Prepare hashbrown for inclusion in the standard library

Author: Amanieu

Cargo.toml

@@ -11,13 +11,15 @@ categories = ["data-structures", "no-std"]
 exclude = [".travis.yml", "bors.toml"]
 
 [dependencies]
-byteorder = { version = "1.0", default-features = false }
-scopeguard = { version = "0.3", default-features = false }
-
 # For external trait impls
 rayon = { version = "1.0", optional = true }
 serde = { version = "1.0.25", default-features = false, optional = true }
 
+# When built as part of libstd
+core = { version = "1.0.0", optional = true, package = "rustc-std-workspace-core" }
+compiler_builtins = { version = "0.1.2", optional = true }
+alloc = { version = "1.0.0", optional = true, package = "rustc-std-workspace-alloc" }
+
 [dev-dependencies]
 lazy_static = "~1.2"
 rand = "0.5.1"
@@ -28,3 +30,4 @@ serde_test = "1.0"
 [features]
 default = []
 nightly = []
+rustc-dep-of-std = ["nightly", "core", "compiler_builtins", "alloc"]

src/fx.rs

@@ -5,8 +5,6 @@ use core::hash::{BuildHasherDefault, Hasher};
 use core::mem::size_of;
 use core::ops::BitXor;
 
-use byteorder::{ByteOrder, NativeEndian};
-
 /// Type alias for a `HashBuilder` using the `fx` hash algorithm.
 pub type FxHashBuilder = BuildHasherDefault<FxHasher>;
 
@@ -47,25 +45,29 @@ impl FxHasher {
 impl Hasher for FxHasher {
     #[inline]
     fn write(&mut self, mut bytes: &[u8]) {
-        #[cfg(target_pointer_width = "32")]
-        #[allow(clippy::cast_possible_truncation)]
-        let read_usize = |bytes| NativeEndian::read_u32(bytes) as usize;
-        #[cfg(target_pointer_width = "64")]
-        #[allow(clippy::cast_possible_truncation)]
-        let read_usize = |bytes| NativeEndian::read_u64(bytes) as usize;
+        macro_rules! read_bytes {
+            ($ty:ty, $src:expr) => {{
+                assert!(size_of::<$ty>() <= $src.len());
+                let mut data: $ty = 0;
+                unsafe {
+                    $src.as_ptr().copy_to_nonoverlapping(&mut data as *mut $ty as *mut u8, size_of::<$ty>());
+                }
+                data
+            }};
+        }
 
         let mut hash = Self { hash: self.hash };
         assert!(size_of::<usize>() <= 8);
         while bytes.len() >= size_of::<usize>() {
-            hash.add_to_hash(read_usize(bytes));
+            hash.add_to_hash(read_bytes!(usize, bytes) as usize);
             bytes = &bytes[size_of::<usize>()..];
         }
         if (size_of::<usize>() > 4) && (bytes.len() >= 4) {
-            hash.add_to_hash(NativeEndian::read_u32(bytes) as usize);
+            hash.add_to_hash(read_bytes!(u32, bytes) as usize);
             bytes = &bytes[4..];
         }
         if (size_of::<usize>() > 2) && bytes.len() >= 2 {
-            hash.add_to_hash(NativeEndian::read_u16(bytes) as usize);
+            hash.add_to_hash(read_bytes!(u16, bytes) as usize);
             bytes = &bytes[2..];
         }
         if (size_of::<usize>() > 1) && !bytes.is_empty() {

src/lib.rs

@@ -26,19 +26,16 @@
 #![allow(clippy::module_name_repetitions)]
 
 #[cfg(test)]
-#[macro_use]
-#[allow(unused_imports)]
+#[cfg_attr(feature = "rayon", macro_use)]
 extern crate std;
 #[cfg(test)]
 extern crate rand;
 
 #[cfg(feature = "nightly")]
 #[cfg_attr(test, macro_use)]
 extern crate alloc;
-extern crate byteorder;
 #[cfg(feature = "rayon")]
 extern crate rayon;
-extern crate scopeguard;
 #[cfg(feature = "serde")]
 extern crate serde;
 #[cfg(not(feature = "nightly"))]
@@ -53,11 +50,16 @@ mod fx;
 mod map;
 mod raw;
 mod set;
+#[cfg(feature = "rustc-dep-of-std")]
+mod rustc_entry;
 
 pub mod hash_map {
     //! A hash map implemented with quadratic probing and SIMD lookup.
     pub use map::*;
 
