hydrusnetwork/hydrus v543

Version 543

on GitHub

misc

• a new string converter rule now allows for extremely easy date parsing, thanks to the dateparser library. all old 'datestring to timestamp' rules remain as they are, but are now called '(advanced)'. a new option, 'datestring to timestamp (easy)', which has exactly zero variables to fiddle with, just eats up pretty much any date string you can think of, including timezone conversions, and even stuff like '2 hours ago'. you need the dateparser library for this to work, so if you run from source, you might like to rebuild your venv this week. your dateparser import status is in help->about
• thanks to the user who added it recently, PSD rendering is now much faster and uses less memory. if you do a lot of PSD work, let me know how this goes. if PSDs now load pretty much like large pngs, I think we'll set them, by default, to show as normal in the preview viewer
• thanks to a user, we now have description note parsing for the default e621 downloader
• the program now supports bitmap files as-is. until now, I automatically converted them to png on import, but this was a mistake--despite this file format being a waste 99.7% of the time, hydrus's philosophy is not to alter files on import, and this long-time exception resulted in several awkward bumps in the code that I'm happy to be rid of now
• fixed a couple desync bugs in the migrate database dialog where you could change a location's weight (particularly between 0 and 1) and not get the correct flip of the 'files need to be moved'/'files are all good' state until you re-opened the dialog

PDFs

• I screwed something up with the PDF thumbnail generation at the last minute last week, fixing it on non-PySide6, but introducing some logspam and--for at least one user--adding instability. the logspam is now gone and I believe the instability is fixed. now it is basically the same as the SVG thumbnail code, which hasn't given us any trouble. if we still see some crashes, I'm going to have to overhaul these two thumbnail generation methods
• when PDFs fail to generate thumbs, a little text about the error is now printed to the log
• help->about now has lines for QtCharts and QtPdf, and if there is a PDF problem, it puts the import trace in a popup

mr bones

• mr bones can now take any file search. if you want to see the average filesize of your pngs, or the archive/inbox ratio of creator x's webms, just set that search on the new panel and the numbers will update for that subset
• this turned, characteristically, into a bottomless rabbit hole, and I culled the more complicated features lest the ride consume me. searching a multiple file domain means deleted numbers cannot be calculated, nor can the 'earliest import' time, and searching deleted domains will generally give you some gonk numbers (and likely reveal some interesting legacy bugs, like inbox count amongst deleted files)
• the old search was highly optimised, but this has few guard rails. if you give this thing a super difficult query, it'll take a long time. there is now a cancel button that should interrupt all but the weirdest operations fairly promptly, however, just in case it is really lagging. note that hitting 'searching immediately' will pause updates as normal, if you need to set up something complex
• assuming deleted numbers are available, the stats now include total views/viewtime for deleted files too
• potential dupe counts are basically a search of 'at least one of the files matches the file list, can be pixel dupes, max distance 8'

more boring work, file storage and misc

• wrote a new object to handle the base storage location for file/thumbnail subfolders. it can do over/underweight calculations and handles the pending max_num_bytes setting for database migration locations
• all the new subfolder objects now track their base location using this new object, and all related load/save/display/edit code is now throwing this thing around instead of raw paths
• the underlying migration determination code is now ready to redistribute according to a max_num_bytes option. I've just got to update the UI, and, fingers crossed, I'll be able to add it next week
• added a bunch of unit tests for the new base storage location object. it separately reports whether it needs to shrink, wants to shrink, is able to expand, or is eager to expand
• improved how updated objects are substituted into all multi-column lists, it fixes a couple of odd storage/display sync bugs here and there
• a core image data loading/conversion tool inside the program is now a bit simpler and faster, and I think it also saves memory. it should speed up various sorts of unusual file loading