diff --git a/lib/std/hash/cityhash.zig b/lib/std/hash/cityhash.zig index 8040c99b84..464e876b4a 100644 --- a/lib/std/hash/cityhash.zig +++ b/lib/std/hash/cityhash.zig @@ -342,64 +342,21 @@ pub const CityHash64 = struct { } }; -fn SMHasherTest(comptime hash_fn: anytype) u32 { - const HashResult = @typeInfo(@TypeOf(hash_fn)).Fn.return_type.?; - - var key: [256]u8 = undefined; - var hashes_bytes: [256 * @sizeOf(HashResult)]u8 = undefined; - - @memset(&key, 0); - @memset(&hashes_bytes, 0); - - var i: u32 = 0; - while (i < 256) : (i += 1) { - key[i] = @as(u8, @intCast(i)); - - var h: HashResult = hash_fn(key[0..i], 256 - i); - - // comptime can't really do reinterpret casting yet, - // so we need to write the bytes manually. - for (hashes_bytes[i * @sizeOf(HashResult) ..][0..@sizeOf(HashResult)]) |*byte| { - byte.* = @as(u8, @truncate(h)); - h = h >> 8; - } - } - - return @as(u32, @truncate(hash_fn(&hashes_bytes, 0))); -} - fn CityHash32hashIgnoreSeed(str: []const u8, seed: u32) u32 { _ = seed; return CityHash32.hash(str); } +const verify = @import("verify.zig"); + test "cityhash32" { - const Test = struct { - fn doTest() !void { - // Note: SMHasher doesn't provide a 32bit version of the algorithm. - // Note: The implementation was verified against the Google Abseil version. - try std.testing.expectEqual(SMHasherTest(CityHash32hashIgnoreSeed), 0x68254F81); - try std.testing.expectEqual(SMHasherTest(CityHash32hashIgnoreSeed), 0x68254F81); - } - }; - try Test.doTest(); - // TODO This is uncommented to prevent OOM on the CI server. Re-enable this test - // case once we ship stage2. - //@setEvalBranchQuota(50000); - //comptime Test.doTest(); + // Note: SMHasher doesn't provide a 32bit version of the algorithm. + // Note: The implementation was verified against the Google Abseil version. + try std.testing.expectEqual(verify.smhasher(CityHash32hashIgnoreSeed), 0x68254F81); } test "cityhash64" { - const Test = struct { - fn doTest() !void { - // Note: This is not compliant with the SMHasher implementation of CityHash64! - // Note: The implementation was verified against the Google Abseil version. - try std.testing.expectEqual(SMHasherTest(CityHash64.hashWithSeed), 0x5FABC5C5); - } - }; - try Test.doTest(); - // TODO This is uncommented to prevent OOM on the CI server. Re-enable this test - // case once we ship stage2. - //@setEvalBranchQuota(50000); - //comptime Test.doTest(); + // Note: This is not compliant with the SMHasher implementation of CityHash64! + // Note: The implementation was verified against the Google Abseil version. + try std.testing.expectEqual(verify.smhasher(CityHash64.hashWithSeed), 0x5FABC5C5); } diff --git a/lib/std/hash/fnv.zig b/lib/std/hash/fnv.zig index 3175f26820..c95f19c04e 100644 --- a/lib/std/hash/fnv.zig +++ b/lib/std/hash/fnv.zig @@ -4,7 +4,7 @@ // // https://tools.ietf.org/html/draft-eastlake-fnv-14 -const std = @import("../std.zig"); +const std = @import("std"); const testing = std.testing; pub const Fnv1a_32 = Fnv1a(u32, 0x01000193, 0x811c9dc5); diff --git a/lib/std/hash/murmur.zig b/lib/std/hash/murmur.zig index f6af810c6e..b9b2ddeca2 100644 --- a/lib/std/hash/murmur.zig +++ b/lib/std/hash/murmur.zig @@ -279,26 +279,10 @@ pub const Murmur3_32 = struct { } }; -fn SMHasherTest(comptime hash_fn: anytype, comptime hashbits: u32) u32 { - const hashbytes = hashbits / 8; - var key: [256]u8 = [1]u8{0} ** 256; - var hashes: [hashbytes * 256]u8 = [1]u8{0} ** (hashbytes * 256); - - var i: u32 = 0; - while (i < 256) : (i += 1) { - key[i] = @as(u8, @truncate(i)); - - var h = hash_fn(key[0..i], 256 - i); - if (native_endian == .Big) - h = @byteSwap(h); - @memcpy(hashes[i * hashbytes ..][0..hashbytes], @as([*]u8, @ptrCast(&h))); - } - - return @as(u32, @truncate(hash_fn(&hashes, 0))); -} +const verify = @import("verify.zig"); test "murmur2_32" { - try testing.expectEqual(SMHasherTest(Murmur2_32.hashWithSeed, 32), 0x27864C1E); + try testing.expectEqual(verify.smhasher(Murmur2_32.hashWithSeed), 0x27864C1E); var v0: u32 = 0x12345678; var v1: u64 = 0x1234567812345678; var v0le: u32 = v0; @@ -312,7 +296,7 @@ test "murmur2_32" { } test "murmur2_64" { - try std.testing.expectEqual(SMHasherTest(Murmur2_64.hashWithSeed, 64), 0x1F0D3804); + try std.testing.expectEqual(verify.smhasher(Murmur2_64.hashWithSeed), 0x1F0D3804); var v0: u32 = 0x12345678; var v1: u64 = 0x1234567812345678; var v0le: u32 = v0; @@ -326,7 +310,7 @@ test "murmur2_64" { } test "murmur3_32" { - try std.testing.expectEqual(SMHasherTest(Murmur3_32.hashWithSeed, 32), 0xB0F57EE3); + try std.testing.expectEqual(verify.smhasher(Murmur3_32.hashWithSeed), 0xB0F57EE3); var v0: u32 = 0x12345678; var v1: u64 = 0x1234567812345678; var v0le: u32 = v0; diff --git a/lib/std/hash/verify.zig b/lib/std/hash/verify.zig new file mode 100644 index 0000000000..bfc6a1f22c --- /dev/null +++ b/lib/std/hash/verify.zig @@ -0,0 +1,35 @@ +const std = @import("std"); + +fn hashMaybeSeed(comptime hash_fn: anytype, seed: anytype, buf: []const u8) @typeInfo(@TypeOf(hash_fn)).Fn.return_type.? { + const HashFn = @typeInfo(@TypeOf(hash_fn)).Fn; + if (HashFn.params.len > 1) { + if (@typeInfo(HashFn.params[0].type.?) == .Int) { + return hash_fn(@intCast(seed), buf); + } else { + return hash_fn(buf, @intCast(seed)); + } + } else { + return hash_fn(buf); + } +} + +// Returns a verification code, the same as user by SMHasher. +// +// Hash keys of the form {0}, {0,1}, {0,1,2}... up to N=255, using 256-N as seed. +// First four-bytes of the hash, interpreted as little-endian is the verification code. +pub fn smhasher(comptime hash_fn: anytype) u32 { + const HashFnTy = @typeInfo(@TypeOf(hash_fn)).Fn; + const HashResult = HashFnTy.return_type.?; + const hash_size = @sizeOf(HashResult); + + var buf: [256]u8 = undefined; + var buf_all: [256 * hash_size]u8 = undefined; + + for (0..256) |i| { + buf[i] = @intCast(i); + const h = hashMaybeSeed(hash_fn, 256 - i, buf[0..i]); + std.mem.writeIntLittle(HashResult, buf_all[i * hash_size ..][0..hash_size], h); + } + + return @truncate(hashMaybeSeed(hash_fn, 0, buf_all[0..])); +} diff --git a/lib/std/hash/wyhash.zig b/lib/std/hash/wyhash.zig index acce3a7732..741b117f5d 100644 --- a/lib/std/hash/wyhash.zig +++ b/lib/std/hash/wyhash.zig @@ -196,6 +196,7 @@ pub const Wyhash = struct { } }; +const verify = @import("verify.zig"); const expectEqual = std.testing.expectEqual; const TestVector = struct { @@ -229,6 +230,10 @@ test "test vectors at comptime" { } } +test "smhasher" { + try expectEqual(verify.smhasher(Wyhash.hash), 0xBD5E840C); +} + test "test vectors streaming" { const step = 5; diff --git a/lib/std/hash/xxhash.zig b/lib/std/hash/xxhash.zig index 8e25b9ce7b..6b6208b854 100644 --- a/lib/std/hash/xxhash.zig +++ b/lib/std/hash/xxhash.zig @@ -438,6 +438,8 @@ fn validateType(comptime T: type) void { } } +const verify = @import("verify.zig"); + fn testExpect(comptime H: type, seed: anytype, input: []const u8, expected: u64) !void { try expectEqual(expected, H.hash(0, input)); @@ -455,6 +457,8 @@ test "xxhash64" { try testExpect(H, 0, "abcdefghijklmnopqrstuvwxyz", 0xcfe1f278fa89835c); try testExpect(H, 0, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789", 0xaaa46907d3047814); try testExpect(H, 0, "12345678901234567890123456789012345678901234567890123456789012345678901234567890", 0xe04a477f19ee145d); + + try expectEqual(verify.smhasher(H.hash), 0x024B7CF4); } test "xxhash32" { @@ -467,4 +471,6 @@ test "xxhash32" { try testExpect(H, 0, "abcdefghijklmnopqrstuvwxyz", 0x63a14d5f); try testExpect(H, 0, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789", 0x9c285e64); try testExpect(H, 0, "12345678901234567890123456789012345678901234567890123456789012345678901234567890", 0x9c05f475); + + try expectEqual(verify.smhasher(H.hash), 0xBA88B743); }