291 lines
11 KiB
Cython
291 lines
11 KiB
Cython
from libc.stdint cimport int64_t
|
|
from libc.stdlib cimport free, malloc
|
|
|
|
from av.codec.context cimport CodecContext, wrap_codec_context
|
|
from av.container.streams cimport StreamContainer
|
|
from av.dictionary cimport _Dictionary
|
|
from av.error cimport err_check
|
|
from av.packet cimport Packet
|
|
from av.stream cimport Stream, wrap_stream
|
|
from av.utils cimport avdict_to_dict
|
|
|
|
from av.dictionary import Dictionary
|
|
|
|
|
|
cdef close_input(InputContainer self):
|
|
self.streams = StreamContainer()
|
|
if self.input_was_opened:
|
|
with nogil:
|
|
# This causes `self.ptr` to be set to NULL.
|
|
lib.avformat_close_input(&self.ptr)
|
|
self.input_was_opened = False
|
|
|
|
|
|
cdef class InputContainer(Container):
|
|
def __cinit__(self, *args, **kwargs):
|
|
cdef CodecContext py_codec_context
|
|
cdef unsigned int i
|
|
cdef lib.AVStream *stream
|
|
cdef lib.AVCodec *codec
|
|
cdef lib.AVCodecContext *codec_context
|
|
|
|
# If we have either the global `options`, or a `stream_options`, prepare
|
|
# a mashup of those options for each stream.
|
|
cdef lib.AVDictionary **c_options = NULL
|
|
cdef _Dictionary base_dict, stream_dict
|
|
if self.options or self.stream_options:
|
|
base_dict = Dictionary(self.options)
|
|
c_options = <lib.AVDictionary**>malloc(self.ptr.nb_streams * sizeof(void*))
|
|
for i in range(self.ptr.nb_streams):
|
|
c_options[i] = NULL
|
|
if i < len(self.stream_options) and self.stream_options:
|
|
stream_dict = base_dict.copy()
|
|
stream_dict.update(self.stream_options[i])
|
|
lib.av_dict_copy(&c_options[i], stream_dict.ptr, 0)
|
|
else:
|
|
lib.av_dict_copy(&c_options[i], base_dict.ptr, 0)
|
|
|
|
self.set_timeout(self.open_timeout)
|
|
self.start_timeout()
|
|
with nogil:
|
|
# This peeks are the first few frames to:
|
|
# - set stream.disposition from codec.audio_service_type (not exposed);
|
|
# - set stream.codec.bits_per_coded_sample;
|
|
# - set stream.duration;
|
|
# - set stream.start_time;
|
|
# - set stream.r_frame_rate to average value;
|
|
# - open and closes codecs with the options provided.
|
|
ret = lib.avformat_find_stream_info(
|
|
self.ptr,
|
|
c_options
|
|
)
|
|
self.set_timeout(None)
|
|
self.err_check(ret)
