Reading the TOC of a disc can be slow

[pimzand]

Reading the TOC of a disc can be very slow, especially when extracting MCN and ISRC, which is done by default.

There are a number of strategies that could improve the speed, preferably without disabling MCN and ISRC extraction.

• add debug statements to measure the time that (lib)discid and cdparanoia are taking to list the contents of a CD.
• measure discid vs libdiscid speed
• never enable MCN and ISRC extraction when we are only looking for a matching existing jack.toc
• use the read_sparse() instead of the read() method of libdiscid, and check whether discid has such a method

[pimzand]

Debugging discid.py shows:

• reading the TOC of "Spike" by Elvis Costello, which has an MCN but no ISRC, takes 40 seconds
• there is no speed difference between discid and libdiscid, they are both wrappers for the same c-library
• for a CD that is not in your collection, jack will read the TOC when searching for an existing jack.toc, when creating jack.toc, and for each track to be ripped
• disabling ISRC extraction makes the read time less than a second

discid and libdiscid do not expose read_sparse(). It appears they call read_sparse() automatically, and fall back to read() if read_sparse() is not available.

[pimzand]

Oviously, we should reduce the number of times jack reads the toc. The first read, used to search an existing jack.toc, should never use the MCN or ISRC features. The second read, used to create the jack.toc, should use the MCN and ISRC features, if the user hasn't disabled it. The third, per-track read should be completely avoided.

[pimzand]

For now, the quick and dirty approach was to cache the results of jack.functions.gettoc, so we only call libdiscid.read() once. This still leaves some room for improvement, as searching for existing jack.toc files still reads the disc with MCN and ISRC features enabled.

But already, there's a big improvement while keeping MCN and ISRC features enabled for future use.

[pimzand]

works quite well, closing