Skip to content

Filtered search: keep invalid entries in a min-heap - #450

Draft
yuejiaointel wants to merge 7 commits into
mainfrom
feature/filtered-search-invalid-heap
Draft

yuejiaointel wants to merge 7 commits into
mainfrom
feature/filtered-search-invalid-heap

Conversation

@yuejiaointel

Copy link
Copy Markdown
Contributor

Built on top of #442 (the branch includes its commits). Only the last two commits are new here: c4d7396...3145c73

Idea

With a filter, the dynamic search buffer (MutableBuffer) keeps every entry that fails the filter in its sorted list, so the walk can go through them. At low pass rates the list fills with failing entries and every insert shifts them.

This change keeps the failing (invalid) entries in a min-heap next to the sorted list. The sorted list holds only valid entries and stays short. Invalid entries are still visited in distance order, with the same rule as before: always while fewer than a search window of valid entries is found, afterwards only if closer than the window's last valid entry.

Changes

  • MutableBuffer (dynamic_search_buffer.h), opt-in via enable_invalid_heap():
    • insert / push_back: invalid entries go to the heap (once per id per search).
    • next(): returns the heap top when it is closer than the best unvisited valid entry.
    • done(): also waits for heap entries that are still worth visiting.
    • clear() / cleanup(): empty the heap.
  • MutableVamanaIndex::search turns it on only when a filter is passed. Unfiltered search, the static index, greedy_search and the batch iterator are unchanged.
  • Test: MutableBuffer "Invalid heap" section.

Not measured yet

Build and benchmark pending.

🤖 Generated with Claude Code

Reuse the MutableBuffer valid/invalid mechanism: a FilteredValidBuilder marks an
entry invalid if it is deleted or the predicate rejects its external id. New
MutableVamanaIndex::search overload taking a predicate, padding rows with
id -1 / +inf and returning per-query found counts. Supplement path checks the
predicate before computing the distance. Orchestrator virtual appended last.
Replace the separate filtered search overload with two optional trailing
arguments on MutableVamanaIndex::search (an id predicate and a per-query
result count). Without a predicate the existing code path runs unchanged.
check_and_supplement_search_buffer gets an optional predicate that is
checked before the distance is computed. The filter predicate type is now
a type-erased std::function, so FilteredValidBuilder is no longer a
template. The orchestrator keeps one appended virtual that forwards to the
index search.
The optional filter on MutableVamanaIndex::search now only chooses which
builder the existing greedy search closure uses; both cases call the same
per_thread_batch_search. FilteredValidBuilder derives from ValidBuilder.
The buffer top-up in single_search applies the filter when the search
closure carries one. The per-query result count out-parameter is removed
(callers count non-padding ids).

per_thread_batch_search now pads result rows with fewer than k entries
(id size_t(-1), distance +inf) instead of reading past the end of the
search buffer, and the post-search id translation skips the padding.
Give ValidBuilder an optional predicate and pass it down through
internal_search_builder, greedy_search_closure and search() as optional
arguments, instead of a separate builder and search path. Filtered-only
checks: the supplement fallback skips failing ids, copy-out pads short rows
with id -1 and the worst score, translation skips padded ids.
The orchestrator reports filtered search as unsupported for indexes whose
search has no filter argument (the concurrent index), instead of failing to
compile. In the filtered test, drop the unfiltered-padding section (padding is
filtered-only) and add an inner-product case checking padding uses -inf.
With a filter, MutableBuffer keeps entries that fail it in a min-heap
instead of the sorted buffer. The sorted buffer then holds only valid
entries, so inserts no longer shift long runs of invalid entries.
Invalid entries are still visited in distance order, using the same
rule as before.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant