module.ae

// std.hash — fast NON-cryptographic hashes for checksums and hashmap
// seeding.
//
//   fnv32 / fnv64  — FNV-1a (Fowler–Noll–Vo), simple and fast for short keys
//   murmur3_32     — MurmurHash3 x86 32-bit, good distribution for hashmaps
//   siphash24      — SipHash-2-4, a KEYED PRF: DoS-resistant hashmap seeding
//                    and short-message MACs. Not a general crypto hash, but
//                    unlike the others it is keyed and hash-flood resistant.
//   crc32          — CRC-32 (IEEE 802.3, the reflected 0xEDB88320
//                    polynomial): the checksum PNG chunks, gzip members and
//                    ZIP entries carry. crc32_update continues one across
//                    several pieces of input.
//
// Input is a `byte[]` slice whose length travels with it (#2301): pass a
// string as `string.bytes(s)`, a prefix as `string.bytes(s)[0..n]`, a
// `std.bytes` buffer as `bytes.view(buf)`. Embedded NULs are honoured.
// Results are the exact 32-/64-bit unsigned values the reference algorithms
// produce, carried in Aether's signed `int`/`long` (mask with 0xFFFFFFFF /
// interpret bit-for-bit as unsigned). fnv/murmur are ported from C3's
// `std::hash::*` (MIT, Copyright (c) 2022-2026 Christoffer Lernö and
// contributors); siphash from C3's std::hash::siphash (MIT, Copyright (c)
// 2025 Zack Puhl), both re-expressed in Aether.

import std.bits

// A slice with no bound (`p as byte[]`, `.len` -1) has no length to hash.
fn unbounded(fname: string) {
    panic("hash.${fname}: the slice has no bound (.len -1); give it one, as `(p as byte[])[0..n]`")
}

exports(
    fnv32, fnv64, murmur3_32, siphash24, crc32, crc32_update
)

// FNV-1a 32-bit. offset basis 0x811c9dc5, prime 0x01000193. Returned as a
// non-negative 32-bit value in a `long` (masked each step so the multiply
// wraps like unsigned uint32 regardless of Aether's signed int width).
fnv32(data: byte[]) -> long {
    if data.len < 0 {
        unbounded("fnv32")
    }
    int length = data.len
    let mask: long = 0xFFFFFFFF
    let h: long = 0x811c9dc5
    i = 0
    while i < length {
        b = (data[i] & 255) as long
        h = (h ^ b) & mask
        h = (h * 0x01000193) & mask
        i = i + 1
    }
    return h
}

// FNV-1a 64-bit. offset basis 0xcbf29ce484222325, prime 0x100000001b3.
// Uses wrapping 64-bit multiply so the product wraps like unsigned uint64.
fnv64(data: byte[]) -> long {
    if data.len < 0 {
        unbounded("fnv64")
    }
    int length = data.len
    h = 0xcbf29ce484222325
    prime = 0x100000001b3
    i = 0
    while i < length {
        b = (data[i] & 255) as long
        h = h ^ b
        h = bits.wrapping_mul64(h, prime)
        i = i + 1
    }
    return h
}

