# This file is based on lzmaffi/_lzmamodule2.py from lzmaffi version 0.3.0.

# PyPy changes:
# - added __getstate__() methods that raise TypeError on pickling.
# - ported to CFFI 1.0

import threading
import functools
import collections
import weakref
import sys
import io
import __pypy__

from _lzma_cffi import ffi, lib as m

SUPPORTED_STREAM_FLAGS_VERSION = 0

__all__ = ['CHECK_CRC32',
 'CHECK_CRC64',
 'CHECK_ID_MAX',
 'CHECK_NONE',
 'CHECK_SHA256',
 'CHECK_UNKNOWN',
 'FILTER_ARM',
 'FILTER_ARMTHUMB',
 'FILTER_DELTA',
 'FILTER_IA64',
 'FILTER_LZMA1',
 'FILTER_LZMA2',
 'FILTER_POWERPC',
 'FILTER_SPARC',
 'FILTER_X86',
 'FORMAT_ALONE',
 'FORMAT_AUTO',
 'FORMAT_RAW',
 'FORMAT_XZ',
 'FORMAT_BLOCK',
 'LZMACompressor',
 'LZMADecompressor',
 'LZMAError',
 'MF_BT2',
 'MF_BT3',
 'MF_BT4',
 'MF_HC3',
 'MF_HC4',
 'MODE_FAST',
 'MODE_NORMAL',
 'PRESET_DEFAULT',
 'PRESET_EXTREME',
 'STREAM_HEADER_SIZE',
 'decode_block_header_size',
 'decode_stream_header',
 'decode_stream_footer',
 'decode_index',
 '_decode_filter_properties',
 '_encode_filter_properties',
 'is_check_supported']

_owns = weakref.WeakKeyDictionary()

def _new_lzma_stream():
    ret = ffi.new('lzma_stream*')
    m._pylzma_stream_init(ret)
    return ffi.gc(ret, m.lzma_end)

def _release_lzma_stream(st):
    ffi.gc(st, None)
    m.lzma_end(st)

def add_constant(c):
    globals()[c] = getattr(m, 'LZMA_' + c)

if sys.version_info >= (2,7):
    def to_bytes(data):
        return memoryview(data).tobytes()
else:
    def to_bytes(data):
        if not isinstance(data, basestring):
            raise TypeError("lzma: must be str/unicode, got %s" % (type(data),))
        return bytes(data)

if sys.version_info >= (3,0):
    long = int

for c in ['CHECK_CRC32', 'CHECK_CRC64', 'CHECK_ID_MAX', 'CHECK_NONE', 'CHECK_SHA256', 'FILTER_ARM', 'FILTER_ARMTHUMB', 'FILTER_DELTA', 'FILTER_IA64', 'FILTER_LZMA1', 'FILTER_LZMA2', 'FILTER_POWERPC', 'FILTER_SPARC', 'FILTER_X86', 'MF_BT2', 'MF_BT3', 'MF_BT4', 'MF_HC3', 'MF_HC4', 'MODE_FAST', 'MODE_NORMAL', 'PRESET_DEFAULT', 'PRESET_EXTREME', 'STREAM_HEADER_SIZE']:
    add_constant(c)

def _parse_format(format):
    if isinstance(format, (int, long)):
        return format
    else:
        raise TypeError

CHECK_UNKNOWN = CHECK_ID_MAX + 1
FORMAT_AUTO, FORMAT_XZ, FORMAT_ALONE, FORMAT_RAW, FORMAT_BLOCK = range(5)

BCJ_FILTERS = (m.LZMA_FILTER_X86,
    m.LZMA_FILTER_POWERPC,
    m.LZMA_FILTER_IA64,
    m.LZMA_FILTER_ARM,
    m.LZMA_FILTER_ARMTHUMB,
    m.LZMA_FILTER_SPARC)

class LZMAError(Exception):
    """Call to liblzma failed."""

def is_check_supported(check):
    """is_check_supported(check_id) -> bool
    
    Test whether the given integrity check is supported.
    
