Hello!

I recently replaced dr_wav with your wonderful AudioFile library. I've been getting reports from some of my users that .wav files that previously loaded and play fine no longer work. I've found that the reason is that the .wav files that they were using were greater than than 24-bit.

The bitrates supported by dr_wav are:

Unsigned 8-bit PCM Signed 12-bit PCM Signed 16-bit PCM Signed 24-bit PCM Signed 32-bit PCM IEEE 32-bit floating point IEEE 64-bit floating point

Would it be possible to support these .wav file formats?

Thanks! Bret

When using this library, I get a few warnings related to -Wsign-compare. Are there any plans on fixing it? It shouldn't be too hard.

In file included from src/main.cpp:3:
src/../include/AudioFile.h: In instantiation of ‘bool AudioFile<T>::decodeWaveFile(std::vector<unsigned char>&) [with T = double]’:
src/../include/AudioFile.h:403:16:   required from ‘bool AudioFile<T>::load(std::string) [with T = double; std::string = std::__cxx11::basic_string<char>]’
src/main.cpp:20:40:   required from here
src/../include/AudioFile.h:468:28: warning: comparison of integer expressions of different signedness: ‘int32_t’ {aka ‘int’} and ‘uint32_t’ {aka ‘unsigned int’
} [-Wsign-compare]
  468 |     if ((numBytesPerSecond != (numChannels * sampleRate * bitDepth) / 8) || (numBytesPerBlock != (numChannels * numBytesPerSample)))
      |         ~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/../include/AudioFile.h: In instantiation of ‘bool AudioFile<T>::decodeAiffFile(std::vector<unsigned char>&) [with T = double]’:
src/../include/AudioFile.h:407:16:   required from ‘bool AudioFile<T>::load(std::string) [with T = double; std::string = std::__cxx11::basic_string<char>]’
src/main.cpp:20:40:   required from here
src/../include/AudioFile.h:599:90: warning: comparison of integer expressions of different signedness: ‘int’ and ‘std::vector<unsigned char>::size_type’ {aka
long unsigned int’} [-Wsign-compare]
  599 |     if ((soundDataChunkSize - 8) != totalNumAudioSampleBytes || totalNumAudioSampleBytes > (fileData.size() - samplesStartIndex))
      |                                                                 ~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/../include/AudioFile.h: In instantiation of ‘int AudioFile<T>::getIndexOfString(std::vector<unsigned char>&, std::string) [with T = double; std::string = s
td::__cxx11::basic_string<char>]’:
src/../include/AudioFile.h:428:28:   required from ‘bool AudioFile<T>::decodeWaveFile(std::vector<unsigned char>&) [with T = double]’
src/../include/AudioFile.h:403:16:   required from ‘bool AudioFile<T>::load(std::string) [with T = double; std::string = std::__cxx11::basic_string<char>]’
src/main.cpp:20:40:   required from here
src/../include/AudioFile.h:1007:23: warning: comparison of integer expressions of different signedness: ‘int’ and ‘std::vector<unsigned char>::size_type’ {aka
‘long unsigned int’} [-Wsign-compare]
 1007 |     for (int i = 0; i < source.size() - stringLength;i++)
      |                     ~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
src/../include/AudioFile.h: In instantiation of ‘void AudioFile<T>::clearAudioBuffer() [with T = double]’:
src/../include/AudioFile.h:490:5:   required from ‘bool AudioFile<T>::decodeWaveFile(std::vector<unsigned char>&) [with T = double]’
src/../include/AudioFile.h:403:16:   required from ‘bool AudioFile<T>::load(std::string) [with T = double; std::string = std::__cxx11::basic_string<char>]’
src/main.cpp:20:40:   required from here
src/../include/AudioFile.h:950:23: warning: comparison of integer expressions of different signedness: ‘int’ and ‘std::vector<std::vector<double, std::allocato
r<double> >, std::allocator<std::vector<double, std::allocator<double> > > >::size_type’ {aka ‘long unsigned int’} [-Wsign-compare]
  950 |     for (int i = 0; i < samples.size();i++)

Great job on the library by the way!! Super useful and lightweight :)

There is a link problem in AudioFile.cpp at aiffSampleRateTable, please change the class in unordered_map template definition from int to long or uint32_t, etc..., to fix it.

std::unordered_map <long, std::vector<uint8_t> > aiffSampleRateTable

Nice code btw.

Added support for multi-channel files. Did not have to do a lot - mainly adding tests to make sure it works. Currently testing writing 8-channel files for all formats and reading with 8-channel wav 24bit, 48k Updated python script to generate headers as well.

Hi Mr. Adam Stark,

May I know how to get the size of an AudioFile because I want to assign to OpenAL Soft generated buffer to store the AudioBuffer. I am referring to the tutorial here (LINE 157, 163, 169). Sorry for the inconvenience.

Thank you Mr. Adam Stark

It would be very helpful for directly loading decrypted .wav files. I noted it is almost done with AudioFile::decodeWaveFile, but it is a private method, so maybe an overloaded option of AudioFile::load would be great, accepting (void* data, size_t size) or std::vector<uint8_t> Thanks for this awesome library

Replaced the std::cout statements with exceptions, because a library shouldn't display any error messages.

Fixed bug in load function, where end iterator was not initialized. Introduced rawread function to read the bytes instead.

I added support for reading and writing of wave files with 32bit resolution with corresponding tests.

