GeoRepair

OGC geometry repair and validation for Rust. Passes the GEOS XML validation suite.

This crate is experimental. The API is actively evolving: expect breaking changes between 0.x releases. Core algorithms, I/O backends, and feature flags are all subject to change as we improve correctness and performance.

Detects and fixes invalid GIS geometries (self-intersections, unclosed rings, degenerate shapes, NaN coordinates) using algorithms selected by geometry type. Built-in I/O for WKB, WKT, and a custom binary batch format with no extra dependencies.

The Structure strategy (default) mirrors GEOS ST_MakeValid: planar graph extraction, face walking, winding-number assembly. The Arrange strategy is a CDT-based fallback for complex topologies.

Performance on the 1,579,030-polygon production dataset (i5-12400F, release, parallel batch, GEOS 3.14.1 conda-forge as reference):

Dataset	GeoRepair	GEOS	vs GEOS
Validation (1.58M)	3.0-3.2 s	3.6-3.9 s	0.75-0.9x
Invalid subset	0 polys (2026-08-06)	0	-
Full dataset (1.58M polys)	3.8-4.0 s	3.3-3.5 s	1.1-1.2x

Performance

Real-world dataset (1,579,030 polygons)

Structure batch on a production GIS dataset (813 features from a GeoPackage, flattened to 1,579,030 polygon parts). i5-12400F (6C/12T), release LTO, Rayon 12-thread batch; GEOS is conda-forge 3.14.1, serial per-call, run concurrently via Rayon, geometries built via CoordSeq direct construction (the one-time 1.5 s pre-build is excluded). Ratio is GEOS / GeoRepair: >1 means GeoRepair is faster (same convention as the synthetic tables).

Dataset	GeoRepair	GEOS	Ratio
Validation (1.58M)	2.5 s (1.6 µs/poly)	3.5 s (2.2 µs/poly)	1.4x
Full pass (1.58M)	4.0 s (2.6 µs/poly)	3.5 s (2.2 µs/poly)	0.87x

Validity agreement with GEOS: 100% (0/0 disagreements, 2026-08-07).

Synthetic benchmarks

Full table, parallel batch (µs); ratio = GEOS / GeoRepair, >1 means we win. Methodology: docs/BENCHMARKS.md. Regenerate: python scripts/readme_bench_table.py --update <bench.json>; CI gate: python scripts/readme_bench_table.py --check <bench.json>.