    Always returns True for CHECK_NONE and CHECK_CRC32."""
    return bool(m.lzma_check_is_supported(check))

def catch_lzma_error(fun, *args, ignore_buf_error=False):
    try:
        lzret = fun(*args)
    except:
        raise
    if lzret in (m.LZMA_OK, m.LZMA_GET_CHECK, m.LZMA_NO_CHECK, m.LZMA_STREAM_END):
        return lzret
    elif lzret == m.LZMA_DATA_ERROR:
        raise LZMAError("Corrupt input data")
    elif lzret == m.LZMA_UNSUPPORTED_CHECK:
        raise LZMAError("Unsupported integrity check")
    elif lzret == m.LZMA_FORMAT_ERROR:
        raise LZMAError("Input format not supported by decoder")
    elif lzret == m.LZMA_OPTIONS_ERROR:
        raise LZMAError("Invalid or unsupported options")
    elif lzret == m.LZMA_BUF_ERROR:
        if ignore_buf_error:
            return m.LZMA_OK
        raise LZMAError("Insufficient buffer space")
    elif lzret == m.LZMA_PROG_ERROR:
        raise LZMAError("Internal error")
    elif lzret == m.LZMA_MEM_ERROR:
        raise MemoryError
    else:
        raise LZMAError("Unrecognised error from liblzma: %d" % lzret)

def parse_filter_spec_delta(id, dist=1):
    ret = ffi.new('lzma_options_delta*')
    ret.type = m.LZMA_DELTA_TYPE_BYTE
    ret.dist = dist
    return ret

def parse_filter_spec_bcj(id, start_offset=0):
    ret = ffi.new('lzma_options_bcj*')
    ret.start_offset = start_offset
    return ret

def parse_filter_spec_lzma(id, preset=m.LZMA_PRESET_DEFAULT, **kwargs):
    ret = ffi.new('lzma_options_lzma*')
    if m.lzma_lzma_preset(ret, preset):
        raise LZMAError("Invalid compression preset: %s" % preset)
    for arg, val in kwargs.items():
        if arg in ('dict_size', 'lc', 'lp', 'pb', 'nice_len', 'depth'):
            setattr(ret, arg, val)
        elif arg in ('mf', 'mode'):
            setattr(ret, arg, int(val))
        else:
            raise ValueError("Invalid filter specifier for LZMA filter")
    return ret

def parse_filter_spec(spec):
    if not isinstance(spec, collections.Mapping):
        raise TypeError("Filter specifier must be a dict or dict-like object")
    ret = ffi.new('lzma_filter*')
    try:
        ret.id = spec['id']
    except KeyError:
        raise ValueError("Filter specifier must have an \"id\" entry")
    if ret.id in (m.LZMA_FILTER_LZMA1, m.LZMA_FILTER_LZMA2):
        try:
            options = parse_filter_spec_lzma(**spec)
        except TypeError:
            raise ValueError("Invalid filter specifier for LZMA filter")
    elif ret.id == m.LZMA_FILTER_DELTA:
        try:
            options = parse_filter_spec_delta(**spec)
        except TypeError:
            raise ValueError("Invalid filter specifier for delta filter")
    elif ret.id in BCJ_FILTERS:
        try:
            options = parse_filter_spec_bcj(**spec)
        except TypeError:
            raise ValueError("Invalid filter specifier for BCJ filter")
    else:
        raise ValueError("Invalid %d" % (ret.id,))

    ret.options = options
    _owns[ret] = options
    return ret

def _encode_filter_properties(filterspec):
    """_encode_filter_properties(filter) -> bytes

    Return a bytes object encoding the options (properties) of the filter
    specified by *filter* (a dict).