|
|
|
|
# Cleanup all of our options.
|
|
if c_options:
|
|
for i in range(self.ptr.nb_streams):
|
|
lib.av_dict_free(&c_options[i])
|
|
free(c_options)
|
|
|
|
at_least_one_accelerated_context = False
|
|
|
|
self.streams = StreamContainer()
|
|
for i in range(self.ptr.nb_streams):
|
|
stream = self.ptr.streams[i]
|
|
codec = lib.avcodec_find_decoder(stream.codecpar.codec_id)
|
|
if codec:
|
|
# allocate and initialise decoder
|
|
codec_context = lib.avcodec_alloc_context3(codec)
|
|
err_check(lib.avcodec_parameters_to_context(codec_context, stream.codecpar))
|
|
codec_context.pkt_timebase = stream.time_base
|
|
py_codec_context = wrap_codec_context(codec_context, codec, self.hwaccel)
|
|
if py_codec_context.is_hwaccel:
|
|
at_least_one_accelerated_context = True
|
|
else:
|
|
# no decoder is available
|
|
py_codec_context = None
|
|
self.streams.add_stream(wrap_stream(self, stream, py_codec_context))
|
|
|
|
if self.hwaccel and not self.hwaccel.allow_software_fallback and not at_least_one_accelerated_context:
|
|
raise RuntimeError("Hardware accelerated decode requested but no stream is compatible")
|
|
|
|
self.metadata = avdict_to_dict(self.ptr.metadata, self.metadata_encoding, self.metadata_errors)
|
|
|
|
def __dealloc__(self):
|
|
close_input(self)
|
|
|
|
@property
|
|
def start_time(self):
|
|
self._assert_open()
|
|
if self.ptr.start_time != lib.AV_NOPTS_VALUE:
|
|
return self.ptr.start_time
|
|
|
|
@property
|
|
def duration(self):
|
|
self._assert_open()
|
|
if self.ptr.duration != lib.AV_NOPTS_VALUE:
|
|
return self.ptr.duration
|
|
|
|
@property
|
|
def bit_rate(self):
|
|
self._assert_open()
|
|
return self.ptr.bit_rate
|
|
|
|
@property
|
|
def size(self):
|
|
self._assert_open()
|
|
return lib.avio_size(self.ptr.pb)
|
|
|
|
def close(self):
|
|
close_input(self)
|
|
|
|
def demux(self, *args, **kwargs):
|
|
"""demux(streams=None, video=None, audio=None, subtitles=None, data=None)
|
|
|
|
Yields a series of :class:`.Packet` from the given set of :class:`.Stream`::
|
|
|
|
for packet in container.demux():
|
|
# Do something with `packet`, often:
|
|
for frame in packet.decode():
|
|
# Do something with `frame`.
|
|
|
|
.. seealso:: :meth:`.StreamContainer.get` for the interpretation of
|
|
the arguments.
|
|
|
|
.. note:: The last packets are dummy packets that when decoded will flush the buffers.
|
|
|
|
"""
|
|
self._assert_open()
|
|
|
|
# For whatever reason, Cython does not like us directly passing kwargs
|
|
# from one method to another. Without kwargs, it ends up passing a
|
|
# NULL reference, which segfaults. So we force it to do something with it.
|
|
# This is likely a bug in Cython; see https://github.com/cython/cython/issues/2166
|
|
# (and others).
|
|
id(kwargs)
|
|
|
|
streams = self.streams.get(*args, **kwargs)
|
|
|
|
cdef bint *include_stream = <bint*>malloc(self.ptr.nb_streams * sizeof(bint))
|
|
if include_stream == NULL:
|
|
raise MemoryError()
|
|
|
|
cdef unsigned int i
|
|
cdef Packet packet
|
|
cdef int ret
|
|
|
|
self.set_timeout(self.read_timeout)
|
|
try:
|
|
for i in range(self.ptr.nb_streams):
|
|
include_stream[i] = False
|
|
for stream in streams:
|
|
i = stream.index
|
|
if i >= self.ptr.nb_streams:
|
|
raise ValueError(f"stream index {i} out of range")
|
|
include_stream[i] = True
|
|
|
|
while True:
|
|
packet = Packet()
|
|
try:
|
|
self.start_timeout()
|
|
with nogil:
|
|
ret = lib.av_read_frame(self.ptr, packet.ptr)
|
|
self.err_check(ret)
|
|
except EOFError:
|
|
break
|
|
|
|
if include_stream[packet.ptr.stream_index]:
|
|
# If AVFMTCTX_NOHEADER is set in ctx_flags, then new streams
|
|
# may also appear in av_read_frame().
|
|
# http://ffmpeg.org/doxygen/trunk/structAVFormatContext.html
|
|
# TODO: find better way to handle this
|
|
if packet.ptr.stream_index < len(self.streams):
|
|
packet._stream = self.streams[packet.ptr.stream_index]