+    #[cfg(feature = "rustc-dep-of-std")]
+    pub use rustc_entry::*;
+
     #[cfg(feature = "rayon")]
     /// [rayon]-based parallel iterator types for hash maps.
     /// You will rarely need to interact with it directly unless you have need

src/map.rs

@@ -150,7 +150,7 @@ pub use fx::FxHashBuilder as DefaultHashBuilder;
 /// }
 ///
 /// impl Viking {
-///     /// Create a new Viking.
+///     /// Creates a new Viking.
 ///     fn new(name: &str, country: &str) -> Viking {
 ///         Viking { name: name.to_string(), country: country.to_string() }
 ///     }
@@ -186,12 +186,12 @@ pub use fx::FxHashBuilder as DefaultHashBuilder;
 
 #[derive(Clone)]
 pub struct HashMap<K, V, S = DefaultHashBuilder> {
-    hash_builder: S,
+    pub(crate) hash_builder: S,
     pub(crate) table: RawTable<(K, V)>,
 }
 
 #[inline]
-fn make_hash<K: Hash + ?Sized>(hash_builder: &impl BuildHasher, val: &K) -> u64 {
+pub(crate) fn make_hash<K: Hash + ?Sized>(hash_builder: &impl BuildHasher, val: &K) -> u64 {
     let mut state = hash_builder.build_hasher();
     val.hash(&mut state);
     state.finish()
@@ -232,8 +232,6 @@ impl<K, V> HashMap<K, V, DefaultHashBuilder> {
 }
 
 impl<K, V, S> HashMap<K, V, S>
-where
-    S: BuildHasher,
 {
     /// Creates an empty `HashMap` which will use the given hash builder to hash
     /// keys.
@@ -486,7 +484,7 @@ where
         self.table.len()
     }
 
-    /// Returns true if the map contains no elements.
+    /// Returns `true` if the map contains no elements.
     ///
     /// # Examples
     ///
@@ -757,7 +755,7 @@ where
             })
     }
 
-    /// Returns true if the map contains a value for the specified key.
+    /// Returns `true` if the map contains a value for the specified key.
     ///
     /// The key may be any borrowed form of the map's key type, but
     /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for
@@ -1220,7 +1218,7 @@ impl<'a, K, V: Debug> fmt::Debug for Values<'a, K, V> {
 /// [`drain`]: struct.HashMap.html#method.drain
 /// [`HashMap`]: struct.HashMap.html
 pub struct Drain<'a, K: 'a, V: 'a> {
-    pub(super) inner: RawDrain<'a, (K, V)>,
+    inner: RawDrain<'a, (K, V)>,
 }
 
 impl<'a, K, V> Drain<'a, K, V> {
@@ -1300,9 +1298,8 @@ pub struct RawEntryBuilder<'a, K: 'a, V: 'a, S: 'a> {
 impl<'a, K, V, S> RawEntryBuilderMut<'a, K, V, S>
 where
     S: BuildHasher,
-    K: Eq + Hash,
 {
-    /// Create a `RawEntryMut` from the given key.
+    /// Creates a `RawEntryMut` from the given key.
     #[inline]
     #[allow(clippy::wrong_self_convention)]
     pub fn from_key<Q: ?Sized>(self, k: &Q) -> RawEntryMut<'a, K, V, S>
@@ -1315,7 +1312,7 @@ where
         self.from_key_hashed_nocheck(hasher.finish(), k)
     }
 
-    /// Create a `RawEntryMut` from the given key and its hash.
+    /// Creates a `RawEntryMut` from the given key and its hash.
     #[inline]
     #[allow(clippy::wrong_self_convention)]
     pub fn from_key_hashed_nocheck<Q: ?Sized>(self, hash: u64, k: &Q) -> RawEntryMut<'a, K, V, S>
@@ -1331,7 +1328,7 @@ impl<'a, K, V, S> RawEntryBuilderMut<'a, K, V, S>
 where
     S: BuildHasher,
 {
-    /// Create a `RawEntryMut` from the given hash.
+    /// Creates a `RawEntryMut` from the given hash.
     #[inline]
     #[allow(clippy::wrong_self_convention)]
     pub fn from_hash<F>(self, hash: u64, is_match: F) -> RawEntryMut<'a, K, V, S>
@@ -1706,7 +1703,7 @@ pub enum Entry<'a, K: 'a, V: 'a, S: 'a> {
     Vacant(VacantEntry<'a, K, V, S>),
 }
 
-impl<'a, K: 'a + Debug + Eq + Hash, V: 'a + Debug, S: 'a> Debug for Entry<'a, K, V, S> {
+impl<'a, K: 'a + Debug, V: 'a + Debug, S: 'a> Debug for Entry<'a, K, V, S> {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         match *self {
             Entry::Vacant(ref v) => f.debug_tuple("Entry").field(v).finish(),
@@ -1759,17 +1756,13 @@ pub struct VacantEntry<'a, K: 'a, V: 'a, S: 'a> {
     table: &'a mut HashMap<K, V, S>,
 }
 
