dsk/lzw.c

/*
 * LZW algorithm as used in the IBM .DSK floppy disk image format.
 * Uses a fixed 12bit code size and an LRU dictionary entry replacement policy.
 * The only special code is 0 and it's used to represent the end of the
 * compressed stream. Single byte strings (0x00 to 0xff) occupy codes from
 * 1 to 256.
 */

#include "lzw.h"
#include <stddef.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>

void lzw_init(struct lzw_ctx* c)
{
    uint16_t i;

    c->input_buffer = 0;
    c->dict[0].usecount = 256;
    c->dict[0].length = 0;
    c->dict[0].prefix = c->dict[0].lru_prev = c->dict[0].lru_next = 0;
    for (i = 1; i <= 256; i++) {
        c->dict[i].length = 1;
        c->dict[i].prefix = 0;
        c->dict[i].last = i - 1;
        c->dict[i].first = i - 1;
        c->dict[i].usecount = 1; /* so that this code will never end up in the lru list */
        c->dict[i].lru_prev = c->dict[i].lru_next = 0;
    }
    for (i = 257; i < 4096; i++) {
        c->dict[i].length = 0;
        c->dict[i].prefix = 0;
        c->dict[i].last = 0;
        c->dict[i].first = 0;
        c->dict[i].usecount = 0;
        c->dict[i].lru_prev = (i == 257 ? 0 : i - 1);
        c->dict[i].lru_next = (i == 4095 ? 0 : i + 1);
    }
    c->lru_head = 257;
    c->lru_tail = 4095;
    c->last_emitted_code = 0;
    c->output_buffer_start = 0;
    c->output_buffer_used = 0;
    c->input_code = c->input_code_bits = 0;
    c->uncompressed_bytes_processed = c->compressed_bytes_processed = c->codes_processed = 0;
    c->eos = 0;
}

/* Remove code from the lru list */
static void lzw_lru_unlink(struct lzw_ctx* c, uint16_t code)
{
    uint16_t next, prev;

    next = c->dict[code].lru_next;
    prev = c->dict[code].lru_prev;
    if (prev != 0)
        c->dict[prev].lru_next = next;
    else
        c->lru_head = next;
    if (next != 0)
        c->dict[next].lru_prev = prev;
    else
        c->lru_tail = prev;
}

/* Add the given code to the back of the lru list. The code must be removed from the list before calling this function. */
static void lzw_lru_append(struct lzw_ctx* c, uint16_t code)
{
    c->dict[code].lru_next = 0;
    c->dict[code].lru_prev = c->lru_tail;
    c->dict[c->lru_tail].lru_next = code;
    c->lru_tail = code;
}

#ifdef DEBUG
static void lzw_print_lru(struct lzw_ctx* c)
{
    printf("head=%03x tail=%03x\n", c->lru_head, c->lru_tail);
    for (int i = 0; i < 4096; i++)
        printf("[%03x] next=%03x prev=%03x\n", i, c->dict[i].lru_next,
            c->dict[i].lru_prev);
}

static void lzw_validate_lru(struct lzw_ctx* c)
{
    for (uint16_t code = c->lru_head; code != 0; code = c->dict[code].lru_next) {
        printf("%03x", code);
        if (c->dict[code].usecount != 0)
            putchar('!');
        if (c->dict[code].lru_next != 0 && c->dict[c->dict[code].lru_next].lru_prev != code)
            putchar('>');
        putchar(' ');
    }
    putchar('\n');
}
#endif

static int lzw_build_entry(struct lzw_ctx* c, uint16_t code)
{
    uint16_t pos;

    pos = c->lru_head;
    lzw_lru_unlink(c, pos);

    if (c->dict[pos].prefix != 0) {
        c->dict[c->dict[pos].prefix].usecount--;
        if (c->dict[c->dict[pos].prefix].usecount == 0)
            lzw_lru_append(c, c->dict[pos].prefix);
    }

    c->dict[pos].prefix = c->last_emitted_code;
    c->dict[pos].last = c->dict[(code == pos ? c->dict[pos].prefix : code)].first;
    c->dict[pos].first = c->dict[c->dict[pos].prefix].first;
    c->dict[pos].length = c->dict[c->dict[pos].prefix].length + 1;
    c->dict[pos].usecount = 0;

    if (c->dict[c->dict[pos].prefix].usecount == 0)
        lzw_lru_unlink(c, c->dict[pos].prefix);
    c->dict[c->dict[pos].prefix].usecount++;
    lzw_lru_append(c, pos);

    return pos;
}

#define MIN(a, b) ((a)<(b)?(a):(b))

int lzw_decompress(struct lzw_ctx* c,
    uint8_t* src, size_t src_size, size_t* src_consumed,
    uint8_t* dst, size_t dst_size, size_t* dst_consumed)
{
    int bytes_written = 0;
    uint16_t code;
    uint16_t code_len, current_code;

    /* First, flush out any remaining data in the string buffer */
    /* NOLINTBEGIN(clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling) */
    if (c->output_buffer_used > 0) {
        bytes_written = MIN(c->output_buffer_used, dst_size - *dst_consumed);
        memcpy(&dst[*dst_consumed], &c->output_buffer[c->output_buffer_start], bytes_written);
        c->output_buffer_used -= bytes_written;
        c->output_buffer_start += bytes_written;
        c->uncompressed_bytes_processed += bytes_written;
        *dst_consumed += bytes_written;
    }
    if (*dst_consumed >= dst_size)
        return bytes_written;
    /* NOLINTEND(clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling) */
    while (*src_consumed < src_size && *dst_consumed < dst_size && c->eos == 0) {
        /* Get next 12bit code from the input buffer */
        while (c->input_code_bits < 12 && *src_consumed < src_size) {
            c->input_code = (c->input_code << 8) | src[(*src_consumed)++];
            c->input_code_bits += 8;
            c->compressed_bytes_processed++;
        }
        if (c->input_code_bits < 12)
            return bytes_written;
        code = (c->input_code >> (c->input_code_bits - 12)) & 0x0fff;
        c->input_code_bits -= 12;
        c->codes_processed++;
        if (code == 0) {
            c->eos = 1;
            return bytes_written;
        }
        /* Build the new dictionary entry */
        if (c->last_emitted_code != 0)
            lzw_build_entry(c, code);
        /* Output the corresponding string */
        c->output_buffer_start = 0;
        for (code_len = c->dict[code].length, current_code = code; current_code != 0 && code_len != 0; code_len--, current_code = c->dict[current_code].prefix) {
            if (code_len <= dst_size - *dst_consumed) {
                dst[*dst_consumed + code_len - 1] = c->dict[current_code].last;
                bytes_written++;
                c->uncompressed_bytes_processed++;
            } else {
                c->output_buffer[code_len - (dst_size - *dst_consumed) - 1] = c->dict[current_code].last;
                c->output_buffer_used++;
            }
        }
        if (c->dict[code].length < dst_size - *dst_consumed)
            *dst_consumed += c->dict[code].length;
        else
            *dst_consumed = dst_size;
        c->last_emitted_code = code;
    }

    return bytes_written;
}