view release on metacpan or  search on metacpan

ext/zstd/doc/educational_decoder/zstd_decompress.c  view on Meta::CPAN

            // "When Field_Size is 2, the offset of 256 is added."
            header->frame_content_size += 256;
        }
    } else {
        header->frame_content_size = 0;
    }

    if (single_segment_flag) {
        // "The Window_Descriptor byte is optional. It is absent when
        // Single_Segment_flag is set. In this case, the maximum back-reference
        // distance is the content size itself, which can be any value from 1 to
        // 2^64-1 bytes (16 EB)."
        header->window_size = header->frame_content_size;
    }
}

/// A dictionary acts as initializing values for the frame context before
/// decompression, so we implement it by applying it's predetermined
/// tables and content to the context before beginning decompression
static void frame_context_apply_dict(frame_context_t *const ctx,
                                     const dictionary_t *const dict) {
    // If the content pointer is NULL then it must be an empty dict
    if (!dict || !dict->content)
        return;

    // If the requested dictionary_id is non-zero, the correct dictionary must
    // be present
    if (ctx->header.dictionary_id != 0 &&
        ctx->header.dictionary_id != dict->dictionary_id) {
        ERROR("Wrong dictionary provided");
    }

    // Copy the dict content to the context for references during sequence
    // execution
    ctx->dict_content = dict->content;
    ctx->dict_content_len = dict->content_size;

    // If it's a formatted dict copy the precomputed tables in so they can
    // be used in the table repeat modes
    if (dict->dictionary_id != 0) {
        // Deep copy the entropy tables so they can be freed independently of
        // the dictionary struct
        HUF_copy_dtable(&ctx->literals_dtable, &dict->literals_dtable);
        FSE_copy_dtable(&ctx->ll_dtable, &dict->ll_dtable);
        FSE_copy_dtable(&ctx->of_dtable, &dict->of_dtable);
        FSE_copy_dtable(&ctx->ml_dtable, &dict->ml_dtable);

        // Copy the repeated offsets
        memcpy(ctx->previous_offsets, dict->previous_offsets,
               sizeof(ctx->previous_offsets));
    }
}

/// Decompress the data from a frame block by block
static void decompress_data(frame_context_t *const ctx, ostream_t *const out,
                            istream_t *const in) {
    // "A frame encapsulates one or multiple blocks. Each block can be
    // compressed or not, and has a guaranteed maximum content size, which
    // depends on frame parameters. Unlike frames, each block depends on
    // previous blocks for proper decoding. However, each block can be
    // decompressed without waiting for its successor, allowing streaming
    // operations."
    int last_block = 0;
    do {
        // "Last_Block
        //
        // The lowest bit signals if this block is the last one. Frame ends
        // right after this block.
        //
        // Block_Type and Block_Size
        //
        // The next 2 bits represent the Block_Type, while the remaining 21 bits
        // represent the Block_Size. Format is little-endian."
        last_block = IO_read_bits(in, 1);
        const int block_type = IO_read_bits(in, 2);
        const size_t block_len = IO_read_bits(in, 21);

        switch (block_type) {
        case 0: {
            // "Raw_Block - this is an uncompressed block. Block_Size is the
            // number of bytes to read and copy."
            const u8 *const read_ptr = IO_get_read_ptr(in, block_len);
            u8 *const write_ptr = IO_get_write_ptr(out, block_len);

            // Copy the raw data into the output
            memcpy(write_ptr, read_ptr, block_len);

            ctx->current_total_output += block_len;
            break;
        }
        case 1: {
            // "RLE_Block - this is a single byte, repeated N times. In which
            // case, Block_Size is the size to regenerate, while the
            // "compressed" block is just 1 byte (the byte to repeat)."
            const u8 *const read_ptr = IO_get_read_ptr(in, 1);
            u8 *const write_ptr = IO_get_write_ptr(out, block_len);

            // Copy `block_len` copies of `read_ptr[0]` to the output
            memset(write_ptr, read_ptr[0], block_len);

            ctx->current_total_output += block_len;
            break;
        }
        case 2: {
            // "Compressed_Block - this is a Zstandard compressed block,
            // detailed in another section of this specification. Block_Size is
            // the compressed size.

            // Create a sub-stream for the block
            istream_t block_stream = IO_make_sub_istream(in, block_len);
            decompress_block(ctx, out, &block_stream);
            break;
        }
        case 3:
            // "Reserved - this is not a block. This value cannot be used with
            // current version of this specification."
            CORRUPTION();
            break;
        default:
            IMPOSSIBLE();
        }