    The result does not include the filter ID itself, only the options."""
    filter = parse_filter_spec(filterspec)
    size = ffi.new("uint32_t*")
    catch_lzma_error(m.lzma_properties_size, size, filter)
    result = ffi.new('uint8_t[]', size[0])
    catch_lzma_error(m.lzma_properties_encode, filter, result)
    return ffi.buffer(result)[:]

def parse_filter_chain_spec(filterspecs):
    if len(filterspecs) > m.LZMA_FILTERS_MAX:
        raise ValueError(
            "Too many filters - liblzma supports a maximum of %s" %
            m.LZMA_FILTERS_MAX)
    filters = ffi.new('lzma_filter[]', m.LZMA_FILTERS_MAX+1)
    _owns[filters] = children = []
    for i in range(m.LZMA_FILTERS_MAX+1):
        try:
            filterspec = filterspecs[i]
        except KeyError:
            raise TypeError
        except IndexError:
            filters[i].id = m.LZMA_VLI_UNKNOWN
        else:
            filter = parse_filter_spec(filterspecs[i])
            children.append(filter)
            filters[i].id = filter.id
            filters[i].options = filter.options
    return filters

def build_filter_spec(filter):
    spec = {'id': filter.id}
    def add_opts(options_type, *opts):
        options = ffi.cast('%s*' % (options_type,), filter.options)
        for v in opts:
            spec[v] = getattr(options, v)
    if filter.id == m.LZMA_FILTER_LZMA1:
        add_opts('lzma_options_lzma', 'lc', 'lp', 'pb', 'dict_size')
    elif filter.id == m.LZMA_FILTER_LZMA2:
        add_opts('lzma_options_lzma', 'dict_size')
    elif filter.id == m.LZMA_FILTER_DELTA:
        add_opts('lzma_options_delta', 'dist')
    elif filter.id in BCJ_FILTERS:
        add_opts('lzma_options_bcj', 'start_offset')
    else:
        raise ValueError("Invalid filter ID: %s" % filter.id)
    return spec

def _decode_filter_properties(filter_id, encoded_props):
    """_decode_filter_properties(filter_id, encoded_props) -> dict

    Return a dict describing a filter with ID *filter_id*, and options
    (properties) decoded from the bytes object *encoded_props*."""
    filter = ffi.new('lzma_filter*')
    filter.id = filter_id
    catch_lzma_error(m.lzma_properties_decode,
        filter, ffi.NULL, encoded_props, len(encoded_props))
    try:
        return build_filter_spec(filter)
    finally:
        # TODO do we need this, the only use of m.free?
        m.free(filter.options)

def _decode_stream_header_or_footer(decode_f, in_bytes):
    footer_o = ffi.new('char[]', to_bytes(in_bytes))
    stream_flags = ffi.new('lzma_stream_flags*')
    catch_lzma_error(decode_f, stream_flags, footer_o)
    return StreamFlags(stream_flags)

decode_stream_footer = functools.partial(_decode_stream_header_or_footer,
    m.lzma_stream_footer_decode)

decode_stream_header = functools.partial(_decode_stream_header_or_footer,
    m.lzma_stream_header_decode)

def decode_block_header_size(in_byte):
    # lzma_block_header_size_decode(b) (((uint32_t)(b) + 1) * 4)
    return (ord(in_byte) + 1) * 4

def decode_index(s, stream_padding=0):
    indexp = ffi.new('lzma_index**')
    memlimit = ffi.new('uint64_t*')
    memlimit[0] = m.UINT64_MAX
    allocator = ffi.NULL
    in_buf = ffi.new('char[]', to_bytes(s))
    in_pos = ffi.new('size_t*')
    in_pos[0] = 0
    catch_lzma_error(m.lzma_index_buffer_decode, indexp,
        memlimit, allocator, in_buf, in_pos, len(s))
    return Index(indexp[0], allocator, stream_padding)

class Index(object):
    def __init__(self, i, allocator, stream_padding=0):
        self.i = i
        self.allocator = allocator
        m.lzma_index_stream_padding(i, stream_padding)

    @property
    def uncompressed_size(self):
        return m.lzma_index_uncompressed_size(self.i)

    @property
    def block_count(self):
        return m.lzma_index_block_count(self.i)