|
|
# Keep track of this so that remuxing is easier.
|
|
packet.ptr.time_base = packet._stream.ptr.time_base
|
|
yield packet
|
|
|
|
# Flush!
|
|
for i in range(self.ptr.nb_streams):
|
|
if include_stream[i]:
|
|
packet = Packet()
|
|
packet._stream = self.streams[i]
|
|
packet.ptr.time_base = packet._stream.ptr.time_base
|
|
yield packet
|
|
|
|
finally:
|
|
self.set_timeout(None)
|
|
free(include_stream)
|
|
|
|
def decode(self, *args, **kwargs):
|
|
"""decode(streams=None, video=None, audio=None, subtitles=None, data=None)
|
|
|
|
Yields a series of :class:`.Frame` from the given set of streams::
|
|
|
|
for frame in container.decode():
|
|
# Do something with `frame`.
|
|
|
|
.. seealso:: :meth:`.StreamContainer.get` for the interpretation of
|
|
the arguments.
|
|
|
|
"""
|
|
self._assert_open()
|
|
id(kwargs) # Avoid Cython bug; see demux().
|
|
for packet in self.demux(*args, **kwargs):
|
|
for frame in packet.decode():
|
|
yield frame
|
|
|
|
def seek(
|
|
self, offset, *, bint backward=True, bint any_frame=False, Stream stream=None,
|
|
bint unsupported_frame_offset=False, bint unsupported_byte_offset=False
|
|
):
|
|
"""seek(offset, *, backward=True, any_frame=False, stream=None)
|
|
|
|
Seek to a (key)frame nearsest to the given timestamp.
|
|
|
|
:param int offset: Time to seek to, expressed in``stream.time_base`` if ``stream``
|
|
is given, otherwise in :data:`av.time_base`.
|
|
:param bool backward: If there is not a (key)frame at the given offset,
|
|
look backwards for it.
|
|
:param bool any_frame: Seek to any frame, not just a keyframe.
|
|
:param Stream stream: The stream who's ``time_base`` the ``offset`` is in.
|
|
|
|
:param bool unsupported_frame_offset: ``offset`` is a frame
|
|
index instead of a time; not supported by any known format.
|
|
:param bool unsupported_byte_offset: ``offset`` is a byte
|
|
location in the file; not supported by any known format.
|
|
|
|
After seeking, packets that you demux should correspond (roughly) to
|
|
the position you requested.
|
|
|
|
In most cases, the defaults of ``backwards = True`` and ``any_frame = False``
|
|
are the best course of action, followed by you demuxing/decoding to
|
|
the position that you want. This is because to properly decode video frames
|
|
you need to start from the previous keyframe.
|
|
|
|
.. seealso:: :ffmpeg:`avformat_seek_file` for discussion of the flags.
|
|
|
|
"""
|
|
self._assert_open()
|
|
|
|
# We used to take floats here and assume they were in seconds. This
|
|
# was super confusing, so lets go in the complete opposite direction
|
|
# and reject non-ints.
|
|
if not isinstance(offset, int):
|
|
raise TypeError("Container.seek only accepts integer offset.", type(offset))
|
|
|
|
cdef int64_t c_offset = offset
|
|
|
|
cdef int flags = 0
|
|
cdef int ret
|
|
|
|
if backward:
|
|
flags |= lib.AVSEEK_FLAG_BACKWARD
|
|
if any_frame:
|
|
flags |= lib.AVSEEK_FLAG_ANY
|
|
|
|
# If someone really wants (and to experiment), expose these.
|
|
if unsupported_frame_offset:
|
|
flags |= lib.AVSEEK_FLAG_FRAME
|
|
if unsupported_byte_offset:
|
|
flags |= lib.AVSEEK_FLAG_BYTE
|
|
|
|
cdef int stream_index = stream.index if stream else -1
|
|
with nogil:
|
|
ret = lib.av_seek_frame(self.ptr, stream_index, c_offset, flags)
|
|
err_check(ret)
|
|
|
|
self.flush_buffers()
|
|
|
|
cdef flush_buffers(self):
|
|
self._assert_open()
|
|
|
|
cdef Stream stream
|
|
cdef CodecContext codec_context
|
|
|
|
for stream in self.streams:
|
|
codec_context = stream.codec_context
|
|
if codec_context:
|
|
codec_context.flush_buffers()
|