-impl<'a, K: 'a + Debug + Eq + Hash, V: 'a, S> Debug for VacantEntry<'a, K, V, S> {
+impl<'a, K: 'a + Debug, V: 'a, S> Debug for VacantEntry<'a, K, V, S> {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         f.debug_tuple("VacantEntry").field(self.key()).finish()
     }
 }
 
-impl<'a, K, V, S> IntoIterator for &'a HashMap<K, V, S>
-where
-    K: Eq + Hash,
-    S: BuildHasher,
-{
+impl<'a, K, V, S> IntoIterator for &'a HashMap<K, V, S> {
     type Item = (&'a K, &'a V);
     type IntoIter = Iter<'a, K, V>;
 
@@ -1779,11 +1772,7 @@ where
     }
 }
 
-impl<'a, K, V, S> IntoIterator for &'a mut HashMap<K, V, S>
-where
-    K: Eq + Hash,
-    S: BuildHasher,
-{
+impl<'a, K, V, S> IntoIterator for &'a mut HashMap<K, V, S> {
     type Item = (&'a K, &'a mut V);
     type IntoIter = IterMut<'a, K, V>;
 
@@ -1793,11 +1782,7 @@ where
     }
 }
 
-impl<K, V, S> IntoIterator for HashMap<K, V, S>
-where
-    K: Eq + Hash,
-    S: BuildHasher,
-{
+impl<K, V, S> IntoIterator for HashMap<K, V, S> {
     type Item = (K, V);
     type IntoIter = IntoIter<K, V>;
 

src/raw/mod.rs

@@ -7,7 +7,6 @@ use core::mem;
 use core::mem::ManuallyDrop;
 use core::ops::Range;
 use core::ptr::NonNull;
-use scopeguard::guard;
 use CollectionAllocErr;
 
 // Branch prediction hint. This is currently only available on nightly but it
@@ -60,6 +59,9 @@ cfg_if! {
 
 mod bitmask;
 
+mod scopeguard;
+
+use self::scopeguard::guard;
 use self::bitmask::BitMask;
 use self::imp::Group;
 
@@ -714,8 +716,10 @@ impl<T> RawTable<T> {
                 // This may panic.
                 let hash = hasher(item.as_ref());
 
-                // We can use a simpler version of insert() here since there are no
-                // DELETED entries.
+                // We can use a simpler version of insert() here since:
+                // - there are no DELETED entries.
+                // - we know there is enough space in the table.
+                // - all elements are unique.
                 let index = new_table.find_insert_slot(hash);
                 new_table.set_ctrl(index, h2(hash));
                 new_table.bucket(index).write(item.read());
@@ -737,7 +741,16 @@ impl<T> RawTable<T> {
     #[inline]
     pub fn insert(&mut self, hash: u64, value: T, hasher: impl Fn(&T) -> u64) -> Bucket<T> {
         self.reserve(1, hasher);
+        self.insert_no_grow(hash, value)
+    }
 
+    /// Inserts a new element into the table, without growing the table.
+    ///
+    /// There must be enough space in the table to insert the new element.
+    ///
+    /// This does not check if the given element already exists in the table.
+    #[inline]
+    pub fn insert_no_grow(&mut self, hash: u64, value: T) -> Bucket<T> {
         unsafe {
             let index = self.find_insert_slot(hash);
             let bucket = self.bucket(index);
@@ -941,7 +954,7 @@ impl<T> IntoIterator for RawTable<T> {
     }
 }
 
-/// Iterator over a a sub-range of a table. Unlike `RawIter` this iterator does
+/// Iterator over a sub-range of a table. Unlike `RawIter` this iterator does
 /// not track an item count.
 pub struct RawIterRange<T> {
     // Using *const here for covariance

src/raw/scopeguard.rs

@@ -0,0 +1,49 @@
+// Extracted from the scopeguard crate
+use core::ops::{Deref, DerefMut};
+
+pub struct ScopeGuard<T, F>
+where
+    F: FnMut(&mut T),
+{
+    dropfn: F,
+    value: T,
+}
+
+#[inline]
+pub fn guard<T, F>(value: T, dropfn: F) -> ScopeGuard<T, F>
+where
+    F: FnMut(&mut T),
+{
+    ScopeGuard { dropfn, value }
+}
+
+impl<T, F> Deref for ScopeGuard<T, F>
+where
+    F: FnMut(&mut T),
+{
+    type Target = T;
+    #[inline]
+    fn deref(&self) -> &T {
+        &self.value
+    }
+}
+
+impl<T, F> DerefMut for ScopeGuard<T, F>
+where
+    F: FnMut(&mut T),
+{
+    #[inline]
+    fn deref_mut(&mut self) -> &mut T {
+        &mut self.value
+    }
+}
+
+impl<T, F> Drop for ScopeGuard<T, F>
+where
+    F: FnMut(&mut T),
+{
+    #[inline]
+    fn drop(&mut self) {
+        (self.dropfn)(&mut self.value)
+    }
+}