/*
 * sha256.h — Standalone RFC 6234 SHA-256, header-only (static inline).
 * No external dependencies beyond <stdint.h> and <string.h>.
 *
 * Test vector:
 *   sha256("abc", 3, d) →
 *   ba7816bf 8f01cfea 414140de 5dae2ec7 3b00361b bde327b6 0b82c10f 46850c58
 */
#ifndef SHA256_H
#define SHA256_H

#include <stdint.h>
#include <stddef.h>
#include <string.h>

typedef struct {
    uint32_t state[8];
    uint64_t count;      /* total bits processed */
    uint8_t  buf[64];    /* partial block */
} sha256_ctx;

/* ── Round constants (first 32 bits of cube roots of first 64 primes) ── */
static const uint32_t SHA256_K[64] = {
    0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
    0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
    0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
    0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
    0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
    0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
    0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
    0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
    0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
    0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
    0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
    0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
    0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
    0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
    0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
    0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
};

/* ── Bit operations ── */
#define SHA256_ROTR(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#define SHA256_CH(e, f, g)   (((e) & (f)) ^ (~(e) & (g)))
#define SHA256_MAJ(a, b, c)  (((a) & (b)) ^ ((a) & (c)) ^ ((b) & (c)))
#define SHA256_EP0(a) (SHA256_ROTR(a,  2) ^ SHA256_ROTR(a, 13) ^ SHA256_ROTR(a, 22))
#define SHA256_EP1(e) (SHA256_ROTR(e,  6) ^ SHA256_ROTR(e, 11) ^ SHA256_ROTR(e, 25))
#define SHA256_SIG0(w)(SHA256_ROTR(w,  7) ^ SHA256_ROTR(w, 18) ^ ((w) >>  3))
#define SHA256_SIG1(w)(SHA256_ROTR(w, 17) ^ SHA256_ROTR(w, 19) ^ ((w) >> 10))

/* ── Process one 64-byte block ── */
static inline void sha256_transform(sha256_ctx *ctx, const uint8_t blk[64])
{
    uint32_t w[64];
    uint32_t a, b, c, d, e, f, g, h, t1, t2;
    int i;

    for (i = 0; i < 16; i++)
        w[i] = ((uint32_t)blk[i*4  ] << 24) | ((uint32_t)blk[i*4+1] << 16)
             | ((uint32_t)blk[i*4+2] <<  8) |  (uint32_t)blk[i*4+3];
    for (i = 16; i < 64; i++)
        w[i] = SHA256_SIG1(w[i-2]) + w[i-7] + SHA256_SIG0(w[i-15]) + w[i-16];

    a = ctx->state[0]; b = ctx->state[1]; c = ctx->state[2]; d = ctx->state[3];
    e = ctx->state[4]; f = ctx->state[5]; g = ctx->state[6]; h = ctx->state[7];

    for (i = 0; i < 64; i++) {
        t1 = h + SHA256_EP1(e) + SHA256_CH(e, f, g) + SHA256_K[i] + w[i];
        t2 = SHA256_EP0(a) + SHA256_MAJ(a, b, c);
        h = g; g = f; f = e; e = d + t1;
        d = c; c = b; b = a; a = t1 + t2;
    }

    ctx->state[0] += a; ctx->state[1] += b; ctx->state[2] += c; ctx->state[3] += d;
    ctx->state[4] += e; ctx->state[5] += f; ctx->state[6] += g; ctx->state[7] += h;
}

/* ── Public API ── */
static inline void sha256_init(sha256_ctx *ctx)
{
    ctx->state[0] = 0x6a09e667; ctx->state[1] = 0xbb67ae85;
    ctx->state[2] = 0x3c6ef372; ctx->state[3] = 0xa54ff53a;
    ctx->state[4] = 0x510e527f; ctx->state[5] = 0x9b05688c;
    ctx->state[6] = 0x1f83d9ab; ctx->state[7] = 0x5be0cd19;
    ctx->count = 0;
    memset(ctx->buf, 0, sizeof(ctx->buf));
}

static inline void sha256_update(sha256_ctx *ctx, const uint8_t *data, size_t len)
{
    /* bytes already in partial block */
    size_t used = (size_t)((ctx->count / 8) % 64);
    ctx->count += (uint64_t)len * 8;

    if (used) {
        size_t fill = 64 - used;
        if (len < fill) {
            memcpy(ctx->buf + used, data, len);
            return;
        }
        memcpy(ctx->buf + used, data, fill);
        sha256_transform(ctx, ctx->buf);
        data += fill;
        len  -= fill;
    }

    while (len >= 64) {
        sha256_transform(ctx, data);
        data += 64;
        len  -= 64;
    }

    if (len)
        memcpy(ctx->buf, data, len);
}

static inline void sha256_final(sha256_ctx *ctx, uint8_t digest[32])
{
    /* Padding: append 0x80, zeros, then 64-bit big-endian bit count */
    uint64_t bits = ctx->count;
    size_t   used = (size_t)((ctx->count / 8) % 64);
    uint8_t  pad[64];
    int i;

    memset(pad, 0, sizeof(pad));
    pad[0] = 0x80;

    if (used < 56) {
        /* Enough room in current block: pad to byte 56 */
        sha256_update(ctx, pad, 56 - used);
    } else {
        /* Need an extra block: fill to end of current, then 56 zero bytes */
        sha256_update(ctx, pad, 64 - used + 56);
    }

    /* Append 64-bit bit-length (big-endian) */
    for (i = 0; i < 8; i++)
        pad[i] = (uint8_t)(bits >> (56 - 8 * i));
    sha256_update(ctx, pad, 8);

    /* Extract digest from state */
    for (i = 0; i < 8; i++) {
        digest[i*4 + 0] = (uint8_t)(ctx->state[i] >> 24);
        digest[i*4 + 1] = (uint8_t)(ctx->state[i] >> 16);
        digest[i*4 + 2] = (uint8_t)(ctx->state[i] >>  8);
        digest[i*4 + 3] = (uint8_t)(ctx->state[i]      );
    }
}

/* One-shot convenience: init + update + final */
static inline void sha256(const uint8_t *data, size_t len, uint8_t digest[32])
{
    sha256_ctx ctx;
    sha256_init(&ctx);
    sha256_update(&ctx, data, len);
    sha256_final(&ctx, digest);
}

#endif /* SHA256_H */