Version 557

misc

• optimised large tag filter edit UI. you can now paste 5,000 items into an empty tag filter blacklist in less than a second, and if you have a big tag filter, removing or adding one thing is now instant (previously, this stuff would lag 4 seconds or more, sometimes multiple minutes!!)
• the ugoira 'num frames' counting method now discludes files ending in .js/.json, to catch future bundling of frame timings
• the cbz scanning tech should now recognise cbzs with four or fewer pages
• a legacy 'is this image all good?' check that happens on PIL-loading is now gone. this improves rendering for a variety of truncated files and clarifies some error messages (previously, this thing was just failing silently)
• fixed the delete file pre-flight logic so users on the non-advanced delete dialog can now delete repository updates. previously, they saw the menu entry, but hitting it was a no-op

better hash predicate parsing

• system:hash labels are a little different now. they'll say system:hash (md5) is abcd..., with the algorithm after the "hash". hash is omitted for sha256 (the hydrus default). this eases parsing
• system:similar to data labels are a little different. they'll say 'distance' instead of 'max hamming', and the number and type of hashes they hold, and if they hold only pixel hashes, the distance is not stated
• system:hash predicate parsing is now more flexible. you can put the hash type pretty much anywhere now.
• system:similar to and system:similar to data predicate parsing is now more flexible. more combinations are allowed, and you can not include distance and it'll be fine
• these three hash predicates now copy to clipboard with all their hashes explicitly enumerated, making strings that are fully parsable! this is a big step forward in a completely sealed import-export predicate parsing loop; now I have the tech set up to export a different phrase to clipboard than what you see in the label, I just need the examples of where it goes wrong. if there is a system predicate that copies to clipboard in a way that won't parse back, let me know and I'll see if I can fix it.
• added more unit tests for this parsing

documentation and cleanup

• wrote a guide on how to install 'Git for Windows' for the 'running from source' help. although most of the settings in its marathon 12-page install wizard can be left as default, the technical questions can be intimidating, so I've written them all out for a nice simple install. also brushed up some of the surrounding help here
• added a warning to the regular 'installing and updating' help regarding the danger of test-running extract releases before updating (you can overwrite your database by accident)
• thanks to a user, the filetypes help document is updated with Ugoira and CBZ info
• all the 'HydrusFiletypeHandling' files are refactored to a new 'files' module. there's a bunch of them these days!
• the hydrus.core.images module is moved beneath this 'files' module too
• the file log list panel right-click menu now says 'open URLs'/'open files' locations' depending on whether you are looking at a URL import log or local HDD import log

client api

• the file_metadata call now returns filetype_forced and, if so, also original_mime to talk about the new forced filetype system
• the client api help and unit tests are updated to test this is working ok
• fixed a typo that was causing too much work in the updated file info manager call (and was often returning 'null' results for half-cached file_metadata requests with only_return_basic_information=true)
• thanks to a user, the /add_urls/get_url_info Client API call now has a cache timeout of ten minutes, and the /add_urls/get_url_files call now has a timeout of 30 seconds if all the files are 'already in db'. this should automatically reduce some overhead for several programs that talk to the Client API a lot about URLs
• the client api version is now 58