    @property
    def index_size(self):
        return m.lzma_index_size(self.i)

    @property
    def blocks_size(self):
        return m.lzma_index_total_size(self.i)

    def __iter__(self):
        return self.iterator()

    def iterator(self, type=m.LZMA_INDEX_ITER_BLOCK):
        iterator = ffi.new('lzma_index_iter*')
        m.lzma_index_iter_init(iterator, self.i)
        while not m.lzma_index_iter_next(iterator, type):
            yield (IndexStreamData(iterator.stream), IndexBlockData(iterator.block))

    def find(self, offset):
        iterator = ffi.new('lzma_index_iter*')
        m.lzma_index_iter_init(iterator, self.i)
        if m.lzma_index_iter_locate(iterator, offset):
            # offset too high
            return None
        return (IndexStreamData(iterator.stream), IndexBlockData(iterator.block))

    def __del__(self):
        m.lzma_index_end(self.i, self.allocator)

    def copy(self):
        new_i = m.lzma_index_dup(self.i, self.allocator)
        return Index(new_i, self.allocator)

    deepcopy = copy

    def append(self, other_index):
        # m.lzma_index_cat frees its second parameter so we
        # must copy it first
        other_index_i = m.lzma_index_dup(other_index.i, self.allocator)
        catch_lzma_error(m.lzma_index_cat, self.i, 
            other_index_i, self.allocator)

class _StructToPy(object):
    __slots__ = ()
    def __init__(self, struct_obj):
        # TODO make PyPy-fast
        for attr in self.__slots__:
            setattr(self, attr, getattr(struct_obj, attr))
    def __repr__(self):
        descriptions = ('%s=%r' % (attr, getattr(self, attr)) for attr in self.__slots__)
        return "<%s %s>" % (type(self).__name__, ' '.join(descriptions))

class IndexStreamData(_StructToPy):
    __slots__ = ('number', 'block_count', 'compressed_offset', 'uncompressed_offset',
        'compressed_size', 'uncompressed_size')

class IndexBlockData(_StructToPy):
    __slots__ = ('number_in_file', 'compressed_file_offset', 'uncompressed_file_offset',
        'compressed_stream_offset', 'uncompressed_stream_offset',
        'uncompressed_size', 'unpadded_size', 'total_size')

class StreamFlags(object):
    def __init__(self, i):
        self.i = i

    version = property(lambda self: self.i.version)
    check = property(lambda self: self.i.check)
    backward_size = property(lambda self: self.i.backward_size)

    @property
    def supported(self):
        return self.version > SUPPORTED_STREAM_FLAGS_VERSION

    def check_supported(self):
        if not self.supported:
            raise LZMAError("Stream is too new for liblzma version")

    def matches(self, other):
        return m.lzma_stream_flags_compare(self.i, other.i) == m.LZMA_OK

    def copy(self):
        other_i = ffi.new('lzma_stream_flags*', self.i)
        return StreamFlags(other_i)

class Allocator(object):
    def __init__(self):
        self.owns = {}
        self.lzma_allocator = ffi.new('lzma_allocator*')
        alloc = self.owns['a'] = ffi.callback("void*(void*, size_t, size_t)", self.__alloc)
        free = self.owns['b'] = ffi.callback("void(void*, void*)", self.__free)
        self.lzma_allocator.alloc = alloc
        self.lzma_allocator.free = free
        self.lzma_allocator.opaque = ffi.NULL
    def __alloc(self, _opaque, _nmemb, size):
        new_mem = ffi.new('char[]', size)
        self.owns[self._addr(new_mem)] = new_mem
        return new_mem
    def _addr(self, ptr):
        return long(ffi.cast('uintptr_t', ptr))
    def __free(self, _opaque, ptr):
        if self._addr(ptr) == 0: return
        del self.owns[self._addr(ptr)]

class LZMADecompressor(object):
    """
    LZMADecompressor(format=FORMAT_AUTO, memlimit=None, filters=None)

    Create a decompressor object for decompressing data incrementally.

    format specifies the container format of the input stream. If this is
    FORMAT_AUTO (the default), the decompressor will automatically detect
    whether the input is FORMAT_XZ or FORMAT_ALONE. Streams created with
    FORMAT_RAW cannot be autodetected.

    memlimit can be specified to limit the amount of memory used by the
    decompressor. This will cause decompression to fail if the input
    cannot be decompressed within the given limit.

    filters specifies a custom filter chain. This argument is required for
    FORMAT_RAW, and not accepted with any other format. When provided,
    this should be a sequence of dicts, each indicating the ID and options
    for a single filter.

