std.hash.int: use anytype instead of explicit type parameter

also
* allow signed ints, simply bitcast them to unsigned
* handle odd bit sizes by upcasting and then truncating
* naming conventions
* remove redundant code
* better use of testing API
This commit is contained in:
Andrew Kelley 2024-11-01 11:57:39 -07:00
parent aee6f7d7ee
commit 5ad44c14b0

View File

@ -37,25 +37,31 @@ pub const XxHash3 = xxhash.XxHash3;
pub const XxHash64 = xxhash.XxHash64;
pub const XxHash32 = xxhash.XxHash32;
/// Deprecated: use std.hash.int(comptime T, input T) T where T is an unsigned integer type.
/// This is handy if you have a u32 and want a u32 and don't want to take a
/// detour through many layers of abstraction elsewhere in the std.hash
/// namespace.
/// Deprecated in favor of `int`.
pub fn uint32(input: u32) u32 {
return int(u32, input);
return int(input);
}
/// Applies a bit-mangling transformation to an unsigned integer type `T`.
/// Optimized per type: for `u16` and `u32`, Skeeto's xorshift-multiply; for `u64`, Maiga's mx3.
/// Falls back on an avalanche pattern for other unsigned types, ensuring high entropy.
/// Only unsigned types are accepted; signed types will raise a compile-time error.
pub fn int(comptime T: type, input: T) T {
const tInfo = @typeInfo(T).int;
if (tInfo.signedness != .unsigned) @compileError("type has to be unsigned integer");
/// Falls back on an avalanche pattern for other integer types, ensuring high entropy.
pub fn int(input: anytype) @TypeOf(input) {
const info = @typeInfo(@TypeOf(input)).int;
if (info.signedness == .signed) {
const Unsigned = @Type(.{ .int = .{ .signedness = .unsigned, .bits = info.bits } });
const casted: Unsigned = @bitCast(input);
return @bitCast(int(casted));
} else if (info.bits < 16) {
return @truncate(int(@as(u16, input)));
} else if (info.bits < 32) {
return @truncate(int(@as(u32, input)));
} else if (info.bits < 64) {
return @truncate(int(@as(u64, input)));
}
var x = input;
switch (T) {
u16 => {
//https://github.com/skeeto/hash-prospector
switch (info.bits) {
16 => {
// https://github.com/skeeto/hash-prospector
// 3-round xorshift-multiply (-Xn3)
// bias = 0.0045976709018820602
x = (x ^ (x >> 7)) *% 0x2993;
@ -63,36 +69,34 @@ pub fn int(comptime T: type, input: T) T {
x = (x ^ (x >> 9)) *% 0x0235;
x = x ^ (x >> 10);
},
u32 => {
32 => {
// https://github.com/skeeto/hash-prospector
x = (x ^ (x >> 17)) *% 0xed5ad4bb;
x = (x ^ (x >> 11)) *% 0xac4c1b51;
x = (x ^ (x >> 15)) *% 0x31848bab;
x = x ^ (x >> 14);
},
u64 => {
64 => {
// https://github.com/jonmaiga/mx3
// https://github.com/jonmaiga/mx3/blob/48924ee743d724aea2cafd2b4249ef8df57fa8b9/mx3.h#L17
const C = 0xbea225f9eb34556d;
x = (x ^ (x >> 32)) *% C;
x = (x ^ (x >> 29)) *% C;
x = (x ^ (x >> 32)) *% C;
const c = 0xbea225f9eb34556d;
x = (x ^ (x >> 32)) *% c;
x = (x ^ (x >> 29)) *% c;
x = (x ^ (x >> 32)) *% c;
x = x ^ (x >> 29);
},
else => {
// this construction provides robust avalanche properties, but it is not optimal for any given size.
const Tsize = @bitSizeOf(T);
if (Tsize < 4) @compileError("not implemented.");
const hsize = Tsize >> 1;
const C = comptime blk: {
const max = (1 << Tsize) - 1;
// This construction provides robust avalanche properties, but it is not optimal for any given size.
const hsize = info.bits >> 1;
const c = comptime blk: {
const max = (1 << info.bits) - 1;
var mul = 1;
while (mul * 3 < max) mul *= 3;
break :blk ((mul ^ (mul >> hsize)) | 1);
};
inline for (0..2) |_| {
x = (x ^ (x >> hsize + 1)) *% C;
x = (x ^ (x >> hsize - 1)) *% C;
x = (x ^ (x >> hsize + 1)) *% c;
x = (x ^ (x >> hsize - 1)) *% c;
}
x ^= (x >> hsize);
},
@ -100,14 +104,15 @@ pub fn int(comptime T: type, input: T) T {
return x;
}
test "bit manglers" {
const expect = @import("std").testing.expect;
try expect(int(u4, 1) == 0xC);
try expect(int(u8, 1) == 0x4F);
try expect(int(u16, 1) == 0x2880);
try expect(int(u32, 1) == 0x42741D6);
try expect(int(u64, 1) == 0x71894DE00D9981F);
try expect(int(u128, 1) == 0x50BC2BB18910C3DE0BAA2CE0D0C5B83E);
test int {
const expectEqual = @import("std").testing.expectEqual;
try expectEqual(0xC, int(@as(u4, 1)));
try expectEqual(0x4F, int(@as(u8, 1)));
try expectEqual(0x4F, int(@as(i8, 1)));
try expectEqual(0x2880, int(@as(u16, 1)));
try expectEqual(0x42741D6, int(@as(u32, 1)));
try expectEqual(0x71894DE00D9981F, int(@as(u64, 1)));
try expectEqual(0x50BC2BB18910C3DE0BAA2CE0D0C5B83E, int(@as(u128, 1)));
}
test {