feat: support ST_DWithin pushdown in vortex

[HarukiMoriarty]

Summary

Insert non-throwing geo predicate vortex_dwithin in DuckDB, which later pushdown into vortex, call distance scalar function on scanning, significantly improve Q1/Q3 performance in SpatialBench.

What changes are included in this PR?

1. adding new geo predicate vortex_dwithin in DuckDB, which is non-throwing and can be pushdown.
2. adding SQL rewrite so for geo native type, ST_dwithin is text rewritten into vortex_dwithin.

Performance

┏━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━┓
┃ Query ┃ duckdb:parquet (base) ┃   duckdb:vortex ┃ duckdb:vortex-native ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━┩
│ 1     │               179.6ms │ 132.1ms (0.74x) │       29.0ms (0.16x) │
│ 2     │               268.1ms │ 255.3ms (0.95x) │      312.5ms (1.17x) │
│ 3     │               247.8ms │ 216.4ms (0.87x) │      140.2ms (0.57x) │
│ 4     │               199.6ms │ 146.9ms (0.74x) │      144.5ms (0.72x) │
│ 5     │                 3.40s │   3.43s (1.01x) │        3.03s (0.89x) │
│ 6     │               528.1ms │ 359.2ms (0.68x) │      455.4ms (0.86x) │
│ 7     │               984.0ms │ 983.2ms (1.00x) │      887.7ms (0.90x) │
│ 8     │                 1.08s │ 945.4ms (0.87x) │      997.6ms (0.92x) │
│ 9     │                34.2ms │  33.5ms (0.98x) │       43.6ms (1.27x) │
└───────┴───────────────────────┴─────────────────┴──────────────────────┘

Takeaways: Q1 and Q3 is significantly improved due to the single table geo predicate is pushdown into vortex scan. Q1 is more benefit from the manually fast path without going through geo crate for calculation. Q3 is also potentially can be benefit from manually calculation.

[myrrc]

I think CREATE MACRO in init.sql is an easier approach than rewriting the query manually.

[HarukiMoriarty]

Here the queries call ST_GeomFromWKB/ST_DWithin, which the spatial extension already defines, and you can't shadow those: CREATE MACRO ST_DWithin(...) -> Catalog Error: Macro Function with name "ST_DWithin" already exists (same for ST_GeomFromWKB, both tested).

[myrrc]

But this means we won't be able to use this function on real queries since you don't have access to the query string in Duckdb. Let's think of another way. Maybe we can delete the spatial function and then create ours? This won't survive extension re-LOAD, but is a temporary good solution.
Otherwise we can write a small pass replacing ST_DWithin with vortex_DWIthin in the graph, but doing nothing else, this should be a single Visitor override.

[HarukiMoriarty]

It is actually messy than what I thought. Tried replacing ST_DWithin in the catalog with our own bind-cleared copy. Single-table filters push (Q1 ~8.6×, Q3 ~4.9×), but it regresses the join queries.
DuckDB's ST_DWithin's bind erases the constant radius into bind_data, leaving a 2-arg predicate, which is exactly what spatial's spatial-join optimizer requires (first gate: func.children.size() != 2). Clearing the bind to expose the radius gives 3 args, trips that gate, and Q8 drops from SPATIAL_JOIN to BLOCKWISE_NL_JOIN. Confirmed via EXPLAIN (override off → SPATIAL_JOIN, on → BLOCKWISE_NL_JOIN).
So the radius-erase is what we need gone to push but spatial needs present to optimize joins, a global replacement can't have both.

[HarukiMoriarty]

Working on SpatialBench infra itself so we never need GeomFromWKB anymore, though selective of rewritten or overwritten or Macro of ST_DWithin is still needed. Or maybe we just rename it as ST_Distance <= xxx?

[myrrc]

Why do we need a transaction here? Can we register the function in the global catalog without it?

[HarukiMoriarty]

It looks like we cannot not. DuckDB's catalog is MVCC, so CreateFunction needs an active transaction (it throws ActiveTransaction called without active transaction otherwise). A fresh Connection only begins a transaction per query, and we're calling the catalog API directly, so RunFunctionInTransaction is just the "begin → run → commit" wrapper to supply one.
I confirmed by removing it: registration aborts and queries then fail with vortex_dwithin does not exist.

[codspeed-hq]

Merging this PR will improve performance by 12.06%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 2 improved benchmarks
✅ 1587 untouched benchmarks
⏩ 4 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	244.4 ns	215.3 ns	+13.55%
⚡	Simulation	bitwise_not_vortex_buffer_mut[1024]	304.7 ns	275.6 ns	+10.58%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing nemo/geo-native-pushdown (94674e0) with nemo/duckdb-native-geometry (69a4f90)}

Footnotes

1. 4 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