| 1 | //! This crate provides foldhash, a fast, non-cryptographic, minimally |
| 2 | //! DoS-resistant hashing algorithm designed for computational uses such as |
| 3 | //! hashmaps, bloom filters, count sketching, etc. |
| 4 | //! |
| 5 | //! When should you **not** use foldhash: |
| 6 | //! |
| 7 | //! - You are afraid of people studying your long-running program's behavior |
| 8 | //! to reverse engineer its internal random state and using this knowledge to |
| 9 | //! create many colliding inputs for computational complexity attacks. |
| 10 | //! |
| 11 | //! - You expect foldhash to have a consistent output across versions or |
| 12 | //! platforms, such as for persistent file formats or communication protocols. |
| 13 | //! |
| 14 | //! - You are relying on foldhash's properties for any kind of security. |
| 15 | //! Foldhash is **not appropriate for any cryptographic purpose**. |
| 16 | //! |
| 17 | //! Foldhash has two variants, one optimized for speed which is ideal for data |
| 18 | //! structures such as hash maps and bloom filters, and one optimized for |
| 19 | //! statistical quality which is ideal for algorithms such as |
| 20 | //! [HyperLogLog](https://en.wikipedia.org/wiki/HyperLogLog) and |
| 21 | //! [MinHash](https://en.wikipedia.org/wiki/MinHash). |
| 22 | //! |
| 23 | //! Foldhash can be used in a `#![no_std]` environment by disabling its default |
| 24 | //! `"std"` feature. |
| 25 | //! |
| 26 | //! # Usage |
| 27 | //! |
| 28 | //! The easiest way to use this crate with the standard library [`HashMap`] or |
| 29 | //! [`HashSet`] is to import them from `foldhash` instead, along with the |
| 30 | //! extension traits to make [`HashMap::new`] and [`HashMap::with_capacity`] |
| 31 | //! work out-of-the-box: |
| 32 | //! |
| 33 | //! ```rust |
| 34 | //! use foldhash::{HashMap, HashMapExt}; |
| 35 | //! |
| 36 | //! let mut hm = HashMap::new(); |
| 37 | //! hm.insert(42, "hello" ); |
| 38 | //! ``` |
| 39 | //! |
| 40 | //! You can also avoid the convenience types and do it manually by initializing |
| 41 | //! a [`RandomState`](fast::RandomState), for example if you are using a different hash map |
| 42 | //! implementation like [`hashbrown`](https://docs.rs/hashbrown/): |
| 43 | //! |
| 44 | //! ```rust |
| 45 | //! use hashbrown::HashMap; |
| 46 | //! use foldhash::fast::RandomState; |
| 47 | //! |
| 48 | //! let mut hm = HashMap::with_hasher(RandomState::default()); |
| 49 | //! hm.insert("foo" , "bar" ); |
| 50 | //! ``` |
| 51 | //! |
| 52 | //! The above methods are the recommended way to use foldhash, which will |
| 53 | //! automatically generate a randomly generated hasher instance for you. If you |
| 54 | //! absolutely must have determinism you can use [`FixedState`](fast::FixedState) |
| 55 | //! instead, but note that this makes you trivially vulnerable to HashDoS |
| 56 | //! attacks and might lead to quadratic runtime when moving data from one |
| 57 | //! hashmap/set into another: |
| 58 | //! |
| 59 | //! ```rust |
| 60 | //! use std::collections::HashSet; |
| 61 | //! use foldhash::fast::FixedState; |
| 62 | //! |