    For one-shot decompression, use the decompress() function instead.
    """
    def __init__(self, format=FORMAT_AUTO, memlimit=None, filters=None,
                 header=None, check=None, unpadded_size=None):
        decoder_flags = m.LZMA_TELL_ANY_CHECK | m.LZMA_TELL_NO_CHECK
        if memlimit is not None:
            if format == FORMAT_RAW:
                raise ValueError("Cannot specify memory limit with FORMAT_RAW")
        else:
            memlimit = m.UINT64_MAX

        if format == FORMAT_RAW and filters is None:
            raise ValueError("Must specify filters for FORMAT_RAW")
        elif format != FORMAT_RAW and filters is not None:
            raise ValueError("Cannot specify filters except with FORMAT_RAW")

        if format == FORMAT_BLOCK and (header is None or unpadded_size is None or check is None):
            raise ValueError("Must specify header, unpadded_size and check "
                             "with FORMAT_BLOCK")
        elif format != FORMAT_BLOCK and (header is not None or unpadded_size is not None or check is not None):
            raise ValueError("Cannot specify header, unpadded_size or check "
                             "except with FORMAT_BLOCK")

        format = _parse_format(format)
        self.lock = threading.Lock()
        self.check = CHECK_UNKNOWN
        self.unused_data = b''
        self.eof = False
        self.lzs = _new_lzma_stream()
        self._bufsiz = max(8192, io.DEFAULT_BUFFER_SIZE)
        self.needs_input = True
        self._input_buffer = ffi.NULL
        self._input_buffer_size = 0

        if format == FORMAT_AUTO:
            catch_lzma_error(m.lzma_auto_decoder, self.lzs, memlimit, decoder_flags)
        elif format == FORMAT_XZ:
            catch_lzma_error(m.lzma_stream_decoder, self.lzs, memlimit, decoder_flags)
        elif format == FORMAT_ALONE:
            self.check = CHECK_NONE
            catch_lzma_error(m.lzma_alone_decoder, self.lzs, memlimit)
        elif format == FORMAT_RAW:
            self.check = CHECK_NONE
            filters = parse_filter_chain_spec(filters)
            catch_lzma_error(m.lzma_raw_decoder, self.lzs,
                filters)
        elif format == FORMAT_BLOCK:
            self.__block = block = ffi.new('lzma_block*')
            block.version = 0
            block.check = check
            block.header_size = len(header)
            block.filters = self.__filters = ffi.new('lzma_filter[]', m.LZMA_FILTERS_MAX+1)
            header_b = ffi.new('char[]', to_bytes(header))
            catch_lzma_error(m.lzma_block_header_decode, block, self.lzs.allocator, header_b)
            if unpadded_size is not None:
                catch_lzma_error(m.lzma_block_compressed_size, block, unpadded_size)
            self.expected_size = block.compressed_size
            catch_lzma_error(m.lzma_block_decoder, self.lzs, block)
        else:
            raise ValueError("invalid container format: %s" % format)

    def pre_decompress_left_data(self, buf, buf_size):
        # in this case there is data left that needs to be processed before the first
        # argument can be processed

