Implemented WAV source, sink and backend

[Pekureda]

Implemented issue: 576

• Implemented source, sink and backend for WAV file read and write operation handling,
• Added dr_wav single header library,
• Read operations of WAV files are performed using dr_wav,
• Implemented WavHeader to store and handle headers of written WAV files,
• Changed lib download links from ftp to https based.

[github-actions]

🤖 Upon creation, pull request description does not have a link to an issue. If there is a related issue, please add it to the description using any of the supported formats.

[gavv]

Thanks for PR! Will review in upcoming days!

[gavv]

Thanks for the patch, here are a few comments.

[gavv]

I guess the code can be simpler and shorter if we would make WavHeader a struct and write it directly to file?

Currently we're defining all fields in .h and then repeating all fields in this function, which doubles the code size. Instead, we could just fwrite() the whole struct.

We'd need to use ROC_ATTR_PACKED_BEGIN/ROC_ATTR_PACKED_END for the struct to ensure that compiler does not insert any padding.

If WavHeader becomes a struct, I guess we can define in directly in wav_sink.cpp (in anon namespace), since it will be small and won't be used anywhere else.

[Pekureda]

I've decided to leave this class but move header data to a POD struct, which memory is this time managed by the WavHeader instead of the user, so it should be cheaper to use it. In case of inclusion of some other, extra parameters to the header, it should be a bit easier to extend this class.

Technically the number of samples written so far can be stored in WavSink, so I'm not really set on any of those solutions, because they're similarly good to me.

[gavv]

I think current version is nice!

[Pekureda]

Thanks for detailed review and finding all those issues! I've updated the commit and fixed mentioned problems.

[gavv]

Thanks,

It seems you forgot to commit your changes or pushed a wrong commit?

Seems that the new cimmit is identical to the old one.

https://github.com/roc-streaming/roc-toolkit/compare/155d58f67566a3a8c2833059268dc80572743a1b..d7378fa820baea482d65ae5afd9242532cca1e3a

[gavv]

Hi, I've pushed commit 03d29eb to develop branch that adds new fields sample_format and pcm_format to SampleSpec.

It mostly does not affect this PR, but SampleSpec ctor now has new parameter pcm_format, which you can just set to audio::Sample_RawFormat (which means that source/sink produces/consumes samples in "raw" format - 32-bit PCM floats).

So please rebase on fresh develop.

[Pekureda]

It's rebased and adjusted to newest changes on develop branch. I must've not force pushed the amended commit and not noticed that.

[gavv]

Thanks, looks and works great!

I've pushed a small follow-up commit: fc0fab2

Changes:

• Array::grow() call is now not needed
• In WavSink, if channels or rate is not specified, use default values (otherwise roc-send panics)
• In WavSource, ignore sample_spec from config (we don't and can't use it because we can't do any conversions)
• Use Sample_RawFormat instead of PcmFormat_Float32_Le when reporting sample spec. Sample_RawFormat corresponds to format of sample_t type. It's equal to PcmFormat_Float32_Le only on little-endian machines, but it's PcmFormat_Float32_Be on big-endian ones.

[gavv]

We also forgot one small thing, I've created a new issue for it: #684