| 63 | //! let mut hm = HashSet::with_hasher(FixedState::with_seed(42)); |
| 64 | //! hm.insert([1, 10, 100]); |
| 65 | //! ``` |
| 66 | //! |
| 67 | //! If you rely on statistical properties of the hash for the correctness of |
| 68 | //! your algorithm, such as in [HyperLogLog](https://en.wikipedia.org/wiki/HyperLogLog), |
| 69 | //! it is suggested to use the [`RandomState`](quality::RandomState) |
| 70 | //! or [`FixedState`](quality::FixedState) from the [`quality`] module instead |
| 71 | //! of the [`fast`] module. The latter is optimized purely for speed in hash |
| 72 | //! tables and has known statistical imperfections. |
| 73 | //! |
| 74 | //! Finally, you can also directly use the [`RandomState`](quality::RandomState) |
| 75 | //! or [`FixedState`](quality::FixedState) to manually hash items using the |
| 76 | //! [`BuildHasher`](std::hash::BuildHasher) trait: |
| 77 | //! ```rust |
| 78 | //! use std::hash::BuildHasher; |
| 79 | //! use foldhash::quality::RandomState; |
| 80 | //! |
| 81 | //! let random_state = RandomState::default(); |
| 82 | //! let hash = random_state.hash_one("hello world" ); |
| 83 | //! ``` |
| 84 | |
| 85 | #![cfg_attr (all(not(test), not(feature = "std" )), no_std)] |
| 86 | #![warn (missing_docs)] |
| 87 | |
| 88 | use core::hash::Hasher; |
| 89 | |
| 90 | #[cfg (feature = "std" )] |
| 91 | mod convenience; |
| 92 | mod seed; |
| 93 | |
| 94 | #[cfg (feature = "std" )] |
| 95 | pub use convenience::*; |
| 96 | |
| 97 | // Arbitrary constants with high entropy. Hexadecimal digits of pi were used. |
| 98 | const ARBITRARY0: u64 = 0x243f6a8885a308d3; |
| 99 | const ARBITRARY1: u64 = 0x13198a2e03707344; |
| 100 | const ARBITRARY2: u64 = 0xa4093822299f31d0; |
| 101 | const ARBITRARY3: u64 = 0x082efa98ec4e6c89; |
| 102 | const ARBITRARY4: u64 = 0x452821e638d01377; |
| 103 | const ARBITRARY5: u64 = 0xbe5466cf34e90c6c; |
| 104 | const ARBITRARY6: u64 = 0xc0ac29b7c97c50dd; |
| 105 | const ARBITRARY7: u64 = 0x3f84d5b5b5470917; |
| 106 | const ARBITRARY8: u64 = 0x9216d5d98979fb1b; |
| 107 | const ARBITRARY9: u64 = 0xd1310ba698dfb5ac; |
| 108 | |
| 109 | #[inline (always)] |
| 110 | const fn folded_multiply(x: u64, y: u64) -> u64 { |
| 111 | #[cfg (target_pointer_width = "64" )] |
| 112 | { |
| 113 | // We compute the full u64 x u64 -> u128 product, this is a single mul |
| 114 | // instruction on x86-64, one mul plus one mulhi on ARM64. |
| 115 | let full = (x as u128) * (y as u128); |
| 116 | let lo = full as u64; |
| 117 | let hi = (full >> 64) as u64; |
| 118 | |
| 119 | // The middle bits of the full product fluctuate the most with small |
| 120 | // changes in the input. This is the top bits of lo and the bottom bits |
| 121 | // of hi. We can thus make the entire output fluctuate with small |
| 122 | // changes to the input by XOR'ing these two halves. |
| 123 | lo ^ hi |
| 124 | } |
| 125 | |
| 126 | #[cfg (target_pointer_width = "32" )] |
| 127 | { |
| 128 | // u64 x u64 -> u128 product is prohibitively expensive on 32-bit. |
| 129 | // Decompose into 32-bit parts. |
| 130 | let lx = x as u32; |
| 131 | let ly = y as u32; |
| 132 | let hx = (x >> 32) as u32; |
