Releases: RoaringBitmap/roaring-rs
v0.10.9
What's Changed
- fix(no_std): use
core::slice
inbitmap_store.rs
by @GZTimeWalker in #305
Full Changelog: v0.10.8...v0.10.9
v0.10.8
What's Changed
- Feature Request:
RoaringBitmap::from_lsb0_bytes
by @lemolatoon in #288
New Contributors
- @lemolatoon made their first contribution in #288
Full Changelog: v0.10.7...v0.10.8
Version 0.10.7
What's Changed
- Introduce cargo workspaces by @christianschleifer in #285
- Fix warnings when testing with nightly by @Dr-Emann in #293
- fix(typos): fix some typos with
typos
by @GZTimeWalker in #291 - Optimize BitmapIter::next by @Dr-Emann in #292
- Lazily compute Iterator len by @Dr-Emann in #295
- feat: implement advance_to and advance_back_to for Iter by @Dr-Emann in #296
- Implement Clone on the bitmap::Iter type by @Kerollmops in #289
- implement range-based iteration by @Dr-Emann in #297
New Contributors
- @christianschleifer made their first contribution in #285
Version 0.10.6
Version 0.10.5
What's Changed
- Resolve issues with no std by @kirk-baird in #279
- add dedicated fold and rfold impls for iterators by @oskgo in #280
- Direct intersections with serialized RoaringBitmaps by @Kerollmops in #281
New Contributors
Version 0.10.4
What's Changed
- Fix potential overflow in deserialization of Bitmap by @kirk-baird in #271
New Contributors
- @kirk-baird made their first contribution in #271
Full Changelog: v0.10.3...v0.10.4
Version 0.10.3
What's Changed
- Make nightly clippy happy by @Kerollmops in #261
- deps: remove retain_mut by @grim7reaper in #259
- Bump the MSRV to 1.66 by @Kerollmops in #267
- feat: support no_std by @GZTimeWalker in #266
- docs:
RoaringTreemap::select
returnsNone
ifn >= len()
by @shekhirin in #264 - Bump version to 0.10.3 by @Kerollmops in #268
Full Changelog: v0.10.2...v0.10.3
Version 0.10.2
Introduce the RoaringBitmap::remove_smallest/biggests
methods
Thanks to @shimatar0, who added those two methods. They remove the n
smallest or biggest values from a RoaringBitmap
. If n
is bigger than the cardinality of the bitmap, it clears it.
let mut rb = RoaringBitmap::from_iter([1, 5, 7, 9, 10, 15, 45]);
rb.remove_biggest(2);
assert_eq!(rb, RoaringBitmap::from_iter([1, 5, 7, 9, 10]));
rb.remove_smallest(1);
assert_eq!(rb, RoaringBitmap::from_iter([5, 7, 9, 10]));
RoaringBitmaps/Treemap
s support run containers when deserializing
Thanks in part to @josephglanville for the original deserialization code. Bitmaps and treemaps can deserialize run containers which are useful when those come from other libraries from other languages.
Note that this library supports run container operations and only converts the run containers into array or bitmap containers depending on the cardinality and will, therefore, not serialize containers into run ones either.
Create RoaringBitmap
or RoaringTreemap
from an array
Thanks to @michaelmior, you can use the From::from
trait method to construct a bitmap from an array. Building them using the FromIterator::from_iter
trait method was already possible. It is just even from convenient now.
let bitmap = RoaringBitmap::from([10]);
let treemap = RoaringTreemap::from([1,2,3]);
Bumped the MSRV to 1.65
While doing crate maintenance, I had to update different crates and bump the minimal supported Rust version to 1.65, which was released eight months ago.
Version 0.10.1
This release is a patch in which @jonasspinner fixed the MultiOps
union for Result
iterators which was wrongly using the intersection operations.
Version 0.10 - Introducing faster multiple-bitmaps operations 🧨
In this new v0.10 release, we introduce significant improvements for multiple bitmap operations and many new methods for range operations. You can find the documentation on docs.rs. The only reason it is a breaking release is due to the simd
feature that was updated to match the new types of the std::simd
nightly module.
About the new MultiOps
trait
@irevoire based this trait on @saik0's previous work. It brings a new fresh API to help in executing different set operations on a set of RoaringBitmap/Treemap
. With up to 37x improvements on intersection operations and 7x on unions, it clearly makes this PR worth releasing. If you want more numbers, you can look at some benchmarks in the original PR.
The MultiOps
trait is implemented for anything that implements IntoIterator
where Item
is a RoaringBitmap/Treemap
, but there also are try_
versions of it when your Item
is a Result
. It can be very useful when the sets are lazily deserialized, for example.
use roaring::{MultiOps, RoaringBitmap};
let bitmaps = [
RoaringBitmap::from_iter(0..10),
RoaringBitmap::from_iter(10..20),
RoaringBitmap::from_iter(20..30),
];
// Stop doing this
let naive = bitmaps.clone().into_iter().reduce(|a, b| a | b).unwrap_or_default();
// And start doing this instead. It will be much faster!
let iter = bitmaps.union();
assert_eq!(naive, iter);
Improve most of the intersection operations
@RaduBerinde made a huge improvement to the whole set of intersection operations by implementing a better version of the retain method that is now branchless. Thank you very much! This PR does between 1.03x and 1.37x improvement, sometimes slowing down the performances but for the good of all the other benchmarks.
Introduce more range-related methods
While @Dr-Emann was introducing the new contains_range
and range_cardinality
methods on the RoaringBitmap
type, @not-jan was working on the RoaringTreemap::insert_range
method.
use roaring::{RoaringTreemap, RoaringBitmap};
let mut rb = RoaringTreemap::new();
rb.insert_range(2..4);
assert!(rb.contains(2));
assert!(rb.contains(3));
assert!(!rb.contains(4));
let mut rb = RoaringBitmap::new();
rb.insert_range(2..45_000);
assert!(rb.contains_range(2..30));
assert!(!rb.contains_range(0..3));
assert!(!rb.contains_range(6_000..46_001));
assert_eq!(rb.range_cardinality(6_000..45_001), 39_000);
More methods for full sets
While I (@Kerollmops) was reviewing the work on the RoaringBitmap::insert_range
I also introduced more methods for full RoaringBitmap
and RoaringTreemap
. You can now create full bitmaps and treemaps.
use roaring::{RoaringTreemap, RoaringBitmap};
let mut rt = RoaringTreemap::full();
assert!(rt.is_full());
assert!(rt.contains(42));
let mut rb = RoaringBitmap::full();
assert!(rb.is_full());
assert!(!rt.is_empty());
Serde is in the place 🎤
@irevoire didn't only work on the MultiOps
trait but also introduced the support of serde! You can use enable it by specifying the serde
feature and then convert your RoaringBitmap
and RoaringTreemap
into any format supported by serde and back into a bitmap.
use roaring::{RoaringTreemap, RoaringBitmap};
let original = RoaringTreemap::from_range(6..42);
let json = serde_json::to_vec(&original).unwrap();
let output = serde_json::from_slice(&json).unwrap();
assert_eq!(original, output);
let original = RoaringBitmap::from_range(16..142);
let buffer = bincode::serialize(&bitmap).unwrap();
let output = bincode::deserialize(&buffer).unwrap();
assert_eq!(bitmap, output);