Modules/zlibmodule.c (part 6)

Source:

cpython 3.14 @ ab2d84fe1023/Modules/zlibmodule.c ↗

This annotation covers streaming decompression. See modules_zlib5_detail for zlib.compress, decompress, Compress.compress, and Compress.flush.

Map

Lines	Symbol	Role
1-80	Decompress.__init__	Set up decompression state with wbits
81-180	Decompress.decompress	Decompress a chunk
181-280	Decompress.flush	Finish decompression and return remaining output
281-360	Decompress.copy	Duplicate decompression state
361-500	wbits parameter	Control format: zlib, gzip, raw deflate

Reading

Decompress.decompress

// CPython: Modules/zlibmodule.c:680 zlib_Decompress_decompress_impl
static PyObject *
zlib_Decompress_decompress_impl(compobject *self, Py_buffer *data,
                                Py_ssize_t max_length)
{
    self->zst.next_in = (Bytef *)data->buf;
    self->zst.avail_in = (uInt)data->len;
    do {
        /* Expand output buffer as needed */
        arrange_input_buffer(&self->zst, &ibuflen);
        err = inflate(&self->zst, Z_SYNC_FLUSH);
        if (err == Z_NEED_DICT) {
            PyErr_SetString(ZlibError, "decompression requires a preset dictionary");
            return NULL;
        }
    } while (self->zst.avail_in != 0 && err != Z_STREAM_END);
    ...
}

inflate is called in a loop because a single call may not consume all input if the output buffer fills. Z_SYNC_FLUSH requests maximum output; the loop continues until all input is consumed.

wbits parameter

// CPython: Modules/zlibmodule.c:840 wbits meaning
/*
 * wbits:
 *   8 to 15     → zlib format (with zlib header)
 *  -8 to -15    → raw deflate (no header)
 *  24 to 31     → gzip format (wbits = 16 + window_bits)
 *  40 to 47     → auto-detect zlib or gzip (wbits = 32 + window_bits)
 *
 * zlib.decompress(data, wbits=-15) decompresses raw deflate data.
 * zlib.decompress(data, wbits=47) accepts either zlib or gzip.
 */

The wbits value controls both the window size (8-15 bits) and the expected header format. gzip.decompress internally uses wbits=47 to auto-detect headers.

Decompress.copy

// CPython: Modules/zlibmodule.c:780 zlib_Decompress_copy_impl
static PyObject *
zlib_Decompress_copy_impl(compobject *self)
{
    /* Duplicate the entire decompression state */
    compobject *ret_val = newcompobject(&Decomptype);
    if (inflateCopy(&ret_val->zst, &self->zst) != Z_OK) {
        PyErr_SetString(ZlibError, "couldn't copy object");
        return NULL;
    }
    return (PyObject *)ret_val;
}

Decompress.copy() creates an independent copy of the decompression state. This enables "checkpointing": decompress some data, copy the state, then explore multiple branches of decompressed data.

gopy notes

Decompress.decompress is module/zlib.DecompressDecompress in module/zlib/module.go. It wraps flate.NewReader or zlib.NewReader or gzip.NewReader depending on wbits. Streaming is handled via io.ReadAll into a growing bytes.Buffer. Decompress.copy deep-copies the decompressor state.