| 133 | let hy = (y >> 32) as u32; |
| 134 | |
| 135 | // u32 x u32 -> u64 the low bits of one with the high bits of the other. |
| 136 | let afull = (lx as u64) * (hy as u64); |
| 137 | let bfull = (hx as u64) * (ly as u64); |
| 138 | |
| 139 | // Combine, swapping low/high of one of them so the upper bits of the |
| 140 | // product of one combine with the lower bits of the other. |
| 141 | afull ^ bfull.rotate_right(32) |
| 142 | } |
| 143 | } |
| 144 | |
| 145 | /// The foldhash implementation optimized for speed. |
| 146 | pub mod fast { |
| 147 | use super::*; |
| 148 | |
| 149 | pub use seed::fast::{FixedState, RandomState}; |
| 150 | |
| 151 | /// A [`Hasher`] instance implementing foldhash, optimized for speed. |
| 152 | /// |
| 153 | /// It can't be created directly, see [`RandomState`] or [`FixedState`]. |
| 154 | #[derive (Clone)] |
| 155 | pub struct FoldHasher { |
| 156 | accumulator: u64, |
| 157 | sponge: u128, |
| 158 | sponge_len: u8, |
| 159 | fold_seed: u64, |
| 160 | expand_seed: u64, |
| 161 | expand_seed2: u64, |
| 162 | expand_seed3: u64, |
| 163 | } |
| 164 | |
| 165 | impl FoldHasher { |
| 166 | #[inline ] |
| 167 | pub(crate) fn with_seed(per_hasher_seed: u64, global_seed: &[u64; 4]) -> FoldHasher { |
| 168 | FoldHasher { |
| 169 | accumulator: per_hasher_seed, |
| 170 | sponge: 0, |
| 171 | sponge_len: 0, |
| 172 | fold_seed: global_seed[0], |
| 173 | expand_seed: global_seed[1], |
| 174 | expand_seed2: global_seed[2], |
| 175 | expand_seed3: global_seed[3], |
| 176 | } |
| 177 | } |
| 178 | |
| 179 | #[inline (always)] |
| 180 | fn write_num<T: Into<u128>>(&mut self, x: T) { |
| 181 | let bits: usize = 8 * core::mem::size_of::<T>(); |
| 182 | if self.sponge_len as usize + bits > 128 { |
| 183 | let lo = self.sponge as u64; |
| 184 | let hi = (self.sponge >> 64) as u64; |
| 185 | self.accumulator = folded_multiply(lo ^ self.accumulator, hi ^ self.fold_seed); |
| 186 | self.sponge = x.into(); |
| 187 | self.sponge_len = bits as u8; |
| 188 | } else { |
| 189 | self.sponge |= x.into() << self.sponge_len; |
| 190 | self.sponge_len += bits as u8; |
| 191 | } |
| 192 | } |
| 193 | } |
| 194 | |
| 195 | impl Hasher for FoldHasher { |
| 196 | #[inline (always)] |
| 197 | fn write(&mut self, bytes: &[u8]) { |
| 198 | let mut s0 = self.accumulator; |
| 199 | let mut s1 = self.expand_seed; |
| 200 | let len = bytes.len(); |
| 201 | if len <= 16 { |
| 202 | // XOR the input into s0, s1, then multiply and fold. |
| 203 | if len >= 8 { |
| 204 | s0 ^= u64::from_ne_bytes(bytes[0..8].try_into().unwrap()); |
| 205 | s1 ^= u64::from_ne_bytes(bytes[len - 8..].try_into().unwrap()); |
| 206 | } else if len >= 4 { |
| 207 | s0 ^= u32::from_ne_bytes(bytes[0..4].try_into().unwrap()) as u64; |
| 208 | s1 ^= u32::from_ne_bytes(bytes[len - 4..].try_into().unwrap()) as u64; |
| 209 | } else if len > 0 { |
| 210 | let lo = bytes[0]; |
| 211 | let mid = bytes[len / 2]; |
| 212 | let hi = bytes[len - 1]; |
| 213 | s0 ^= lo as u64; |
| 214 | s1 ^= ((hi as u64) << 8) | mid as u64; |
| 215 | } |
| 216 | self.accumulator = folded_multiply(s0, s1); |
