Route constant-geometry spatial-relationship predicates to the TRTREE index#144
Open
estebanzimanyi wants to merge 3 commits into
Open
Route constant-geometry spatial-relationship predicates to the TRTREE index#144estebanzimanyi wants to merge 3 commits into
estebanzimanyi wants to merge 3 commits into
Conversation
… index eIntersects/eContains/eDisjoint/eTouches against a constant geometry are now recognised by the TRTREE scan optimizer, which synthesizes the bounding-box && prefilter from the constant and probes the index (single-box or MEST). The bbox is a lossy superset, so the index scan reports supports_pushdown_type = false and DuckDB keeps the original predicate as an exact recheck filter above the scan, mirroring the lossy-index-always-rechecks contract of PostGIS GiST and MobilityDB's tspatial_supportfn. A parity block proves that a row whose bbox overlaps the query polygon but does not actually intersect it is dropped, identically with and without the index and under MEST multi-entry.
Member
Author
Reviewer's quickstart — ~6 minutesWhat this PR does in one sentence: when a spatial-rel predicate has a constant geometry on one side (e.g. Files (4):
Verification: CREATE INDEX rt ON trips USING TRTREE (traj);
EXPLAIN SELECT count(*) FROM trips WHERE eIntersects(traj, 'POLYGON(...)'::GEOMETRY);
-- Before this PR: SEQ_SCAN. After: TRTREE_INDEX_SCAN.Cross-link: stacked on #143 (MEST multi-entry indexing), which is in turn the foundation for this optimisation. Both will rebuild cleanly once #161 lands (Linux arm64 build). Why it's safe to merge: the pre-existing scan path remains available for non-constant-geometry cases; this only adds a new index-aware branch. |
`meosType` (lower-case) is the **pre-consolidation** MEOS type name; `MeosType` (upper-case) is the **post-consolidation** target that the upstream rename sweep has not yet reached. The current vcpkg pin (`vcpkg_ports/meos/portfile.cmake` REF f11b7443ee98…) is still pre-consolidation: `meos/include/temporal/meos_catalog.h` line 121 declares the typedef as `} meosType;` and every MEOS API uses the lower-case spelling. MobilityDuck's source code consistently uses `meosType` to match — `grep -rn '\bMeosType\b' src/` finds the name only on the alias line and its comment, nowhere else. c8cad6d added `using meosType = MeosType;` as a forward-looking bridge for the eventual consolidation bump. That bridge points at `MeosType`, which the current pin does NOT yet expose, so it breaks every PR's Linux arm64 build with: /duckdb_build_dir/src/include/tydef.hpp:18:18: error: ‘MeosType’ does not name a type; did you mean ‘meosType’? The fix is to drop the premature alias and replace the misleading comment with one that documents the pre/post-consolidation distinction and the resume path for the next pin bump — at that point a reviewer can either restore the bridge (this time it'll be valid because `MeosType` will exist) or sweep the MobilityDuck source from `meosType` to `MeosType` in a single PR. Unblocks every in-flight PR's Linux arm64 build: #126, #130, #149, #158, #159, #160, plus the entire `feat/*_port_core` extended-type stack (#148/#150/#151/#153/#155/#156).
The stage_icu helper mapped only the Linux uname values, so on the macOS arm64 test runner uname -m returned "arm64" and the icu extension was copied to .duckdb/extensions/v1.4.4/arm64 instead of .../osx_arm64, where DuckDB's autoload looks. The hub fallback is not reliably resolvable on that runner, so the osx_arm64 Test step failed to load the extension. Map the OS and architecture to the DuckDB platform string (linux_amd64, linux_arm64, osx_amd64, osx_arm64) so the locally built icu is staged at the path autoload expects on every tested platform; the Linux mapping is unchanged. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
eIntersects, eContains, eDisjoint and eTouches against a constant geometry are recognised by the TRTREE scan optimizer, which synthesizes the bounding-box && prefilter from the constant and probes the index (single-box or MEST). The bbox is a lossy superset, so the index scan reports supports_pushdown_type = false and DuckDB keeps the original predicate as an exact recheck filter above the scan, mirroring the lossy-index-always-rechecks contract of PostGIS GiST and MobilityDB's tspatial_supportfn. A parity block proves that a row whose bounding box overlaps the query polygon but does not actually intersect it is dropped, identically with and without the index and under MEST multi-entry. Stacked on #143; its commits show in the diff until it merges.