points that doesn't match the condition, and have matching
let non_match_filter =
Filter::new_must_not(Condition::Filter(filter)).with_point_ids(point_ids);
Ok(points_by_filter(segments, &non_match_filter, hw_counter)?
.into_iter()
.collect())
}
fn points_by_filter(
segments: &SegmentHolder,
filter: &Filter,
hw_counter: &HardwareCounterCell,
) -> OperationResult<Vec<PointIdType>> {
// we dont want to cancel this filtered read
let is_stopped = AtomicBool::new(false);
let mut has_deferred = false;
let per_segment_points: AHashMap<SegmentId, Vec<PointIdType>> = segments
.iter()
.map(|(segment_id, segment)| {
let segment = segment.get().read();
let point_ids = segment.read_filtered(
None,
None,
Some(filter),
&is_stopped,
hw_counter,
// Read operation used for updates, so we must handle all points
DeferredBehavior::WithDeferred,
)?;
has_deferred |= segment.has_deferred_points();
Ok((segment_id, point_ids))
})
.collect::<OperationResult<_>>()?;
let mut affected_points: Vec<PointIdType> = per_segment_points
.values()
.flat_map(|v| v.iter().copied())
.collect();
// Deferred points corner case: exclude points where the newest version is deferred
// and wasnt matched by the filter (only an old stale copy matched).
if has_deferred {
let to_exclude = deferred_points_to_exclude_by_filter(segments, &per_segment_points);
if !to_exclude.is_empty() {
affected_points.retain(|id| !to_exclude.contains(id));
}
}
Ok(affected_points)
}
fn check_unprocessed_points(
points: &[PointIdType],
processed: &AHashSet<PointIdType>,
) -> OperationResult<usize> {
let first_missed_point = points.iter().copied().find(|p| !processed.contains(p));
match first_missed_point {
None => Ok(processed.len()),
Some(missed_point_id) => Err(OperationError::PointIdError { missed_point_id }),
}
}
#[cfg(test)]
mod test {
use std::sync::Arc;
use common::counter::hardware_counter::HardwareCounterCell;
use parking_lot::RwLock;
use segment::data_types::vectors::{DEFAULT_VECTOR_NAME, only_default_vector};
use segment::entry::ReadSegmentEntry as _;
use segment::entry::entry_point::SegmentEntry as _;
use segment::payload_json;
use segment::types::{
Condition, FieldCondition, Filter, Match, MatchValue, PayloadKeyType, ValueVariants,
};
use tempfile::Builder;
use crate::fixtures::{
build_segment_1, build_segment_2, empty_segment, empty_segment_with_deferred,
};
use crate::segment_holder::SegmentHolder;
use crate::update::{
clear_payload_by_filter, delete_payload_by_filter, delete_points_by_filter,
delete_vectors_by_filter, overwrite_payload_by_filter, set_payload_by_filter,
};
#[test]
fn test_delete_by_filter_version_bump() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let segment1 = build_segment_1(dir.path());
let segment2 = build_segment_2(dir.path());
let hw_counter = HardwareCounterCell::new();
let mut holder = SegmentHolder::default();
let _sid1 = holder.add_new(segment1);
let _sid2 = holder.add_new(segment2);
const DELETE_OP_NUM: u64 = 16;
assert!(
holder
.iter()
.all(|i| i.1.get().read().version() < DELETE_OP_NUM)
);
let old_version = holder
.flush_all(true, false)
.expect("Failed to flush test segment holder");
let segments = Arc::new(RwLock::new(holder));
// A filter that matches no points.
let filter = Filter::new_must(Condition::Field(FieldCondition::new_match(
"color".parse().unwrap(),
Match::Value(MatchValue {
value: ValueVariants::String("white".to_string()),
}),
)));
let deleted_count =
delete_points_by_filter(&segments.read(), DELETE_OP_NUM, &filter, &hw_counter).unwrap();
assert_eq!(deleted_count, 0);
let new_version = segments
.read()
.flush_all(true, false)
.expect("Failed to flush test segment holder");
// Flushing again inrceases by 1 and is now equal to `DELETE_OP_NUM` as we want to acknowledge the empty
// delete operation in WAL.
assert_eq!(old_version + 1, new_version);
assert_eq!(new_version, DELETE_OP_NUM);
}
/// Helper: creates a non-appendable segment with a single point at the given version and city payload.
