A couple of days ago we made a new release available at api.onemusicapi.com. This
included support for Discogs' new approach to image serving, improved performance and also
better accuracy and matching coverage for short CDs in the /disc endpoint.
None of these changes have required a new API version, so you can continue using the API and you will benefit from these changes automatically.
One of the most significant changes behind the scenes has been the adoption of new ways to work
with Discogs images. Discogs have deprecated
their existing /images endpoint on discogs.com and have now disabled it
completely. Images are now served from a dedicated image cluster.
On the upside, in addition, the maximum number of images you can download has been raised significantly, from 1000 per application per day to 300 per minute per IP address.
On the downside, it's yet another change to Discogs' API, which means current integrations have to be re-written. Of course, if you are using OneMusicAPI, no changes are required!
We've taken some big steps to improve performance in this version.
For tag based lookup we've begun using the scores from our search provider to drop some results which are sufficiently irrelevant compared to other results.
That sounds a little clumsy; but that's because the 'scores' we receive are not absolute, they are relative to each other. We have worked to research what relative scores we can safely drop while maintaining good coverage. As a result, we can improve performance because there's less downstream work to do.
In addition we have started to cache, short term, results for lookups from our database.
A number of improvements to
/disc queries...
Matching for shorter discs, e.g. CD singles, has been improved when using Acoustid fingerprint matching. We've tweaked the heuristics for using scoring to make sure we readily match discs where the titles of tracks are "close enough".
We've also added position data for each media item inside associated_releases.
Some smaller items:
To give OneMusicAPI a try all you need to do is sign up and you'll get 1000 free queries. Further queries can be purchased on a rolling monthly basis or as a credit top-up.
Thanks to chintermeyer who made the the image above available for sharing.