Benchmark	GeoRepair	GEOS	Ratio
valid polygon 4v	0.069	0.252	3.6x
valid polygon 10v	0.141	0.318	2.3x
valid polygon 50v	0.349	0.421	1.2x
valid polygon 100v	0.539	0.589	1.1x
valid polygon 500v	2.1	1.2	0.59x
valid polygon 1000v	3.9	2.1	0.55x
valid polygon 5000v	20.5	9.4	0.46x
valid polygon 10000v	39.0	16.9	0.43x
invalid bowtie 4v	0.64	17.6	27x
invalid bowtie 50v	6.2	60.6	9.7x
invalid bowtie 100v	24.2	117	4.8x
invalid bowtie 500v	54.6	488	8.9x
invalid bowtie 1000v	114	967	8.5x
star poly 100v	7.5	7.1	0.94x
star poly 500v	531	110	0.21x
star poly 1000v	246	268	1.1x
spaghetti 500v	4693	2521	0.54x
spaghetti 2000v	32307	16113	0.50x
self-touch 100v	0.672	0.805	1.2x
self-touch 500v	2.2	1.6	0.75x
self-touch 1000v	4.5	3.4	0.77x
collapsed 100v	19.5	119	6.1x
collapsed 500v	86.9	487	5.6x
collapsed 1000v	172	921	5.3x
near-collinear 100v	63.6	166	2.6x
near-collinear 500v	755	1066	1.4x
near-collinear 1000v	2631	2622	1.00x
large coord 1e12 100v	0.733	0.486	0.66x
large coord 1e12 500v	8.1	1.4	0.17x
large coord 1e12 1000v	33.1	2.6	0.08x
valid line	0.009	0.04	4.4x
zero-length line	0.007	0.252	38x
valid ls 4v	0.026	0.042	1.6x
valid ls 10v	0.056	0.05	0.88x
valid ls 50v	0.145	0.071	0.49x
valid ls 100v	0.282	0.121	0.43x
valid ls 500v	1.2	0.536	0.45x
valid ls 1000v	2.6	2.6	0.97x
collinear ls 4v	0.029	0.523	18x
collinear ls 10v	0.075	0.549	7.3x
collinear ls 50v	0.237	0.66	2.8x
collinear ls 100v	0.443	0.983	2.2x
collinear ls 500v	1.8	2.1	1.1x
collinear ls 1000v	3.8	7.8	2.0x
zigzag ls 10v	0.061	1.3	21x
zigzag ls 50v	0.162	4.7	29x
zigzag ls 100v	0.296	10.6	36x
zigzag ls 500v	1.7	62.1	38x
zigzag ls 1000v	2.9	122	41x
spiral ls 10v	0.063	0.836	13x
spiral ls 50v	0.52	4.3	8.4x
spiral ls 100v	1.3	15.7	12x
spiral ls 500v	112	526	4.7x
spiral ls 1000v	384	1807	4.7x
self-int ls 100v	5.5	2.5	0.45x
self-int ls 500v	84.2	3.9	0.05x
self-int ls 1000v	330	6.8	0.02x
dense self ls 10v	0.058	1.3	22x
dense self ls 50v	0.155	5.4	35x
dense self ls 100v	0.304	13.5	44x
dense self ls 500v	1.6	127	80x
dense self ls 1000v	2.5	361	147x
duped ls 100v	0.286	0.692	2.4x
duped ls 500v	0.936	1.6	1.7x
duped ls 1000v	2.2	3.6	1.6x
mls 50x3v	2.8	32.3	12x
mls 250x3v	15.8	132	8.4x
mls 500x3v	38.3	257	6.7x
self-int mls 50x4v	29.7	74.4	2.5x
self-int mls 250x4v	191	435	2.3x
self-int mls 500x4v	520	1091	2.1x
star-burst 10sp	2.5	14.5	5.7x
star-burst 50sp	0.677	258	381x
star-burst 100sp	1.3	1115	867x
star-burst 500sp	7.0	35180	5038x
star-burst 1000sp	16.8	185213	11035x
collinear ov 10seg	0.976	11.7	12x
collinear ov 50seg	6.6	55.4	8.4x
collinear ov 100seg	12.7	104	8.2x
collinear ov 500seg	52.7	534	10x
collinear ov 1000seg	112	1094	9.7x
x-scale 10v	1.7	20.0	12x
x-scale 50v	49.3	376	7.6x
x-scale 100v	234	1541	6.6x
x-scale 500v	16886	53173	3.1x
x-scale 1000v	1719	320188	186x
ringing 100v	11.7	60.8	5.2x
ringing 500v	44.1	344	7.8x
ringing 1000v	123	1009	8.2x
hilbert 256v	35.0	165	4.7x
hilbert 1024v	308	1212	3.9x
lissajous 200v	19.6	142	7.2x
lissajous 500v	42.4	382	9.0x
lissajous 1000v	79.4	723	9.1x
lissajous 2000v	170	1583	9.3x
lissajous 5000v	504	4929	9.8x
lissajous 7:4 500v	33.6	46.8	1.4x
spoke 10sp	2.4	14.0	5.8x
spoke 50sp	0.774	266	343x
spoke 100sp	1.3	1102	816x
spoke 500sp	8.6	36076	4194x
spoke 1000sp	20.7	188121	9095x
star-comb 20sp	0.195	2.6	13x
star-comb 100sp	2.9	32.5	11x
star-comb 500sp	73.9	672	9.1x
star-comb 1000sp	698	2694	3.9x
hole hier 5h	1.5	1.9	1.2x
hole hier 20h	5.5	8.4	1.5x
hole hier 50h	21.9	22.2	1.0x
hole hier 100h	45.6	52.2	1.1x
overlap mp 5sh	1.8	408	224x
overlap mp 20sh	6.8	2428	356x
overlap mp 50sh	21.3	6635	311x
overlap mp 100sh	49.4	13849	280x
dense grid 5x5=25	8.0	1655	206x
dense grid 10x10=100	42.1	11513	273x
dense grid 20x20=400	459	99563	217x
dense grid 30x30=900	1899	361528	190x
sliver 100v	1.5	2.2	1.4x
sliver 500v	11.1	14.4	1.3x
sliver 1000v	16.1	25.9	1.6x
arrange valid 4v	0.059	0.258	4.4x
arrange valid 10v	0.1	0.333	3.3x
arrange valid 50v	0.689	0.462	0.67x
arrange valid 100v	0.902	0.539	0.60x
arrange valid 500v	3.7	1.4	0.38x
arrange valid 1000v	8.0	4.0	0.50x
arrange bowtie 4v	1.2	14.9	13x
arrange bowtie 100v	32.3	101	3.1x
arrange star 10sp	2.2	13.3	6.0x
arrange star 50sp	1.1	251	234x
arrange star 100sp	1.4	1040	733x
arrange star 500sp	7.1	36466	5126x

