[db,ffmpeg] calculate and persist audio hash on local files

[whatdoineed2do]

Contributing to fixing #85

• DB - new audio_hash column
• scanner - optionally calc audio hash for files (based on all packets in file)
• json api, conf, smartpl - supporting work

[ejurgensen]

Thanks for the suggestion @whatdoineed2do. It's not great to solve usability issues with config options, but sometimes its all there is to do. Not sure where this is on that scale. In your experience, by what factor does this slow the scanning?

[whatdoineed2do]

just ran ths off my laptop with an NVEM drive.

1198 real audio files (mostly 20-40mb flacs) locally, destroyed the db on each run, filescan_disable = true for startup, and manually trigger the scan via curl -XPUT http://localhost:3689/api/rescan

• no audio generation: 3 secs (0.001 per file)
• audio generation: 53 sec (0.044 per file)

I guess this is not a surprise, as the server needs to read most/all of each file in chunks (packet-by-packet).

[ejurgensen]

destroyed the db on each run

How so?

[whatdoineed2do]

Destroyed db - I mean deleted the db file before each test

[ejurgensen]

I think the existing avformat_find_stream_info() reads some kB of each file, have you checked if you can get a hash from that?

[whatdoineed2do]

Not seen anything on the existing metadata that the libav* maintains.

[ejurgensen]

Sorry, but I'm not in favour of this option, so I will close the PR. The reasons are because of the performance impact and the "fix-with-a-config-option" approach.