        lzs = self.lzs

        addr_input_buffer = int(ffi.cast('uintptr_t', self._input_buffer))
        addr_next_in = int(ffi.cast('uintptr_t', lzs.next_in))
        avail_now = (addr_input_buffer + self._input_buffer_size) - \
                    (addr_next_in + lzs.avail_in)
        avail_total = self._input_buffer_size - lzs.avail_in
        if avail_total < buf_size:
            # resize the buffer, it is too small!
            offset = addr_next_in - addr_input_buffer
            new_size = self._input_buffer_size + buf_size - avail_now
            # there is no realloc?
            tmp = ffi.cast("uint8_t*",m.malloc(new_size))
            if tmp == ffi.NULL:
                raise MemoryError
            ffi.memmove(tmp, lzs.next_in, lzs.avail_in)
            lzs.next_in = tmp
            m.free(self._input_buffer)
            self._input_buffer = tmp
            self._input_buffer_size = new_size
        elif avail_now < buf_size:
            # the buffer is not too small, but we cannot append it!
            # move all data to the front
            ffi.memmove(self._input_buffer, lzs.next_in, lzs.avail_in)
            lzs.next_in = self._input_buffer
        ffi.memmove(lzs.next_in+lzs.avail_in, buf, buf_size)
        lzs.avail_in += buf_size
        return lzs.next_in, lzs.avail_in

    def post_decompress_avail_data(self):
        lzs = self.lzs
        # free buffer it is to small
        if self._input_buffer is not ffi.NULL and \
           self._input_buffer_size < lzs.avail_in:
            m.free(self._input_buffer)
            self._input_buffer = ffi.NONE

        # allocate if necessary
        if self._input_buffer is ffi.NULL:
            self._input_buffer = ffi.cast("uint8_t*",m.malloc(lzs.avail_in))
            if self._input_buffer == ffi.NULL:
                raise MemoryError
            self._input_buffer_size = lzs.avail_in

        ffi.memmove(self._input_buffer, lzs.next_in, lzs.avail_in)
        lzs.next_in = self._input_buffer

    def clear_input_buffer(self):
        # clean the buffer
        if self._input_buffer is not ffi.NULL:
            m.free(self._input_buffer)
            self._input_buffer = ffi.NULL
            self._input_buffer_size = 0

    def decompress(self, data, max_length=-1):
        """
        decompress(data, max_length=-1) -> bytes

        Provide data to the decompressor object. Returns a chunk of
        decompressed data if possible, or b"" otherwise.

        Attempting to decompress data after the end of the stream is
        reached raises an EOFError. Any data found after the end of the
        stream is ignored, and saved in the unused_data attribute.
        """
        if not isinstance(max_length, int):
            raise TypeError("max_length parameter object cannot be interpreted as an integer")
        with self.lock:
            if self.eof:
                raise EOFError("Already at end of stream")
            lzs = self.lzs
            data = to_bytes(data)
            buf = ffi.new('uint8_t[]', data)
            buf_size = len(data)

            if lzs.next_in:
                buf, buf_size = self.pre_decompress_left_data(buf, buf_size)
                used__input_buffer = True
            else:
                lzs.avail_in = buf_size
                lzs.next_in = ffi.cast("uint8_t*",buf)
                used__input_buffer = False

            # actual decompression
            result = self._decompress(buf, buf_size, max_length)

            if self.eof:
                self.needs_input = False
                if lzs.avail_in > 0:
                    self.unused_data = ffi.buffer(lzs.next_in, lzs.avail_in)[:]
                self.clear_input_buffer()
            elif lzs.avail_in == 0:
                # completed successfully!
                lzs.next_in = ffi.NULL
                if lzs.avail_out == 0:
                    # (avail_in==0 && avail_out==0)
                    # Maybe lzs's internal state still have a few bytes can
                    # be output, try to output them next time.
                    self.needs_input = False
                    assert max_length >= 0   # if < 0, lzs.avail_out always > 0
                else:
                    # Input buffer exhausted, output buffer has space.
                    self.needs_input = True
                self.clear_input_buffer()
            else:
                self.needs_input = False
                if not used__input_buffer:
                    self.post_decompress_avail_data()