The unit tests currently use "hard-coded" reference data (got it through Matlab's audioread) and compare the first 64 samples of audio.

Also, I'm creating the output directory for the unit test in case it doesn't exist yet. Added it to .gitignore as well.

I haven't had time to look into the headers you use as reference, but guess makeHeaders.py is what I'm looking for.

Due to the fact that you're implementing a Template, to avoid issues with different compilers, you should consider having only one .hpp file with the full template definition there.

Make AudioFile.h compatible with -Wall -Wextra -pedantic -Werror compilation options for GCC (And /W4 and /WX for MSVC). This will make it easier project compatible with many projects. You can use GitHub Actions to easly run tests on Linux, Windows and Mac.

Hi Adam,

In a short test, filling the buffer with some sine values, I noticed the resulting wave had a negative peak value, where it should have been the max positive value. It seems to be that a 1.0 float value in the buffer is saved as 32768 (for the 16-bit format). This is the piece of code for saving the file:

            else if (bitDepth == 16)
            {
                int16_t sampleAsInt = (int16_t) (samples[channel][i] * (T)32768.);
                addInt16ToFileData (fileData, sampleAsInt);
            }

The easiest solution would be to replace the 32768 value by a 32767, hereby 'losing' the max nagative -32768 value. Or using an intermediate 32-bit value, which is clipped between 32767 and -32768 afterwards.

Regards and thanks for your code,

Ruud

Hi, i met a error when using short type, "This version of AudioFile only supports floating point sample formats"

Will AudioFile support short type in the future? Thanks.

Hi, Adam: When I use msan to run AudioFile's exmaple program, it outputs the following error message:

Uninitialized bytes in MemcmpInterceptorCommon at offset 48 inside [0x7fffcb727450, 256)
==89831==WARNING: MemorySanitizer: use-of-uninitialized-value
    #0 0x5627f25d0a0e in __interceptor_memcmp (/workspace/test/msan/AudioFile/examples/Examples+0x5da0e) (BuildId: d090aa71c8d408cc1981b1bbc768ff7bf3daf5a3)
    #1 0x7f3f4348d071 in std::ctype<char>::_M_widen_init() const (/lib/x86_64-linux-gnu/libstdc++.so.6+0xd7071) (BuildId: 725ef5da52ee6d881f9024d8238a989903932637)
    #2 0x7f3f434f280f in std::basic_ostream<char, std::char_traits<char> >& std::endl<char, std::char_traits<char> >(std::basic_ostream<char, std::char_traits<char> >&) (/lib/x86_64-linux-gnu/libstdc++.so.6+0x13c80f) (BuildId: 725ef5da52ee6d881f9024d8238a989903932637)
    #3 0x5627f2620132 in examples::writeSineWaveToAudioFile() (/workspace/test/msan/AudioFile/AudioFile/Examples+0xad132) (BuildId: d090aa71c8d408cc1981b1bbc768ff7bf3daf5a3)
    #4 0x5627f2620010 in main (/workspace/test/msan/AudioFile/examples/Examples+0xad010) (BuildId: d090aa71c8d408cc1981b1bbc768ff7bf3daf5a3)
    #5 0x7f3f430b0d8f  (/lib/x86_64-linux-gnu/libc.so.6+0x29d8f) (BuildId: 89c3cb85f9e55046776471fed05ec441581d1969)
    #6 0x7f3f430b0e3f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e3f) (BuildId: 89c3cb85f9e55046776471fed05ec441581d1969)
    #7 0x5627f2597ee4 in _start (/workspace/test/msan/AudioFile/examples/Examples+0x24ee4) (BuildId: d090aa71c8d408cc1981b1bbc768ff7bf3daf5a3)

SUMMARY: MemorySanitizer: use-of-uninitialized-value (/workspace/test/msan/AudioFile/examples/Examples+0x5da0e) (BuildId: d090aa71c8d408cc1981b1bbc768ff7bf3daf5a3) in __interceptor_memcmp
Exiting

Although this does not necessarily bring any high-risk effects, I still think it is an abnormal result, and hope to get your attention

Verification steps：

git clone https://github.com/adamstark/AudioFile.git
cd AudioFile/
git checkout develop
mkdir check_build && cd check_build
cmake ../ -DCMAKE_C_COMPILER=clang  -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_FLAGS="-fsanitize=memory" -DCMAKE_CXX_FLAGS="-fsanitize=memory" 
make -j 
cd examples
./Examples

enviroment: Ubuntu 22.04 LTS Linux lab-pro 4.15.0-147-generic gcc (Ubuntu 11.2.0-19ubuntu1) 11.2.0 Ubuntu clang version 14.0.6

Thanks & Best regards !

Hi Adam,

This is a nice header-only library.

Quick Q: If the header is included in two different cpp files, won't you get two independent copies of the aiff sample rate table, because the static will be scoped to within the cpp file that included it?

// Pre-defined 10-byte representations of common sample rates
static std::unordered_map <uint32_t, std::vector<uint8_t>> aiffSampleRateTable = {
    {8000, {64, 11, 250, 0, 0, 0, 0, 0, 0, 0}},

I am working on a project in which data keeps streaming in and I need to save the frames to wav as they come in. I cannot wait until I have all the data to save and the sequence load_wav-add_samplesa-save_wav seems wasteful and I may lose frames. Is there a better way to "flush" the frames to disk without having to close the file all the time?