Run benchmarks

Run benchmarks

Run benchmarks

Run benchmarks

# Real-world + synthetic with GEOS comparison (system GEOS, conda)
GEOS_LIB_DIR='D:\Miniconda\Library\lib' GEOS_INCLUDE_DIR='D:\Miniconda\Library\include' \
GEOS_VERSION=3.14.1 cargo bench --features bench-geos-system,arrange,structure,parallel,simd --bench real_world
GEOS_LIB_DIR='D:\Miniconda\Library\lib' GEOS_INCLUDE_DIR='D:\Miniconda\Library\include' \
GEOS_VERSION=3.14.1 cargo bench --features bench-geos-system,arrange,structure,parallel,simd --bench bench

# CI regression gate (no GEOS; fixed subset vs benches/bench_baseline.json)
python scripts/bench_gate.py

Measurement rules: always take the settled second run (first-run-after-build is inflated ~18%); never trust a bench binary you cannot trace to a source file. Full GEOS setup: benches/AGENTS.md.

Limitations

1. GEOS comparison is against conda-forge MSVC GEOS (serial per-call, no LTO, no mimalloc) - a static LLVM-built GEOS would improve the GEOS side of every table.
2. Validator strictness is deliberate: eps-class predicates (1e-12 x scale tolerance, fast-FP-first with exact escalation). On the 1.58M real-world dataset it agrees with GEOS 0/0 (winding-agnostic); it diverges on borderline inputs (213 baselined XML cases, mostly the OGC winding contract - docs/BENCHMARKS.md). Repair ships only validator-clean geometry and degrades to an empty GeometryCollection otherwise.
3. W/12 pool-saturation floor: the parallel batch fills all 12 workers with giants; nested intra-poly rayon finds no idle threads in-batch (Amdahl-bound; standalone the parallel check measures 96 ms vs 53 ms in-batch).
4. Giants (>4,096 ring edges) route to the boolean pipeline - on a 200k-edge giant single-pass noding costs 168 ms vs 36 ms for the boolean path (SP_MAX_EDGES in src/core/mod.rs).
5. Mass-overlap repairs are the slowest synthetic class (~0.46 ms/poly at dense grid 20x20) but stay ~230x faster than GEOS makeValid on the same shapes.
6. Python bindings: tests/test_python.py covers the WKT surface (18 tests); GeoJSON bindings were removed deliberately.
7. simd-portable is nightly-only (3 E0554 on stable, expected); hand-written AVX2 beyond the bbox scan measured slower than auto-vectorized scalar (point_in_ring 8.4x, is_ring_ccw 2.8x), and -C target-cpu=native regresses the full pass ~25%.
8. proj requires native PROJ and is mutually exclusive with io-gpkg (sqlite3 link conflict); io-gpkg is a default feature, so proj users build with --no-default-features --features proj,....
9. GEOS XML suite: 937/937 dispatched cases pass; 3,629 overlay/relate cases are skipped (out of scope); 213 masked divergences (documented tolerance gates). The suite's WKT/WKB readers are our own - external wkt/wkb crates are disallowed because they measure slower.
10. Sub-µs synthetic rows are noise (Rayon dispatch overhead); the trustworthy metrics are the real-world batch numbers and the larger synthetic rows.

Integration with the geo ecosystem

geo-repair is built on geo types and plugs into the georust ecosystem two ways:

• geo-traits sources (geo-traits feature). The interop module runs validation and repair over geo_traits::GeometryTrait / geo_traits::GeometryCollectionTrait, the trait layer implemented by geo, geoarrow, geozero, and wkb. Any such source can be validated or repaired in one call (interop::is_valid_geometry, interop::make_valid_geometry, interop::make_valid_geometries, ...) without materializing geo types; results come back as geo::Geometry<f64>.

