Merge pull request #125 from Dr-Emann/roaring64

Bindings CRoaring 3.0, including 64 bit bitmaps

croaring-sys/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "croaring-sys"
-version = "1.1.0"
+version = "2.0.0"
 edition = "2021"
 authors = ["croaring-rs developers"]
 license = "Apache-2.0"

croaring-sys/build.rs

@@ -1,5 +1,4 @@
 use std::env;
-use std::path::PathBuf;
 
 fn main() {
     println!("cargo:rerun-if-changed=CRoaring");

croaring/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "croaring"
-version = "1.0.1"
+version = "1.1.0"
 edition = "2021"
 authors = ["croaring-rs developers"]
 license = "Apache-2.0"
@@ -22,7 +22,7 @@ roaring = "0.10"
 criterion = { version = "0.5", features = ["html_reports"] }
 
 [dependencies]
-ffi = { package = "croaring-sys", path = "../croaring-sys", version = "1.1.0" }
+ffi = { package = "croaring-sys", path = "../croaring-sys", version = "2.0.0" }
 byteorder = "1.4.3"
 
 [[bench]]

croaring/benches/benches.rs

@@ -1,8 +1,9 @@
 use criterion::{
     black_box, criterion_group, criterion_main, BatchSize, BenchmarkId, Criterion, Throughput,
 };
+use std::ops::ControlFlow;
 
-use croaring::{Bitmap, Portable};
+use croaring::{Bitmap, Bitmap64, Portable};
 
 fn new(c: &mut Criterion) {
     c.bench_function("new", |b| b.iter(Bitmap::new));
@@ -130,10 +131,25 @@ fn flip(c: &mut Criterion) {
 }
 
 fn to_vec(c: &mut Criterion) {
-    c.bench_function("to_vec", |b| {
-        let bitmap = Bitmap::of(&[1, 2, 3]);
+    const N: usize = 100_000;
+    let bitmap: Bitmap = random_iter().take(N).collect();
+    let mut g = c.benchmark_group("collect");
+    g.bench_function("to_vec", |b| {
         b.iter(|| bitmap.to_vec());
     });
+    g.bench_function("via_iter", |b| {
+        b.iter(|| bitmap.iter().collect::<Vec<_>>());
+    });
+    g.bench_function("foreach", |b| {
+        b.iter(|| {
+            let mut vec = Vec::with_capacity(bitmap.cardinality() as usize);
+            bitmap.for_each(|item| -> ControlFlow<()> {
+                vec.push(item);
+                ControlFlow::Continue(())
+            });
+            vec
+        });
+    });
 }
 
 fn get_serialized_size_in_bytes(c: &mut Criterion) {
@@ -213,24 +229,35 @@ fn bulk_new(c: &mut Criterion) {
     group.finish();
 }
 
-fn random_iter(c: &mut Criterion) {
+#[derive(Clone)]
+struct RandomIter {
+    x: u32,
+}
+
+impl Iterator for RandomIter {
+    type Item = u32;
+
+    fn next(&mut self) -> Option<u32> {
+        const MULTIPLIER: u32 = 742938285;
+        const MODULUS: u32 = (1 << 31) - 1;
+        self.x = (MULTIPLIER.wrapping_mul(self.x)) % MODULUS;
+        Some(self.x)
+    }
+}
+
+fn random_iter() -> RandomIter {
+    RandomIter { x: 20170705 }
+}
+
+fn create_random(c: &mut Criterion) {
     const N: u32 = 5_000;
     // Clamp values so we get some re-use of containers
     const MAX: u32 = 8 * (u16::MAX as u32 + 1);
 
     let mut group = c.benchmark_group("random_iter");
     group.throughput(Throughput::Elements(N.into()));
 
-    let rand_iter = {
-        const MULTIPLIER: u32 = 742938285;
-        const MODULUS: u32 = (1 << 31) - 1;
-        // Super simple LCG iterator
-        let mut z = 20170705; // seed
-        std::iter::from_fn(move || {
-            z = (MULTIPLIER * z) % MODULUS;
-            Some(z % MAX)
-        })
-    };
+    let rand_iter = random_iter();
 
     group.bench_function("random_adds", |b| {
         b.iter(|| {
@@ -252,6 +279,96 @@ fn random_iter(c: &mut Criterion) {
     });
 }
 
