Build with ffmpeg 7

.github/workflows/smoke.yml

@@ -38,8 +38,10 @@ jobs:
       matrix:
         config:
           - {os: ubuntu-latest, python: "3.8", ffmpeg: "6.1", extras: true}
+          # TODO: - {os: ubuntu-latest, python: "3.8", ffmpeg: "7.0"}
           - {os: ubuntu-latest, python: "3.8", ffmpeg: "6.0"}
           - {os: ubuntu-latest, python: pypy3.9, ffmpeg: "6.1"}
+          - {os: macos-12,  python: "3.8", ffmpeg: "7.0"}
           - {os: macos-12,  python: "3.8", ffmpeg: "6.1"}
 
     env:
@@ -116,6 +118,7 @@ jobs:
       fail-fast: false
       matrix:
         config:
+          - {os: windows-latest, python: "3.8", ffmpeg: "7.0"}
           - {os: windows-latest, python: "3.8", ffmpeg: "6.1"}
           - {os: windows-latest, python: "3.8", ffmpeg: "6.0"}
 

av/audio/codeccontext.pxd

@@ -5,9 +5,7 @@ from av.codec.context cimport CodecContext
 
 
 cdef class AudioCodecContext(CodecContext):
-
     # Hold onto the frames that we will decode until we have a full one.
     cdef AudioFrame next_frame
-
     # For encoding.
     cdef AudioResampler resampler

av/audio/codeccontext.pyi

@@ -11,12 +11,12 @@ class AudioCodecContext(CodecContext):
     frame_size: int
     sample_rate: int
     rate: int
-    channels: int
-    channel_layout: int
     layout: AudioLayout
     format: AudioFormat
     type: Literal["audio"]
 
+    @property
+    def channels(self) -> int: ...
     def encode(self, frame: AudioFrame | None = None) -> list[Packet]: ...
     def encode_lazy(self, frame: AudioFrame | None = None) -> Iterator[Packet]: ...
     def decode(self, packet: Packet | None = None) -> list[AudioFrame]: ...

av/audio/codeccontext.pyx

@@ -11,13 +11,6 @@ cdef class AudioCodecContext(CodecContext):
     cdef _init(self, lib.AVCodecContext *ptr, const lib.AVCodec *codec):
         CodecContext._init(self, ptr, codec)
 
-        # Sometimes there isn't a layout set, but there are a number of
-        # channels. Assume it is the default layout.
-        # TODO: Put this behind `not bare_metal`.
-        # TODO: Do this more efficiently.
-        if self.ptr.channels and not self.ptr.channel_layout:
-            self.ptr.channel_layout = get_audio_layout(self.ptr.channels, 0).layout
-
     cdef _set_default_time_base(self):
         self.ptr.time_base.num = 1
         self.ptr.time_base.den = self.ptr.sample_rate
@@ -62,7 +55,6 @@ cdef class AudioCodecContext(CodecContext):
         """
         return self.ptr.frame_size
 
-
     @property
     def sample_rate(self):
         """
@@ -85,18 +77,9 @@ cdef class AudioCodecContext(CodecContext):
     def rate(self, value):
         self.sample_rate = value
 
-    # TODO: Integrate into AudioLayout.
     @property
     def channels(self):
-        return self.ptr.channels
-
-    @channels.setter
-    def channels(self, value):
-        self.ptr.channels = value
-        self.ptr.channel_layout = lib.av_get_default_channel_layout(value)
-    @property
-    def channel_layout(self):
-        return self.ptr.channel_layout
+        return self.layout.nb_channels
 
     @property
     def layout(self):
@@ -105,13 +88,12 @@ cdef class AudioCodecContext(CodecContext):
 
         :type: AudioLayout
         """
-        return get_audio_layout(self.ptr.channels, self.ptr.channel_layout)
+        return get_audio_layout(self.ptr.ch_layout)
 
     @layout.setter
     def layout(self, value):
         cdef AudioLayout layout = AudioLayout(value)
-        self.ptr.channel_layout = layout.layout
-        self.ptr.channels = layout.nb_channels
+        self.ptr.ch_layout = layout.layout
 
     @property
     def format(self):

av/audio/fifo.pyx

@@ -61,7 +61,7 @@ cdef class AudioFifo:
 
             self.ptr = lib.av_audio_fifo_alloc(
                 <lib.AVSampleFormat>frame.ptr.format,
-                len(frame.layout.channels),  # TODO: Can we safely use frame.ptr.nb_channels?
+                frame.layout.nb_channels,
                 frame.ptr.nb_samples * 2,  # Just a default number of samples; it will adjust.
             )
 
@@ -71,7 +71,7 @@ cdef class AudioFifo:
         # Make sure nothing changed.
         elif (
             frame.ptr.format != self.template.ptr.format or
-            frame.ptr.channel_layout != self.template.ptr.channel_layout or
+            # TODO: frame.ptr.ch_layout != self.template.ptr.ch_layout or
             frame.ptr.sample_rate != self.template.ptr.sample_rate or
             (frame._time_base.num and self.template._time_base.num and (
                 frame._time_base.num != self.template._time_base.num or
@@ -131,7 +131,7 @@ cdef class AudioFifo:
         frame._copy_internal_attributes(self.template)
         frame._init(
             <lib.AVSampleFormat>self.template.ptr.format,
-            self.template.ptr.channel_layout,
+            <lib.AVChannelLayout>self.template.ptr.ch_layout,
             samples,
             1,  # Align?
         )

av/audio/frame.pxd

@@ -7,7 +7,6 @@ from av.frame cimport Frame
 
 
 cdef class AudioFrame(Frame):
-
     # For raw storage of the frame's data; don't ever touch this.
     cdef uint8_t *_buffer
     cdef size_t _buffer_size
@@ -26,7 +25,7 @@ cdef class AudioFrame(Frame):
     :type: AudioFormat
     """
 