fn build_non_appendable_with_city(
path: &std::path::Path,
point_id: u64,
version: u64,
city: &str,
) -> segment::segment::Segment {
let hw_counter = HardwareCounterCell::new();
let mut seg = empty_segment(path);
seg.upsert_point(
version,
point_id.into(),
only_default_vector(&[1.0, 0.0, 0.0, 0.0]),
&hw_counter,
)
.unwrap();
let payload: segment::types::Payload = payload_json! {"city": city.to_owned()};
seg.set_payload(version, point_id.into(), &payload, &None, &hw_counter)
.unwrap();
seg.appendable_flag = false;
seg
}
/// Helper: creates an appendable segment with deferred threshold 0 (all points deferred),
/// containing a single point with the given city payload.
fn build_deferred_with_city(
path: &std::path::Path,
point_id: u64,
version: u64,
city: &str,
) -> segment::segment::Segment {
let hw_counter = HardwareCounterCell::new();
// threshold 0 => every point is deferred
let mut seg = empty_segment_with_deferred(path, 0);
seg.upsert_point(
version,
point_id.into(),
only_default_vector(&[1.0, 0.0, 0.0, 0.0]),
&hw_counter,
)
.unwrap();
let payload: segment::types::Payload = payload_json! {"city": city.to_owned()};
seg.set_payload(version, point_id.into(), &payload, &None, &hw_counter)
.unwrap();
assert!(
seg.point_is_deferred(point_id.into()),
"Point {point_id} should be deferred"
);
seg
}
fn city_filter(city: &str) -> Filter {
Filter::new_must(Condition::Field(FieldCondition::new_match(
"city".parse().unwrap(),
Match::Value(MatchValue {
value: ValueVariants::String(city.to_string()),
}),
)))
}
/// Delete by filter with deferred points corner case:
/// - Non-appendable segment: point 1 at version 1, city=Berlin
/// - Appendable+deferred segment: point 1 at version 2, city=Amsterdam
/// - Delete by filter on city=Amsterdam
///
/// The deferred point (newest) matches the filter, so both copies must be deleted.
#[test]
fn test_delete_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let deleted = delete_points_by_filter(&holder, 10, &filter, &hw_counter).unwrap();
// The deferred version matches the filter => both copies deleted.
assert!(deleted > 0, "Should have deleted at least one copy");
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
!app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy should be deleted (matches filter)"
);
assert!(
!non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy should also be deleted (deferred version matched filter)"
);
}
/// Delete by filter with deferred points corner case:
/// - Non-appendable segment: point 1 at version 1, city=Berlin
/// - Appendable+deferred segment: point 1 at version 2, city=Amsterdam
/// - Delete by filter on city=Berlin
///
/// The old copy matches the filter, but the newest version is deferred and does NOT
/// match city=Berlin. The delete must be skipped for all copies so that after
/// optimization deduplication we're left with the Amsterdam version.
#[test]
fn test_delete_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let _deleted = delete_points_by_filter(&holder, 10, &filter, &hw_counter).unwrap();
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
// The deferred version (Amsterdam) does NOT match the filter (Berlin),
// so both copies must be kept. Once the optimizer kicks in and deduplicates,
// only the Amsterdam version will remain.
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept (does not match filter, is newest)"
);
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept (deferred version is newer and does not match filter)"
);
}
// --- set_payload_by_filter deferred tests ---
/// Set payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Amsterdam (matches deferred copy)
///
/// The deferred point (newest) matches the filter, so the operation should be applied.
#[test]
fn test_set_payload_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
holder.add_new(non_appendable);
holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let payload: segment::types::Payload = payload_json! {"color": "red"};
let updated =
set_payload_by_filter(&holder, 10, &payload, &filter, &None, &hw_counter).unwrap();
assert!(updated > 0, "Should have updated at least one point");
}
/// Set payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Berlin (matches old copy only)
///
/// The deferred version does NOT match, so the operation must be skipped.
#[test]
fn test_set_payload_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let payload: segment::types::Payload = payload_json! {"color": "red"};
let updated =
set_payload_by_filter(&holder, 10, &payload, &filter, &None, &hw_counter).unwrap();
assert_eq!(
updated, 0,
"Operation should be skipped (deferred version does not match filter)"
);
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept"
);
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept"
);
}
// --- delete_payload_by_filter deferred tests ---
/// Delete payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Amsterdam (matches deferred copy)
///
/// The deferred point (newest) matches, so the payload key should be deleted.
#[test]
fn test_delete_payload_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
holder.add_new(non_appendable);
holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let keys: Vec<PayloadKeyType> = vec!["city".parse().unwrap()];
let updated = delete_payload_by_filter(&holder, 10, &filter, &keys, &hw_counter).unwrap();
assert!(updated > 0, "Should have updated at least one point");
}
/// Delete payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Berlin (matches old copy only)
///
/// The deferred version does NOT match, so the operation must be skipped.
