related tags
EDIT: In final testing, this stuff was working slow IRL for me, so I reduced its search space for today. You are going to see a lot of 'search domain is too large' on the PTR. I'll keep working!
Thanks to a user who sent in a great proof of concept, the 'related tags' column (which shows in manage tags and you can turn on under options->tag suggestions) has some new buttons to test. I called them 'new 1' and 'new 2'. The first works fast but may have limited results on repositories, the second--which only shows on repositories--can only search small-count tags but gives deep results.
Secondly, if you select some tags in the main tags list, the 'related tags' search will now only search those tags! If you can't remember a character name, then select a 'series' tag and click for a related tags search, and you'll get a bunch of them!
Thirdly, these new search buttons now work on manage tags launched from multiple thumbnails!
Please give this a go and let me know how you get on with IRL data. The quality of the new search results is better than my old routines, but I need to do a bunch more tuning and optimise their search. Are there too many recommendations (is the tolerance too high?)? Is the sorting good (is the stuff at the top relevant or often just noise?)? Which source tags (e.g. 'page:x') give the worst suggestions?
I'm sufficiently happy with the new performance, particularly with the tag selection, that once this is working nice, I think I'll make it default-on for everyone. Lots of people missed this system, and I always thought it was a bit jank.
other highlights
I wrote some sidecar import/export help here! https://hydrusnetwork.github.io/hydrus/advanced_sidecars.html
I optimised the how the thumbnail cache and downloader import queues do their normal work, particularly when they get tens of thousands of items. Big clients, which may have very busy versions of these objects, should be a little less laggy. Let me know if you notice any difference!
Downloaders that get from multiple sites in one query (aka NGUGs) now do their searches one site at a time, not interleaved (e.g. they now search sites A and B as AAABBBB, rather than ABABABB)!
next week
I've been semi-ill this past two weeks with some long-running something, and I pushed myself a bit hard getting all this ready in time. Next week will be some simple code cleanup, nothing too clever!