NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
crates.io
crates.io, #1410 most downloaded on crates.io
An R*-tree spatial index
Latest stable
0.14.0
Last release 21 days ago
01 Nov 2026
Ships fairly regularly
a new release about every 5 months
Most releases are documented
notes for 17 of 26 stable releases
2 versions withdrawn
withdrawn after publishing
8 years old
28 releases · first in 2018
Drag across the ruler to compare, or tap a release to open it.
Provide a breaking release to get feedback on the trait changes w.r.t…
Provide a breaking release to get feedback on the trait changes w.r.t…
ordered-float crate. (PR)locate_in_envelope_int, locate_in_envelope_intersecting_int, locate_at_point_int, locate_all_at_point_int and their _mut variants) does not clone the envelope of every visited node anymore.locate_* iterators test parent nodes before pushing them onto their stack.remove, remove_at_point, remove_with_selection_function and pop_nearest_neighbor remove the element in place instead of taking the tree apart with a drain iterator, and do not allocate anymore.RTreeNum no longer has a blanket implementation, and no longer requires PartialOrd. It now requires
an associated OrdType: Ord plus an ord() conversion, so that every comparison the r-tree makes is total, avoiding
problems with NaN. Implementations have been added for all common numeric types as well as ordered_float::NotNan
(ordered_float is re-exported as rstar::ordered_float).
If you implement RTreeNum for your own scalar type, you'll need to add an implementation like:impl RTreeNum for MyScalar {
// Any `Ord` type will do; `OrderedFloat` is re-exported from the `ordered-float` crate.
type OrdType = rstar::ordered_float::OrderedFloat<f64>;
fn ord(self) -> Self::OrdType { rstar::ordered_float::OrderedFloat(self.0) }
}
(PR)RTree::bulk_load_parallel and RTree::bulk_load_with_params_parallel behind the new rayon feature build the same tree as bulk_load, with independent subtrees built in parallel on the current Rayon thread pool (#107).bulk_load building trees with leaves on different levels for some sizes, which made later insertions panic (#256). With the default parameters these are e.g. 25–27 and 97–111 elements in 2D, a regression in 0.13.0 (#221), and 49–54 elements in 3D. The partitioning is unchanged and the trees are the same as before for all other sizes; for these sizes, the deeper clusters become nodes with a few more than MAX_SIZE children instead.MIN_SIZE or more than MAX_SIZE children, and made the internal sanity_check accept such trees (#243).drain_within_distance panicking with an arithmetic overflow on an empty tree with integer coordinates. locate_within_distance was already guarded against this, its draining counterpart was not.NaN inputs, and a NaN coordinate no longer strands finite points.
(PR)Your coding agent can read these notes before it upgrades. Set up the MCP server →
Most importantly, this fixes a regression introduced in v0.13.0 where bulk loading produces unbalanced trees.
Most importantly, this fixes a regression introduced in v0.13.0
where bulk loading produces unbalanced trees.
Make a breaking release to get accumulated fixes and additions out to…
Make a breaking release to get accumulated fixes and additions out to…
nearest_neighbor_with_distance_2 and nearest_neighbors_with_distance_2 methods (PR).root doc examples and traversal docsRTreeObject for Arc<T> and Rc<T>Envelope::is_empty. (PR)AABB::from_center utility constructorAABB::from_bounds utility constructorbulk_load from Vec::split_off over-capacityRStar methods take Point and Envelope as owned values where it makes sense (PR)verify_parameters of rstar/src/params.rsFix for buggy 0.12.1 release
Fix for buggy 0.12.1 release
AABB::new_empty while still avoiding overflow panics applying selections on empty trees (PR)Provide selection methods based on internal iteration (PR)
Add optional support for the mint crate
IntoIter for RTree, i.e. added a owning iteratorHash for AABB, Line, and Rectangle, provided the Point used implements Hash itselfDefault for DefaultParamsDrainIterator::nextdistance_2 is not restricted to euclidean distanceheapless=0.8MSRV bump and bulk-loading perf improvements
MSRV bump and bulk-loading perf improvements
CachedEnvelope combinator which simplifies memoizing envelope computations. (PR)Point is now implemented as const generic for any length of RTreeNum arraygeo crate. (PR)Prepare for v0.10.0 release
Prepare for v0.10.0 release
RTree::drain().RTreeParams and added assert for positive MIN_SIZECopy bound from Point and Envelope. (PR)MSRV: 1.5.1, no-std support
Add RTree::drain_* methods to remove and drain selected items. (PR)
Added generic geomwithdata container
Added generic geomwithdata container
GeomWithData (PR)Update Criterion and rewrite deprecated benchmark functions
RTree::nearest_neighbors method based on
spade crate's implementationmin_max_dist_2 (PR).Point::generate function now accepts a impl FnMut. Custom implementations of Point must change to
accept impl FnMut instead of impl Fn. Callers of Point::generate should not require changes.Nothing published for this version
Update dependencies to remove heapless 0.5, which has a known vulnerability
Fixed a rare panic when calling insert (See #45)
insert (See #45)Fine tuned nearest neighbor iterator inline capacity (see #39). This should boost performance in some cases.
Deprecated RTree::nearest_neighbor_iter_with_distance. The name is misleading, use RTree::nearest_neighbor_iter_with_distance_2 instead.
RTree::locate_with_selection_function_mut sometimes returned too many elements for small trees.RTree::nearest_neighbor_iter_with_distance. The name is misleading, use RTree::nearest_neighbor_iter_with_distance_2 instead.nearest_neighbor_iter_with_distance_2 #31RTree::intersection_candidates_with_other_tree can now calculate intersections of trees of different item types (see #23)
RTree::intersection_candidates_with_other_tree can now calculate intersections of trees of different item types (see #23)RTree::remove_with_selection_function
RTree::remove_with_selection_functionRTree::pop_nearest_neighborNothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per quarter.