// MurmurHash3 x86 32-bit. `seed` lets callers salt the hash (0 for a plain
// digest). Processes 4-byte little-endian blocks then the 1..3-byte tail,
// with the standard fmix32 finalizer.
// Implemented in 64-bit `long` throughout with an explicit 32-bit mask after
// every step, so the arithmetic matches unsigned uint32 exactly regardless
// of Aether's signed-int width, and the final value is a non-negative 32-bit
// number. Rotates use the 32-bit rotate on the masked value.
murmur3_32(data: byte[], seed: int) -> long {
    if data.len < 0 {
        unbounded("murmur3_32")
    }
    int length = data.len
    let c1: long = 0xcc9e2d51
    let c2: long = 0x1b873593
    let mask: long = 0xFFFFFFFF
    let h1: long = (seed as long) & mask
    nblocks = length / 4
    i = 0
    while i < nblocks {
        base = i * 4
        b0 = (data[base] & 255) as long
        b1 = (data[base + 1] & 255) as long
        b2 = (data[base + 2] & 255) as long
        b3 = (data[base + 3] & 255) as long
        let k1: long = (b0 | (b1 << 8) | (b2 << 16) | (b3 << 24)) & mask
        k1 = (k1 * c1) & mask
        k1 = rotl32m(k1, 15)
        k1 = (k1 * c2) & mask
        h1 = (h1 ^ k1) & mask
        h1 = rotl32m(h1, 13)
        h1 = ((h1 * 5) + 0xe6546b64) & mask
        i = i + 1
    }
    // tail: remaining 1..3 bytes
    tail = nblocks * 4
    rem = length - tail
    let k1t: long = 0
    if rem == 3 { k1t = k1t ^ (((data[tail + 2] & 255) as long) << 16) }
    if rem >= 2 { k1t = k1t ^ (((data[tail + 1] & 255) as long) << 8) }
    if rem >= 1 {
        k1t = k1t ^ ((data[tail] & 255) as long)
        k1t = (k1t * c1) & mask
        k1t = rotl32m(k1t, 15)
        k1t = (k1t * c2) & mask
        h1 = (h1 ^ k1t) & mask
    }
    // finalization
    h1 = (h1 ^ (length as long)) & mask
    h1 = fmix32(h1)
    return h1
}

// 32-bit left rotate of a value already masked to 32 bits, in `long`.
fn rotl32m(x: long, r: int) -> long {
    let mask: long = 0xFFFFFFFF
    return ((x << r) | (x >> (32 - r))) & mask
}

// MurmurHash3 32-bit finalizer (in `long`, masked).
fn fmix32(h0: long) -> long {
    let mask: long = 0xFFFFFFFF
    h = h0 & mask
    h = (h ^ (h >> 16)) & mask
    h = (h * 0x85ebca6b) & mask
    h = (h ^ (h >> 13)) & mask
    h = (h * 0xc2b2ae35) & mask
    h = (h ^ (h >> 16)) & mask
    return h
}

// ---------------------------------------------------------------------------
// SipHash-2-4 (64-bit output). A KEYED pseudorandom function — the standard
// choice for hash-flood-resistant hashmaps and short-message MACs. Ported
// from C3's std::hash::siphash (MIT, © 2025 Zack Puhl), following the
// reference SipHash construction (Aumasson & Bernstein).
//
// All arithmetic is 64-bit with modular (wrapping) addition — bits.
// wrapping_add64 — and 64-bit left rotates (bits.rotl64). The 128-bit key is
// supplied as two `long` halves (k0 = low 64 bits, k1 = high 64 bits); the
// message is a `byte[]` slice. The returned `long` is the
// 64-bit digest interpreted bit-for-bit (may be negative when the top bit is
// set). c = 2 compression rounds per block, d = 4 finalization rounds.
// ---------------------------------------------------------------------------

// One SipRound over the 4-word state; returns the updated (v0,v1,v2,v3).
fn sipround(v0: long, v1: long, v2: long, v3: long) -> (long, long, long, long) {
    a0 = bits.wrapping_add64(v0, v1)
    b1 = bits.rotl64(v1, 13) ^ a0
    a0b = bits.rotl64(a0, 32)
    a2 = bits.wrapping_add64(v2, v3)
    b3 = bits.rotl64(v3, 16) ^ a2
    a0c = bits.wrapping_add64(a0b, b3)
    b3b = bits.rotl64(b3, 21) ^ a0c
    a2b = bits.wrapping_add64(a2, b1)
    b1b = bits.rotl64(b1, 17) ^ a2b
    a2c = bits.rotl64(a2b, 32)
    return a0c, b1b, a2c, b3b
}

// Read up to 8 little-endian bytes starting at `off` into a 64-bit word
// (missing bytes are 0). `cnt` bytes are taken.
fn le64(data: byte[], off: int, cnt: int) -> long {
    let w: long = 0
    i = 0
    while i < cnt {
        b = (data[off + i] & 255) as long
        w = w | (b << (i * 8))
        i = i + 1
    }
    return w
}