• geo's Validation trait (always available). GeoRepairValidation(&geometry) wraps any &geo::Geometry<f64> and exposes geo_repair's validator through geo's Validation trait (.is_valid(), .check_validation(), .validation_errors()) with geo's Invalid* error taxonomy. The orphan rule prevents implementing geo's trait for geo's own types, so the adapter is the bridge. Mapping is best-effort: geo_repair's stricter gates (32-ulp collinear, T-junction) surface through it, and classes geo does not model (ring closure, orientation, duplicates, ...) are omitted from the geo view.

use geo::algorithm::validation::Validation;
use geo_repair::GeoRepairValidation;

let adapter = GeoRepairValidation(&geometry);
assert!(!adapter.is_valid());

Python bindings

The geo-repair PyPI package exposes the full validation and repair surface over WKB bytes and WKT text, single geometry and batch (including a rayon-backed parallel batch). abi3 wheels (cp38-abi3) serve Python 3.8+; typing stubs are shipped in the wheel.

pip install geo-repair

import geo_repair

fixed = geo_repair.repair_wkt("POLYGON((0 0, 5 5, 5 0, 0 5, 0 0))")
assert geo_repair.is_valid_wkt(fixed)

was_valid, errors, fixed_wkb = geo_repair.validate_and_fix_wkb(wkb_bytes)
results = geo_repair.par_repair_wkb_batch(list_of_wkb_bytes)

Every function exists for WKB and WKT (repair_*, repair_*_batch, par_repair_*_batch, repair_validate_*, is_valid_*, validate_*, validate_and_fix_*, plus batch forms), with method (auto/arrange/structure) and keep_collapsed parameters. A QGIS Processing script (qgis/qgis_geo_repair.py) streams features through the WKB batch API. Full API reference: docs/BINDINGS.md.

C API

The ffi feature exposes a panic-safe C API over WKB and WKT, single geometries and parallel batches:

cargo build --release --features ffi
# output: target/release/geo_repair.{dll,so,dylib} + libgeo_repair.a + include/geo_repair.h

#include "geo_repair.h"

GeoRepairResult r = geo_repair_make_valid(bowtie_wkb, bowtie_wkb_len);
if (r.success) { /* r.wkb_data / r.wkb_len = fixed WKB */ }
geo_repair_free_result(&r);

Every result carries a GeoRepairErrorCode (None/Parse/InvalidInput/ InvalidGeometry/Encode/Panic); batches report per-item outcomes without failing as a whole. All results must be freed with the matching geo_repair_free_* (double-free safe). The ABI (struct layouts, error codes) is fixed from 0.14.2; panic containment requires the release profile's panic = "unwind". Prebuilt libraries for Windows, Linux, and macOS are attached to every GitHub release. Full API reference: docs/BINDINGS.md.

Features

Feature	Description	Default
arrange	CDT-based polygon repair (requires spade)	yes
structure	Structure-based fast path repair	yes
parallel	Rayon parallel processing (non-WASM)	yes
simd	Runtime-dispatched AVX2 bbox kernel (the one SIMD kernel that beat auto-vectorized scalar, 4.5x); all other kernels auto-vectorized scalar	yes
validate	OGC validation predicates	yes
mimalloc	Use mimalloc global allocator	yes
std	Standard library + file I/O. Disable for no_std builds.	yes
simd-portable	Portable SIMD via core::simd (nightly only)	no
memmap	Memory-mapped binary file loading	no
wasm	WASM browser fetch (synchronous XHR)	no
proj	CRS transformation (placeholder)	no
serde	Geometry serde support (geo/serde)	no
ffi	C-compatible FFI bindings	no
geo-traits	Interop surface over geo_traits::GeometryTrait / GeometryCollectionTrait (geo, geoarrow, geozero, wkb sources)	no
python	Python bindings via PyO3	no
io-shp	Shapefile format backend	no
io-wkt	No-op (WKT is built-in, kept for CI compatibility)	no
io-csv	CSV format backend	no
io-gml	GML/XML format backend	no
io-gpkg	GeoPackage format backend (default; gated out on wasm32)	yes
io-all	All opt-in backends except gpkg	no
io-all-native	All opt-in backends including gpkg	no
bench-geos	GEOS comparison benchmarks (build from source, MSVC, no LTO)	no
bench-geos-system	GEOS comparison benchmarks (link against system GEOS, conda-forge MSVC)	no

License

Apache-2.0