| 217 | } else if len < 256 { |
| 218 | self.accumulator = hash_bytes_medium(bytes, s0, s1, self.fold_seed); |
| 219 | } else { |
| 220 | self.accumulator = hash_bytes_long( |
| 221 | bytes, |
| 222 | s0, |
| 223 | s1, |
| 224 | self.expand_seed2, |
| 225 | self.expand_seed3, |
| 226 | self.fold_seed, |
| 227 | ); |
| 228 | } |
| 229 | } |
| 230 | |
| 231 | #[inline (always)] |
| 232 | fn write_u8(&mut self, i: u8) { |
| 233 | self.write_num(i); |
| 234 | } |
| 235 | |
| 236 | #[inline (always)] |
| 237 | fn write_u16(&mut self, i: u16) { |
| 238 | self.write_num(i); |
| 239 | } |
| 240 | |
| 241 | #[inline (always)] |
| 242 | fn write_u32(&mut self, i: u32) { |
| 243 | self.write_num(i); |
| 244 | } |
| 245 | |
| 246 | #[inline (always)] |
| 247 | fn write_u64(&mut self, i: u64) { |
| 248 | self.write_num(i); |
| 249 | } |
| 250 | |
| 251 | #[inline (always)] |
| 252 | fn write_u128(&mut self, i: u128) { |
| 253 | let lo = i as u64; |
| 254 | let hi = (i >> 64) as u64; |
| 255 | self.accumulator = folded_multiply(lo ^ self.accumulator, hi ^ self.fold_seed); |
| 256 | } |
| 257 | |
| 258 | #[inline (always)] |
| 259 | fn write_usize(&mut self, i: usize) { |
| 260 | // u128 doesn't implement From<usize>. |
| 261 | #[cfg (target_pointer_width = "32" )] |
| 262 | self.write_num(i as u32); |
| 263 | #[cfg (target_pointer_width = "64" )] |
| 264 | self.write_num(i as u64); |
| 265 | } |
| 266 | |
| 267 | #[inline (always)] |
| 268 | fn finish(&self) -> u64 { |
| 269 | if self.sponge_len > 0 { |
| 270 | let lo = self.sponge as u64; |
| 271 | let hi = (self.sponge >> 64) as u64; |
| 272 | folded_multiply(lo ^ self.accumulator, hi ^ self.fold_seed) |
| 273 | } else { |
| 274 | self.accumulator |
| 275 | } |
| 276 | } |
| 277 | } |
| 278 | } |
| 279 | |
| 280 | /// The foldhash implementation optimized for quality. |
| 281 | pub mod quality { |
| 282 | use super::*; |
| 283 | |
| 284 | pub use seed::quality::{FixedState, RandomState}; |
| 285 | |
| 286 | /// A [`Hasher`] instance implementing foldhash, optimized for quality. |
| 287 | /// |
| 288 | /// It can't be created directly, see [`RandomState`] or [`FixedState`]. |
| 289 | #[derive (Clone)] |
| 290 | pub struct FoldHasher { |
| 291 | pub(crate) inner: fast::FoldHasher, |
| 292 | } |
| 293 | |
| 294 | impl Hasher for FoldHasher { |
| 295 | #[inline (always)] |
| 296 | fn write(&mut self, bytes: &[u8]) { |
| 297 | self.inner.write(bytes); |
| 298 | } |
| 299 | |
| 300 | #[inline (always)] |
| 301 | fn write_u8(&mut self, i: u8) { |
| 302 | self.inner.write_u8(i); |
| 303 | } |
| 304 | |
| 305 | #[inline (always)] |
| 306 | fn write_u16(&mut self, i: u16) { |
| 307 | self.inner.write_u16(i); |
| 308 | } |
| 309 | |
| 310 | #[inline (always)] |
| 311 | fn write_u32(&mut self, i: u32) { |
| 312 | self.inner.write_u32(i); |
| 313 | } |
| 314 | |
| 315 | #[inline (always)] |
| 316 | fn write_u64(&mut self, i: u64) { |
| 317 | self.inner.write_u64(i); |
| 318 | } |
| 319 | |
| 320 | #[inline (always)] |
| 321 | fn write_u128(&mut self, i: u128) { |
| 322 | self.inner.write_u128(i); |
| 323 | } |
| 324 | |
| 325 | #[inline (always)] |