-    cdef _init(self, lib.AVSampleFormat format, uint64_t layout, unsigned int nb_samples, unsigned int align)
+    cdef _init(self, lib.AVSampleFormat format, lib.AVChannelLayout layout, unsigned int nb_samples, unsigned int align)
     cdef _init_user_attributes(self)
 
 cdef AudioFrame alloc_audio_frame()

av/audio/frame.pyx

@@ -4,10 +4,6 @@ from av.audio.plane cimport AudioPlane
 from av.error cimport err_check
 from av.utils cimport check_ndarray, check_ndarray_shape
 
-import warnings
-
-from av.deprecation import AVDeprecationWarning
-
 
 cdef object _cinit_bypass_sentinel
 
@@ -47,26 +43,22 @@ cdef class AudioFrame(Frame):
         cdef AudioLayout cy_layout = AudioLayout(layout)
         self._init(cy_format.sample_fmt, cy_layout.layout, samples, align)
 
-    cdef _init(self, lib.AVSampleFormat format, uint64_t layout, unsigned int nb_samples, unsigned int align):
-
+    cdef _init(self, lib.AVSampleFormat format, lib.AVChannelLayout layout, unsigned int nb_samples, unsigned int align):
         self.ptr.nb_samples = nb_samples
         self.ptr.format = <int>format
-        self.ptr.channel_layout = layout
+        self.ptr.ch_layout = layout
 
         # Sometimes this is called twice. Oh well.
         self._init_user_attributes()
 
-        # Audio filters need AVFrame.channels to match number of channels from layout.
-        self.ptr.channels = self.layout.nb_channels
-
-        if self.layout.channels and nb_samples:
+        if self.layout.nb_channels != 0 and nb_samples:
             # Cleanup the old buffer.
             lib.av_freep(&self._buffer)
 
             # Get a new one.
             self._buffer_size = err_check(lib.av_samples_get_buffer_size(
                 NULL,
-                len(self.layout.channels),
+                self.layout.nb_channels,
                 nb_samples,
                 format,
                 align
@@ -78,7 +70,7 @@ cdef class AudioFrame(Frame):
             # Connect the data pointers to the buffer.
             err_check(lib.avcodec_fill_audio_frame(
                 self.ptr,
-                len(self.layout.channels),
+                self.layout.nb_channels,
                 <lib.AVSampleFormat>self.ptr.format,
                 self._buffer,
                 self._buffer_size,
@@ -89,7 +81,7 @@ cdef class AudioFrame(Frame):
         lib.av_freep(&self._buffer)
 
     cdef _init_user_attributes(self):
-        self.layout = get_audio_layout(0, self.ptr.channel_layout)
+        self.layout = get_audio_layout(self.ptr.ch_layout)
         self.format = get_audio_format(<lib.AVSampleFormat>self.ptr.format)
 
     def __repr__(self):
@@ -114,7 +106,7 @@ cdef class AudioFrame(Frame):
             )
 
         # check input format
-        nb_channels = len(AudioLayout(layout).channels)
+        nb_channels = AudioLayout(layout).nb_channels
         check_ndarray(array, dtype, 2)
         if AudioFormat(format).is_planar:
             check_ndarray_shape(array, array.shape[0] == nb_channels)
@@ -188,6 +180,6 @@ cdef class AudioFrame(Frame):
         if self.format.is_planar:
             count = self.samples
         else:
-            count = self.samples * len(self.layout.channels)
+            count = self.samples * self.layout.nb_channels
 
         return np.vstack([np.frombuffer(x, dtype=dtype, count=count) for x in self.planes])

av/audio/layout.pxd

@@ -1,26 +1,8 @@
-from libc.stdint cimport uint64_t
+cimport libav as lib
 
 
 cdef class AudioLayout:
+    cdef lib.AVChannelLayout layout
+    cdef _init(self, lib.AVChannelLayout layout)
 
-    # The layout for FFMpeg; this is essentially a bitmask of channels.
-    cdef uint64_t layout
-    cdef int nb_channels
-
-    cdef readonly tuple channels
-    """
-    A tuple of :class:`AudioChannel` objects.
-
-    :type: tuple
-    """
-
-    cdef _init(self, uint64_t layout)
-
-
-cdef class AudioChannel:
-
-    # The channel for FFmpeg.
-    cdef uint64_t channel
-
-
-cdef AudioLayout get_audio_layout(int channels, uint64_t c_layout)
+cdef AudioLayout get_audio_layout(lib.AVChannelLayout c_layout)

av/audio/layout.pyi

@@ -1,12 +1,4 @@
-channel_descriptions: dict[str, str]
-
 class AudioLayout:
     name: str
-    layout: int
     nb_channels: int
-    channels: tuple[AudioChannel, ...]
-    def __init__(self, layout: int | str | AudioLayout): ...
-
-class AudioChannel:
-    name: str
-    description: str
+    def __init__(self, layout: str | AudioLayout): ...