Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@ versions may change the store format and the API; the release notes say how to u

### Changed

- Transitive property paths run on term ids (#11): `?s <p>+ ?o`, `?s ^<p>+ ?o`, and `<p>*` with a constant end, alone or joined, filtered, counted or grouped with other patterns, are found by a search over the predicate's edges on ids instead of one start at a time in the evaluator: from the values the other patterns give when they bind an end (either end), else over all the edges. Same answers. On a tree of 200,000 nodes (1.48 million pairs), a `GROUP BY` count over `?s <p>+ ?o` takes 0.19 s instead of 1.05 s, and the path joined with a class and counted 0.05 s instead of 1.61 s; a path from a few bound starts stays at about 1 ms. `#pragma tt.id-rewrites off` turns it off. See [the store](docs/store.md#design).
- A resource's Turtle document (Linked Data and `/{ds}/describe`) declares prefixes for the namespaces it uses, and every format names blank nodes `b0`, `b1`… in the order they come, so the same description is the same text.
- Web UI: every page uses the full width of the window (the Query, Map, Shapes and admin pages stopped at 1,400 pixels), and the fields outside the top bar (search, number and date boxes, text areas) take the page's colours instead of the browser's grey, most visible in the dark theme.
- `ORDER BY DESC(tt:cosine(?v, "[…]")) LIMIT k`, or `ORDER BY DESC(?score)` with `?score` bound to that cosine, is answered from the copy of the vectors that `tt:nearest` reads (#236): the query is evaluated again with `?v` kept to the vectors most similar, each one compared, so the answer stays exact and other patterns and filters still apply. It works on the union default graph and for a caller limited to some graphs. Among 100,000 vectors of 768 numbers: 40 ms instead of 6.2 s; at most two rounds, so a filter that keeps one row in a hundred costs about 130 ms instead of 93 ms. A query with its own dataset or under row-level rules, with an `OFFSET` and `LIMIT` past a sixteenth of the vectors, or over a property that also has plain strings in the form of a vector, is evaluated as before. See [vector search](docs/vector-search.md#functions).
Expand Down
Loading
Loading