| 326 | fn write_usize(&mut self, i: usize) { |
| 327 | self.inner.write_usize(i); |
| 328 | } |
| 329 | |
| 330 | #[inline (always)] |
| 331 | fn finish(&self) -> u64 { |
| 332 | folded_multiply(self.inner.finish(), ARBITRARY0) |
| 333 | } |
| 334 | } |
| 335 | } |
| 336 | |
| 337 | /// Hashes strings >= 16 bytes, has unspecified behavior when bytes.len() < 16. |
| 338 | fn hash_bytes_medium(bytes: &[u8], mut s0: u64, mut s1: u64, fold_seed: u64) -> u64 { |
| 339 | // Process 32 bytes per iteration, 16 bytes from the start, 16 bytes from |
| 340 | // the end. On the last iteration these two chunks can overlap, but that is |
| 341 | // perfectly fine. |
| 342 | let left_to_right: ChunksExact<'_, u8> = bytes.chunks_exact(chunk_size:16); |
| 343 | let mut right_to_left: RChunksExact<'_, u8> = bytes.rchunks_exact(chunk_size:16); |
| 344 | for lo: &[u8] in left_to_right { |
| 345 | let hi: &[u8] = right_to_left.next().unwrap(); |
| 346 | let unconsumed_start: *const u8 = lo.as_ptr(); |
| 347 | let unconsumed_end: *const u8 = hi.as_ptr_range().end; |
| 348 | if unconsumed_start >= unconsumed_end { |
| 349 | break; |
| 350 | } |
| 351 | |
| 352 | let a: u64 = u64::from_ne_bytes(lo[0..8].try_into().unwrap()); |
| 353 | let b: u64 = u64::from_ne_bytes(lo[8..16].try_into().unwrap()); |
| 354 | let c: u64 = u64::from_ne_bytes(hi[0..8].try_into().unwrap()); |
| 355 | let d: u64 = u64::from_ne_bytes(hi[8..16].try_into().unwrap()); |
| 356 | s0 = folded_multiply(x:a ^ s0, y:c ^ fold_seed); |
| 357 | s1 = folded_multiply(x:b ^ s1, y:d ^ fold_seed); |
| 358 | } |
| 359 | |
| 360 | s0 ^ s1 |
| 361 | } |
| 362 | |
| 363 | /// Hashes strings >= 16 bytes, has unspecified behavior when bytes.len() < 16. |
| 364 | #[cold ] |
| 365 | #[inline (never)] |
| 366 | fn hash_bytes_long( |
| 367 | bytes: &[u8], |
| 368 | mut s0: u64, |
| 369 | mut s1: u64, |
| 370 | mut s2: u64, |
| 371 | mut s3: u64, |
| 372 | fold_seed: u64, |
| 373 | ) -> u64 { |
| 374 | let chunks = bytes.chunks_exact(64); |
| 375 | let remainder = chunks.remainder().len(); |
| 376 | for chunk in chunks { |
| 377 | let a = u64::from_ne_bytes(chunk[0..8].try_into().unwrap()); |
| 378 | let b = u64::from_ne_bytes(chunk[8..16].try_into().unwrap()); |
| 379 | let c = u64::from_ne_bytes(chunk[16..24].try_into().unwrap()); |
| 380 | let d = u64::from_ne_bytes(chunk[24..32].try_into().unwrap()); |
| 381 | let e = u64::from_ne_bytes(chunk[32..40].try_into().unwrap()); |
| 382 | let f = u64::from_ne_bytes(chunk[40..48].try_into().unwrap()); |
| 383 | let g = u64::from_ne_bytes(chunk[48..56].try_into().unwrap()); |
| 384 | let h = u64::from_ne_bytes(chunk[56..64].try_into().unwrap()); |
| 385 | s0 = folded_multiply(a ^ s0, e ^ fold_seed); |
| 386 | s1 = folded_multiply(b ^ s1, f ^ fold_seed); |
| 387 | s2 = folded_multiply(c ^ s2, g ^ fold_seed); |
| 388 | s3 = folded_multiply(d ^ s3, h ^ fold_seed); |
| 389 | } |
| 390 | s0 ^= s2; |
| 391 | s1 ^= s3; |
| 392 | |
| 393 | if remainder > 0 { |
| 394 | hash_bytes_medium(&bytes[bytes.len() - remainder.max(16)..], s0, s1, fold_seed) |
| 395 | } else { |
| 396 | s0 ^ s1 |
| 397 | } |
| 398 | } |
| 399 | |