            return result

    def _decompress(self, buf, buf_len, max_length):
        lzs = self.lzs

        lzs.next_in = buf
        lzs.avail_in = buf_len

        bufsiz = self._bufsiz
        if not (max_length < 0 or max_length > io.DEFAULT_BUFFER_SIZE):
            bufsiz = max_length

        lzs.next_out = orig_out = m.malloc(bufsiz)
        if orig_out == ffi.NULL:
            raise MemoryError

        lzs.avail_out = bufsiz

        data_size = 0

        try:
            while True:
                ret = catch_lzma_error(m.lzma_code, lzs, m.LZMA_RUN,
                    ignore_buf_error=(lzs.avail_in == 0 and lzs.avail_out > 0))
                data_size = int(ffi.cast('uintptr_t', lzs.next_out)) - int(ffi.cast('uintptr_t', orig_out))
                # data_size is the amount lzma_code has already outputted

                if ret in (m.LZMA_NO_CHECK, m.LZMA_GET_CHECK):
                    self.check = m.lzma_get_check(lzs)

                if ret == m.LZMA_STREAM_END:
                    self.eof = True
                    break
                elif lzs.avail_out == 0:
                    # Need to check lzs->avail_out before lzs->avail_in.
                    # Maybe lzs's internal state still have a few bytes
                    # can be output, grow the output buffer and continue
                    # if max_lengh < 0.
                    if data_size == max_length:
                        break
                    # ran out of space in the output buffer, let's grow it
                    bufsiz += (bufsiz >> 3) + 6
                    if max_length > 0 and bufsiz > max_length:
                        bufsiz = max_length
                    next_out = m.realloc(orig_out, bufsiz)
                    if next_out == ffi.NULL:
                        # realloc unsuccessful
                        m.free(orig_out)
                        orig_out = ffi.NULL
                        raise MemoryError

                    orig_out = next_out

                    lzs.next_out = orig_out + data_size
                    lzs.avail_out = bufsiz - data_size
                elif lzs.avail_in == 0:
                    # it ate everything
                    break

            result = ffi.buffer(orig_out, data_size)[:]
        finally:
            m.free(orig_out)

        return result

    def __getstate__(self):
        raise TypeError("cannot serialize '%s' object" %
                        self.__class__.__name__)


# Issue #2579: Setting up the stream for encoding takes around 17MB of
# RAM on my Linux 64 system.  So we call add_memory_pressure(17MB) when
# we create the stream.  In flush(), we actively free the stream even
# though we could just leave it to the GC (but 17MB is too much for
# doing that sanely); at this point we call add_memory_pressure(-17MB)
# to cancel the original increase.
COMPRESSION_STREAM_SIZE = 1024*1024*17


class LZMACompressor(object):
    """
    LZMACompressor(format=FORMAT_XZ, check=-1, preset=None, filters=None)

    Create a compressor object for compressing data incrementally.

    format specifies the container format to use for the output. This can
    be FORMAT_XZ (default), FORMAT_ALONE, or FORMAT_RAW.

    check specifies the integrity check to use. For FORMAT_XZ, the default
    is CHECK_CRC64. FORMAT_ALONE and FORMAT_RAW do not suport integrity
    checks; for these formats, check must be omitted, or be CHECK_NONE.

    The settings used by the compressor can be specified either as a
    preset compression level (with the 'preset' argument), or in detail
    as a custom filter chain (with the 'filters' argument). For FORMAT_XZ
    and FORMAT_ALONE, the default is to use the PRESET_DEFAULT preset
    level. For FORMAT_RAW, the caller must always specify a filter chain;
    the raw compressor does not support preset compression levels.

    preset (if provided) should be an integer in the range 0-9, optionally
    OR-ed with the constant PRESET_EXTREME.

    filters (if provided) should be a sequence of dicts. Each dict should
    have an entry for "id" indicating the ID of the filter, plus
    additional entries for options to the filter.