// SipHash-2-4 of the bytes of `data` under the 128-bit key (k0, k1).
siphash24(data: byte[], k0: long, k1: long) -> long {
    if data.len < 0 {
        unbounded("siphash24")
    }
    int length = data.len
    let v0: long = 0x736f6d6570736575 ^ k0
    let v1: long = 0x646f72616e646f6d ^ k1
    let v2: long = 0x6c7967656e657261 ^ k0
    let v3: long = 0x7465646279746573 ^ k1

    // full 8-byte blocks
    nblocks = length / 8
    b = 0
    while b < nblocks {
        m = le64(data, b * 8, 8)
        v3 = v3 ^ m
        v0, v1, v2, v3 = sipround(v0, v1, v2, v3) // c = 2
        v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
        v0 = v0 ^ m
        b = b + 1
    }

    // final block: remaining bytes + the length in the top byte
    tail = nblocks * 8
    rem = length - tail
    mlast = le64(data, tail, rem)
    mlast = mlast | ((length as long) << 56)
    v3 = v3 ^ mlast
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
    v0 = v0 ^ mlast

    // finalization: d = 4 rounds
    v2 = v2 ^ 0xff
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)
    v0, v1, v2, v3 = sipround(v0, v1, v2, v3)

    return v0 ^ v1 ^ v2 ^ v3
}

// ---- CRC-32 (IEEE 802.3) ----------------------------------------------------
// The reflected table: the CRC of each byte value under the polynomial
// 0xEDB88320, so the checksum costs one lookup per input byte. A constant,
// so it is read-only data shared by every thread, with nothing built at run
// time for two first callers to race on.

const CRC32_TABLE: uint32[256] = [
    0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F,
    0xE963A535, 0x9E6495A3, 0x0EDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988,
    0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91, 0x1DB71064, 0x6AB020F2,
    0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7,
    0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9,
    0xFA0F3D63, 0x8D080DF5, 0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172,
    0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B, 0x35B5A8FA, 0x42B2986C,
    0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59,
    0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423,
    0xCFBA9599, 0xB8BDA50F, 0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924,
    0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D, 0x76DC4190, 0x01DB7106,
    0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433,
    0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D,
    0x91646C97, 0xE6635C01, 0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E,
    0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457, 0x65B0D9C6, 0x12B7E950,
    0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65,
    0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7,
    0xA4D1C46D, 0xD3D6F4FB, 0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0,
    0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9, 0x5005713C, 0x270241AA,
    0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F,
    0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81,
    0xB7BD5C3B, 0xC0BA6CAD, 0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A,
    0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683, 0xE3630B12, 0x94643B84,
    0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1,
    0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB,
    0x196C3671, 0x6E6B06E7, 0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC,
    0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5, 0xD6D6A3E8, 0xA1D1937E,
    0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B,
    0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55,
    0x316E8EEF, 0x4669BE79, 0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236,
    0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F, 0xC5BA3BBE, 0xB2BD0B28,
    0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D,
    0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F,
    0x72076785, 0x05005713, 0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38,
    0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21, 0x86D3D2D4, 0xF1D4E242,
    0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777,
    0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69,
    0x616BFFD3, 0x166CCF45, 0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2,
    0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB, 0xAED16A4A, 0xD9D65ADC,
    0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9,
    0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693,
    0x54DE5729, 0x23D967BF, 0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94,
    0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D
]

// Continues a CRC-32 over the bytes of `data`: pass the value a
// previous crc32 / crc32_update returned (0 to start), and the result is the
// CRC of everything so far, as crc32 of the whole input would be. Returned as
// the unsigned 32-bit value in a `long`.
crc32_update(crc: long, data: byte[]) -> long {
    if data.len < 0 {
        unbounded("crc32_update")
    }
    int length = data.len
    let mask: long = 0xFFFFFFFF
    let c: long = (crc ^ mask) & mask
    i = 0
    while i < length {
        b = (data[i] & 255) as long
        idx = ((c ^ b) & 255) as int
        entry = (CRC32_TABLE[idx] as long) & mask
        c = (entry ^ (c >> 8)) & mask
        i = i + 1
    }
    return (c ^ mask) & mask
}

// CRC-32 of the bytes of `data`.
crc32(data: byte[]) -> long {
    return crc32_update(0, data)
}