+fn collect_bitmap64_to_vec(c: &mut Criterion) {
+    const N: u64 = 1_000_000;
+
+    let mut group = c.benchmark_group("collect_bitmap64_to_vec");
+    group.throughput(Throughput::Elements(N.into()));
+    let bitmap = Bitmap64::from_range(0..N);
+    group.bench_function("to_vec", |b| {
+        b.iter_batched(|| (), |()| bitmap.to_vec(), BatchSize::LargeInput);
+    });
+    group.bench_function("foreach", |b| {
+        b.iter_batched(
+            || (),
+            |()| {
+                let mut vec = Vec::with_capacity(bitmap.cardinality() as usize);
+                bitmap.for_each(|item| -> ControlFlow<()> {
+                    vec.push(item);
+                    ControlFlow::Continue(())
+                });
+                vec
+            },
+            BatchSize::LargeInput,
+        );
+    });
+    group.bench_function("iter", |b| {
+        b.iter_batched(
+            || (),
+            |()| {
+                let mut vec = Vec::with_capacity(bitmap.cardinality() as usize);
+                vec.extend(bitmap.iter());
+                vec
+            },
+            BatchSize::LargeInput,
+        );
+    });
+    group.bench_function("iter_many", |b| {
+        b.iter_batched(
+            || (),
+            |()| {
+                let mut vec = vec![0; bitmap.cardinality() as usize];
+                let mut iter = bitmap.cursor();
+                assert_eq!(iter.read_many(&mut vec), vec.len());
+                vec
+            },
+            BatchSize::LargeInput,
+        );
+    });
+
+    group.finish();
+}
+
+fn iterate_bitmap64(c: &mut Criterion) {
+    const N: u64 = 1_000_000;
+    const END_ITER: u64 = N - 100;
+
+    let mut group = c.benchmark_group("bitmap64_iterate");
+    group.throughput(Throughput::Elements(N.into()));
+    let bitmap = Bitmap64::from_range(0..N);
+    group.bench_function("iter", |b| {
+        b.iter(|| {
+            for x in bitmap.iter() {
+                if x == END_ITER {
+                    break;
+                }
+            }
+        })
+    });
+    group.bench_function("cursor", |b| {
+        b.iter(|| {
+            let mut cursor = bitmap.cursor();
+            while let Some(x) = cursor.next() {
+                if x == END_ITER {
+                    break;
+                }
+            }
+        })
+    });
+    group.bench_function("for_each", |b| {
+        b.iter(|| {
+            bitmap.for_each(|x| -> ControlFlow<()> {
+                if x == END_ITER {
+                    return ControlFlow::Break(());
+                }
+                ControlFlow::Continue(())
+            })
+        })
+    });
+
+    group.finish();
+}
+
 criterion_group!(
     benches,
     new,
@@ -269,6 +386,8 @@ criterion_group!(
     serialize,
     deserialize,
     bulk_new,
-    random_iter,
+    create_random,
+    collect_bitmap64_to_vec,
+    iterate_bitmap64,
 );
 criterion_main!(benches);

croaring/src/bitmap/imp.rs

@@ -1,9 +1,9 @@
+use crate::callback::CallbackWrapper;
 use crate::Bitset;
 use ffi::roaring_bitmap_t;
-use std::convert::TryInto;
 use std::ffi::{c_void, CStr};
-use std::ops::{Bound, RangeBounds};
-use std::{mem, ptr};
+use std::ops::{Bound, ControlFlow, RangeBounds};
+use std::{mem, panic, ptr};
 
 use super::serialization::{Deserializer, Serializer};
 use super::{Bitmap, Statistics};
@@ -21,7 +21,7 @@ impl Bitmap {
         // (it can be moved safely), and can be freed with `free`, without freeing the underlying
         // containers and auxiliary data. Ensure this is still valid every time we update
         // the version of croaring.
-        const _: () = assert!(ffi::ROARING_VERSION_MAJOR == 2 && ffi::ROARING_VERSION_MINOR == 0);
+        const _: () = assert!(ffi::ROARING_VERSION_MAJOR == 3 && ffi::ROARING_VERSION_MINOR == 0);
         ffi::roaring_free(p.cast::<c_void>());
         result
     }
@@ -279,6 +279,29 @@ impl Bitmap {
         unsafe { ffi::roaring_bitmap_remove_checked(&mut self.bitmap, element) }
     }
 