    For one-shot compression, use the compress() function instead.
    """
    def __init__(self, format=FORMAT_XZ, check=-1, preset=None, filters=None):
        if format != FORMAT_XZ and check not in (-1, m.LZMA_CHECK_NONE):
            raise ValueError("Integrity checks are only supported by FORMAT_XZ")
        if preset is not None and filters is not None:
            raise ValueError("Cannot specify both preset and filter chain")
        if preset is None:
            preset = m.LZMA_PRESET_DEFAULT
        format = _parse_format(format)
        self.lock = threading.Lock()
        self.flushed = 0
        self.lzs = _new_lzma_stream()
        __pypy__.add_memory_pressure(COMPRESSION_STREAM_SIZE)
        if format == FORMAT_XZ:
            if filters is None:
                if check == -1:
                    check = m.LZMA_CHECK_CRC64
                catch_lzma_error(m.lzma_easy_encoder, self.lzs,
                    preset, check)
            else:
                filters = parse_filter_chain_spec(filters)
                catch_lzma_error(m.lzma_stream_encoder, self.lzs,
                    filters, check)
        elif format == FORMAT_ALONE:
            if filters is None:
                options = ffi.new('lzma_options_lzma*')
                if m.lzma_lzma_preset(options, preset):
                    raise LZMAError("Invalid compression preset: %s" % preset)
                catch_lzma_error(m.lzma_alone_encoder, self.lzs,
                    options)
            else:
                raise NotImplementedError
        elif format == FORMAT_RAW:
            if filters is None:
                raise ValueError("Must specify filters for FORMAT_RAW")
            filters = parse_filter_chain_spec(filters)
            catch_lzma_error(m.lzma_raw_encoder, self.lzs,
                filters)
        else:
            raise ValueError("invalid container format: %s" % format)

    def compress(self, data):
        """
        compress(data) -> bytes

        Provide data to the compressor object. Returns a chunk of
        compressed data if possible, or b"" otherwise.

        When you have finished providing data to the compressor, call the
        flush() method to finish the conversion process.
        """
        with self.lock:
            if self.flushed:
                raise ValueError("Compressor has been flushed")
            return self._compress(data)

    def _compress(self, data, action=m.LZMA_RUN):
        # TODO use realloc like in LZMADecompressor
        BUFSIZ = 8192

        lzs = self.lzs

        lzs.next_in = input_ = ffi.new('uint8_t[]', to_bytes(data))
        lzs.avail_in = input_len = len(data)
        outs = [ffi.new('uint8_t[]', BUFSIZ)]
        lzs.next_out, = outs
        lzs.avail_out = BUFSIZ

        siz = BUFSIZ

        while True:
            next_out_pos = int(ffi.cast('intptr_t', lzs.next_out))
            ret = catch_lzma_error(m.lzma_code, lzs, action,
                      ignore_buf_error=(input_len==0 and lzs.avail_out > 0))
            data_size = int(ffi.cast('intptr_t', lzs.next_out)) - next_out_pos
            if (action == m.LZMA_RUN and lzs.avail_in == 0) or \
                (action == m.LZMA_FINISH and ret == m.LZMA_STREAM_END):
                break
            elif lzs.avail_out == 0:
                # ran out of space in the output buffer
                #siz = (BUFSIZ << 1) + 6
                siz = 512
                outs.append(ffi.new('uint8_t[]', siz))
                lzs.next_out = outs[-1]
                lzs.avail_out = siz
        last_out = outs.pop()
        last_out_len = siz - lzs.avail_out
        last_out_piece = ffi.buffer(last_out[0:last_out_len], last_out_len)[:]

        return b''.join(ffi.buffer(nn)[:] for nn in outs) + last_out_piece

    def flush(self):
        with self.lock:
            if self.flushed:
                raise ValueError("Repeated call to flush()")
            self.flushed = 1
            result = self._compress(b'', action=m.LZMA_FINISH)
            __pypy__.add_memory_pressure(-COMPRESSION_STREAM_SIZE)
            _release_lzma_stream(self.lzs)
        return result

    def __getstate__(self):
        raise TypeError("cannot serialize '%s' object" %
                        self.__class__.__name__)