Musepack: Support new Musepack SV8 API about mutagen HOT 12 CLOSED

This issue was closed by revision r122 .

Joe? Will you merge this?

Changes:

 - get bitrate using file size; the calculation was wrong for sv4-7: bps/1000+0.5 instead of just bps
 - move sv8_int, peak/gain in functions, IOError in sv8_int after limit is reached
 - move sv8 parsing in a seperate method
 - I've checked the replaygain calculation with GStreamer and it's correct.
 - correctly ignore empty gain/peak fields (this needs to be fixed in QL).
 - include version and bitrate in pprint

Sorry it's taken so long. I guess it's still technically October...

Comments on the patch:

I really don't like the "while 1". Can it be replaced with a real loop invariant? I'd rather just chunk frames as we go rather than repeat the validation logic for each frame type we care about, stopping as soon as we find an AP.

Do we really need to parse every frame in the file just to get the approximate bitrate? It's going to be very slow (and still not accurate at all). Just let the file length / time case take care of it.

I don't like the get_sv8_int staticmethod. This should probably be a bare function or part of the cdata module. I also don't see a reason why it stops after 9 bytes except as a sanity check - in which case it should throw an exception, not return 0, after failing to stop after reading 8 bytes.

If the exceptions are going to claim it's a Musepack file with an invalid frame, it should explain why it's an invalid frame.

I'd like someone from the Musepack team to explain why their new Replay Gain format is so incredibly bad. Until then, Mutagen shouldn't try to parse it.

For the record, this is the #1 thing in my Mutagen pile right now; unfortunately Mutagen is about the #-1 thing in my overall pile. I'm hoping to get to it by October.

> Hi, it would be nice to merge this patch into the trunk :) cheers
another cheers on that :D

Hi, it would be nice to merge this patch into the trunk :) cheers

Thanks Philipp.

Updated patch and test sample attached.

I just played with your patch and found a file that will cause a
MusepackHeaderError("Invalid Frame") in line 84. The attached file is a SV8 file with
zero audio length. I have also attached the WAV source from which I generated the
file. The SV7 version of the same file can be read without problems.

Other files seemed to work so far.

Patch+Testfile attached.

I made the peak calculation the same as with SV7, but I think it is wrong in both
cases, or QL needs to convert them somehow.. I guess you know more about this.

Wow, they made the header format really crap.