+    /// Remove many values from the bitmap
+    ///
+    /// This should be faster than calling `remove` multiple times.
+    ///
+    /// In order to exploit this optimization, the caller should attempt to keep values with the same high 48 bits of
+    /// the value as consecutive elements in `vals`
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use croaring::Bitmap;
+    /// let mut bitmap = Bitmap::of(&[1, 2, 3, 4, 5, 6, 7, 8, 9]);
+    /// bitmap.remove_many(&[1, 2, 3, 4, 5, 6, 7, 8]);
+    /// assert_eq!(bitmap.to_vec(), vec![9]);
+    /// ```
+    #[inline]
+    #[doc(alias = "roaring_bitmap_remove_many")]
+    pub fn remove_many(&mut self, elements: &[u32]) {
+        unsafe {
+            ffi::roaring_bitmap_remove_many(&mut self.bitmap, elements.len(), elements.as_ptr())
+        }
+    }
+
     /// Contains returns true if the integer element is contained in the bitmap
     ///
     /// # Examples
@@ -721,6 +744,49 @@ impl Bitmap {
         unsafe { ffi::roaring_bitmap_flip_inplace(&mut self.bitmap, start, end) }
     }
 
+    /// Iterate over the values in the bitmap in sorted order
+    ///
+    /// If `f` returns `Break`, iteration will stop and the value will be returned,
+    /// Otherwise, iteration continues. If `f` never returns break, `None` is returned after all values are visited.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use croaring::Bitmap;
+    /// use std::ops::ControlFlow;
+    ///
+    /// let bitmap = Bitmap::of(&[1, 2, 3, 14, 20, 21, 100]);
+    /// let mut even_nums_under_50 = vec![];
+    ///
+    /// let first_over_50 = bitmap.for_each(|value| {
+    ///     if value > 50 {
+    ///        return ControlFlow::Break(value);
+    ///     }
+    ///     if value % 2 == 0 {
+    ///         even_nums_under_50.push(value);
+    ///     }
+    ///     ControlFlow::Continue(())
+    /// });
+    ///
+    /// assert_eq!(even_nums_under_50, vec![2, 14, 20]);
+    /// assert_eq!(first_over_50, ControlFlow::Break(100));
+    /// ```
+    #[inline]
+    pub fn for_each<F, O>(&self, f: F) -> ControlFlow<O>
+    where
+        F: FnMut(u32) -> ControlFlow<O>,
+    {
+        let mut callback_wrapper = CallbackWrapper::new(f);
+        let (callback, context) = callback_wrapper.callback_and_ctx();
+        unsafe {
+            ffi::roaring_iterate(&self.bitmap, Some(callback), context);
+        }
+        match callback_wrapper.result() {
+            Ok(cf) => cf,
+            Err(e) => panic::resume_unwind(e),
+        }
+    }
+
     /// Returns a vector containing all of the integers stored in the Bitmap
     /// in sorted order.
     ///
@@ -922,6 +988,13 @@ impl Bitmap {
     /// // Exclusive ranges still step from the start, but do not include it
     /// let bitmap = Bitmap::from_range_with_step((Bound::Excluded(10), Bound::Included(30)), 10);
     /// assert_eq!(bitmap.to_vec(), [20, 30]);
+    ///
+    /// // Ranges including max value
+    /// let bitmap = Bitmap::from_range_with_step((u32::MAX - 1)..=u32::MAX, 1);
+    /// assert_eq!(bitmap.to_vec(), vec![u32::MAX - 1, u32::MAX]);
+    ///
+    /// let bitmap = Bitmap::from_range_with_step((u32::MAX - 1)..=u32::MAX, 3);
+    /// assert_eq!(bitmap.to_vec(), vec![u32::MAX - 1]);
     /// ```
     #[inline]
     #[doc(alias = "roaring_bitmap_from_range")]