#[test]
fn test_delete_payload_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let keys: Vec<PayloadKeyType> = vec!["city".parse().unwrap()];
let updated = delete_payload_by_filter(&holder, 10, &filter, &keys, &hw_counter).unwrap();
assert_eq!(
updated, 0,
"Operation should be skipped (deferred version does not match filter)"
);
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept"
);
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept"
);
}
// --- clear_payload_by_filter deferred tests ---
/// Clear payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Amsterdam (matches deferred copy)
///
/// The deferred point (newest) matches, so the payload should be cleared.
#[test]
fn test_clear_payload_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
holder.add_new(non_appendable);
holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let updated = clear_payload_by_filter(&holder, 10, &filter, &hw_counter).unwrap();
assert!(updated > 0, "Should have updated at least one point");
}
/// Clear payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Berlin (matches old copy only)
///
/// The deferred version does NOT match, so the operation must be skipped.
#[test]
fn test_clear_payload_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let updated = clear_payload_by_filter(&holder, 10, &filter, &hw_counter).unwrap();
assert_eq!(
updated, 0,
"Operation should be skipped (deferred version does not match filter)"
);
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept"
);
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept"
);
}
// --- overwrite_payload_by_filter deferred tests ---
/// Overwrite payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Amsterdam (matches deferred copy)
///
/// The deferred point (newest) matches, so the payload should be overwritten.
#[test]
fn test_overwrite_payload_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
holder.add_new(non_appendable);
holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let payload: segment::types::Payload = payload_json! {"color": "red"};
let updated =
overwrite_payload_by_filter(&holder, 10, &payload, &filter, &hw_counter).unwrap();
assert!(updated > 0, "Should have updated at least one point");
}
/// Overwrite payload by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Berlin (matches old copy only)
///
/// The deferred version does NOT match, so the operation must be skipped.
#[test]
fn test_overwrite_payload_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let payload: segment::types::Payload = payload_json! {"color": "red"};
let updated =
overwrite_payload_by_filter(&holder, 10, &payload, &filter, &hw_counter).unwrap();
assert_eq!(
updated, 0,
"Operation should be skipped (deferred version does not match filter)"
);
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept"
);
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept"
);
}
// --- delete_vectors_by_filter deferred tests ---
/// Delete vectors by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Amsterdam (matches deferred copy)
///
/// The deferred point (newest) matches, so the vector should be deleted.
#[test]
fn test_delete_vectors_by_filter_deferred_filter_matches_deferred() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
holder.add_new(non_appendable);
holder.add_new(appendable);
let filter = city_filter("Amsterdam");
let vector_names = vec![DEFAULT_VECTOR_NAME.into()];
let deleted =
delete_vectors_by_filter(&holder, 10, &filter, &vector_names, &hw_counter).unwrap();
assert!(deleted > 0, "Should have deleted at least one vector");
}
/// Delete vectors by filter with deferred points:
/// - Non-appendable: point 1 v1, city=Berlin
/// - Deferred: point 1 v2, city=Amsterdam
/// - Filter: city=Berlin (matches old copy only)
///
/// The deferred version does NOT match, so the operation must be skipped.
#[test]
fn test_delete_vectors_by_filter_deferred_filter_matches_old_copy() {
let dir = Builder::new().prefix("segment_dir").tempdir().unwrap();
let hw_counter = HardwareCounterCell::new();
let non_appendable = build_non_appendable_with_city(dir.path(), 1, 1, "Berlin");
let appendable = build_deferred_with_city(dir.path(), 1, 2, "Amsterdam");
let mut holder = SegmentHolder::default();
let sid_non_app = holder.add_new(non_appendable);
let sid_app = holder.add_new(appendable);
let filter = city_filter("Berlin");
let vector_names = vec![DEFAULT_VECTOR_NAME.into()];
let deleted =
delete_vectors_by_filter(&holder, 10, &filter, &vector_names, &hw_counter).unwrap();
assert_eq!(
deleted, 0,
"Operation should be skipped (deferred version does not match filter)"
);
let non_app = holder.get(sid_non_app).unwrap().get();
let non_app = non_app.read();
let app = holder.get(sid_app).unwrap().get();
let app = app.read();
assert!(
non_app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Old copy must be kept"
);
assert!(
app.has_point(1.into(), common::types::DeferredBehavior::WithDeferred),
"Deferred copy must be kept"
);
}
}