mirror of https://github.com/H-uru/korman.git
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
104 lines
3.4 KiB
104 lines
3.4 KiB
/* This file is part of Korman. |
|
* |
|
* Korman is free software: you can redistribute it and/or modify |
|
* it under the terms of the GNU General Public License as published by |
|
* the Free Software Foundation, either version 3 of the License, or |
|
* (at your option) any later version. |
|
* |
|
* Korman is distributed in the hope that it will be useful, |
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
* GNU General Public License for more details. |
|
* |
|
* You should have received a copy of the GNU General Public License |
|
* along with Korman. If not, see <http://www.gnu.org/licenses/>. |
|
*/ |
|
|
|
#include "sound.h" |
|
#include "PyHSPlasma_private.h" |
|
|
|
#include <PRP/Audio/plSoundBuffer.h> |
|
#include <Stream/hsStream.h> |
|
#include <vorbis/vorbisfile.h> |
|
|
|
static const int BITS_PER_SAMPLE = 16; |
|
|
|
static size_t _read_stream(void* ptr, size_t size, size_t nmemb, void* datasource) { |
|
hsStream* s = static_cast<hsStream*>(datasource); |
|
// hsStream is a bit overzealous protecting us against overreads, so we need to |
|
// make sure to not tick it off. |
|
size_t request = nmemb * size; |
|
size_t remaining = s->size() - s->pos(); |
|
size_t min = std::min(request, remaining); |
|
if (min == 0) { |
|
return 0; |
|
} else { |
|
size_t read = s->read(min, ptr); |
|
return read / size; |
|
} |
|
} |
|
|
|
static int _seek_stream(void* datasource, ogg_int64_t offset, int whence) { |
|
hsStream* s = static_cast<hsStream*>(datasource); |
|
switch (whence) { |
|
case SEEK_CUR: |
|
s->skip(offset); |
|
break; |
|
case SEEK_SET: |
|
s->seek(offset); |
|
break; |
|
case SEEK_END: |
|
s->seek(s->size() - offset); |
|
break; |
|
} |
|
return 0; |
|
} |
|
|
|
static long _tell_stream(void* datasource) { |
|
hsStream* s = static_cast<hsStream*>(datasource); |
|
return s->pos(); |
|
} |
|
|
|
static ov_callbacks s_callbacks = { |
|
(size_t(*)(void *, size_t, size_t, void *))_read_stream, |
|
(int (*)(void *, ogg_int64_t, int))_seek_stream, |
|
(int (*)(void *))NULL, |
|
(long (*)(void *))_tell_stream, |
|
}; |
|
|
|
PyObject* inspect_vorbisfile(PyObject*, PyObject* args) { |
|
pyStream* stream; |
|
pyWAVHeader* header; |
|
if (!PyArg_ParseTuple(args, "OO", &stream, &header)) { |
|
PyErr_SetString(PyExc_TypeError, "inspect_vorbisfile expects an hsStream, plWAVHeader"); |
|
return NULL; |
|
} |
|
|
|
// The OGG file may actually be in Blender's memory, so we will use hsStream. Therefore, |
|
// we must tell vorbisfile how to do this. |
|
OggVorbis_File vorbis; |
|
int result = ov_open_callbacks(stream->fThis, &vorbis, NULL, 0, s_callbacks); |
|
if (result < 0) { |
|
PyErr_Format(PyExc_RuntimeError, "vorbisfile ov_open_callbacks: %d", result); |
|
return NULL; |
|
} |
|
|
|
header->fThis->setFormatTag(plWAVHeader::kPCMFormatTag); |
|
header->fThis->setBitsPerSample(BITS_PER_SAMPLE); |
|
vorbis_info* info = ov_info(&vorbis, -1); |
|
header->fThis->setNumChannels(info->channels); |
|
header->fThis->setNumSamplesPerSec(info->rate); |
|
unsigned short align = (BITS_PER_SAMPLE * info->channels) >> 3; |
|
header->fThis->setBlockAlign(align); |
|
header->fThis->setAvgBytesPerSec(info->rate * align); |
|
|
|
// This behavior was copied and pasted from CWE |
|
ogg_int64_t size = (ov_pcm_total(&vorbis, -1) - 1) * align; |
|
|
|
// Cleanup |
|
ov_clear(&vorbis); |
|
|
|
// We got the plWAVHeader from Python because we don't link against PyHSPlasma, only libHSPlasma |
|
// Therefore, we only need to return the size. |
|
return PyLong_FromSsize_t(size); |
|
}
|
|
|