search

georust / geojson

Library for serializing the GeoJSON vector GIS file format
https://crates.io/crates/geojson
Apache License 2.0
280 stars 60 forks source link

For Discussion: Using `tinyvec::ArrayVec` for `Position` without allocating #218

Open phayes opened 1 year ago

phayes commented 1 year ago

This is a PR for discussion. I've moved Position from being a Vec<f64> to being an ArrayVec<f64> with a capacity of 4 elements (x, y, z, t). This means that instead of every point resulting in an allocation, we only allocate to create LineString, Polygon etc.

Preliminary cargo-bench results suggest performance improvements up to 79% in some cases.

If there is agreement that this is a good direction, I'll add some convenience functions like Position::from_vec etc to recover some of the ergonomics of working with Vecs.

Benchmarks:

Benchmarking parse (countries.geojson)
Benchmarking parse (countries.geojson): Warming up for 3.0000 s
Benchmarking parse (countries.geojson): Collecting 100 samples in estimated 5.1934 s (2600 iterations)
Benchmarking parse (countries.geojson): Analyzing
parse (countries.geojson)
                        time:   [1.9881 ms 1.9981 ms 2.0086 ms]
                        change: [-13.706% -13.086% -12.475%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 100 measurements (1.00%)
  1 (1.00%) high mild

Benchmarking FeatureReader::features (countries.geojson)
Benchmarking FeatureReader::features (countries.geojson): Warming up for 3.0000 s
Benchmarking FeatureReader::features (countries.geojson): Collecting 100 samples in estimated 5.5955 s (900 iterations)
Benchmarking FeatureReader::features (countries.geojson): Analyzing
FeatureReader::features (countries.geojson)
                        time:   [6.1732 ms 6.1837 ms 6.1956 ms]
                        change: [-6.3597% -6.0745% -5.7713%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  7 (7.00%) high mild
  2 (2.00%) high severe

Benchmarking FeatureReader::deserialize (countries.geojson)
Benchmarking FeatureReader::deserialize (countries.geojson): Warming up for 3.0000 s
Benchmarking FeatureReader::deserialize (countries.geojson): Collecting 100 samples in estimated 5.5427 s (700 iterations)
Benchmarking FeatureReader::deserialize (countries.geojson): Analyzing
FeatureReader::deserialize (countries.geojson)
                        time:   [8.1079 ms 8.1365 ms 8.1734 ms]
                        change: [-2.0054% -1.5319% -0.9778%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
  1 (1.00%) high mild
  1 (1.00%) high severe

Benchmarking FeatureReader::deserialize to geo-types (countries.geojson)
Benchmarking FeatureReader::deserialize to geo-types (countries.geojson): Warming up for 3.0000 s
Benchmarking FeatureReader::deserialize to geo-types (countries.geojson): Collecting 100 samples in estimated 5.5704 s (700 iterations)
Benchmarking FeatureReader::deserialize to geo-types (countries.geojson): Analyzing
FeatureReader::deserialize to geo-types (countries.geojson)
                        time:   [7.9112 ms 7.9288 ms 7.9473 ms]
                        change: [-2.6196% -2.3486% -2.0713%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild

Benchmarking parse (geometry_collection.geojson)
Benchmarking parse (geometry_collection.geojson): Warming up for 3.0000 s
Benchmarking parse (geometry_collection.geojson): Collecting 100 samples in estimated 5.0480 s (293k iterations)
Benchmarking parse (geometry_collection.geojson): Analyzing
parse (geometry_collection.geojson)
                        time:   [17.236 µs 17.274 µs 17.320 µs]
                        change: [-7.0696% -6.7231% -6.3747%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  3 (3.00%) high mild

WARNING: HTML report generation will become a non-default optional feature in Criterion.rs 0.4.0.
This feature is being moved to cargo-criterion (https://github.com/bheisler/cargo-criterion) and will be optional in a future version of Criterion.rs. To silence this warning, either switch to cargo-criterion or enable the 'html_reports' feature in your Cargo.toml.

Gnuplot not found, using plotters backend
Benchmarking serialize geojson::FeatureCollection struct (countries.geojson)
Benchmarking serialize geojson::FeatureCollection struct (countries.geojson): Warming up for 3.0000 s
Benchmarking serialize geojson::FeatureCollection struct (countries.geojson): Collecting 100 samples in estimated 5.0413 s (1700 iterations)
Benchmarking serialize geojson::FeatureCollection struct (countries.geojson): Analyzing
serialize geojson::FeatureCollection struct (countries.geojson)
                        time:   [2.9583 ms 2.9634 ms 2.9690 ms]
                        change: [-2.4235% -1.8933% -1.3582%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 5 outliers among 100 measurements (5.00%)
  1 (1.00%) high mild
  4 (4.00%) high severe

Benchmarking serialize custom struct (countries.geojson)
Benchmarking serialize custom struct (countries.geojson): Warming up for 3.0000 s
Benchmarking serialize custom struct (countries.geojson): Collecting 100 samples in estimated 5.1482 s (2300 iterations)
Benchmarking serialize custom struct (countries.geojson): Analyzing
serialize custom struct (countries.geojson)
                        time:   [2.2299 ms 2.2331 ms 2.2364 ms]
                        change: [-2.4677% -2.2513% -2.0310%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  2 (2.00%) high mild
  1 (1.00%) high severe

Benchmarking serialize custom struct to geo-types (countries.geojson)
Benchmarking serialize custom struct to geo-types (countries.geojson): Warming up for 3.0000 s
Benchmarking serialize custom struct to geo-types (countries.geojson): Collecting 100 samples in estimated 5.0427 s (2200 iterations)
Benchmarking serialize custom struct to geo-types (countries.geojson): Analyzing
serialize custom struct to geo-types (countries.geojson)
                        time:   [2.2959 ms 2.3017 ms 2.3081 ms]
                        change: [-11.260% -11.003% -10.713%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  7 (7.00%) high mild
  2 (2.00%) high severe

WARNING: HTML report generation will become a non-default optional feature in Criterion.rs 0.4.0.
This feature is being moved to cargo-criterion (https://github.com/bheisler/cargo-criterion) and will be optional in a future version of Criterion.rs. To silence this warning, either switch to cargo-criterion or enable the 'html_reports' feature in your Cargo.toml.

Gnuplot not found, using plotters backend
Benchmarking quick_collection
Benchmarking quick_collection: Warming up for 3.0000 s
Benchmarking quick_collection: Collecting 100 samples in estimated 5.1283 s (71k iterations)
Benchmarking quick_collection: Analyzing
quick_collection        time:   [72.840 µs 73.421 µs 74.013 µs]
                        change: [-79.655% -79.490% -79.317%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild
michaelkirk commented 1 year ago

See also: https://github.com/georust/geojson/issues/88, and in particular, my comment