changelog¶

Note

This is the new changelog, only the most recent builds. For all versions, see the old changelog.

Version 673 ¶

misc¶

the file history chart now has a custom y axis range. also, the chart now remembers if you have set either axis custom and new searches will auto-refit or maintain current dimensions as appropriate. hide/showing the lines will only recalculate the non-user-customised Y axis; let's see how that goes
added a sanity check to the new fast 'give me the average character width' calculation, which is used for some scaling-agnostic UI sizing. one user (on a monospaced font, no less) had extremely wide average character width; I guess the font has funny kerning or extended characters or something. if the average character width is more than twice the reported height (which appears to be more reliable), I now fall back to a slower but more accurate calculation
you can now edit the Access Key of a Client API permissions entry (a user mentioned they were migrating to a new client and updating every existing script to use new random keys was a pain). since you don't want to do this casually, it works through a button that gives a little spiel and tests the new key for validity and such, and the final ok will bail out if you paste something already in the system
updated some system predicate parsing to support <= and >= operators, along with some variants like 'less than or equal to'. the types now supporting this are: width, height, duration, number of frames, number of words (issue #2019)

new help docs for the recovery.txts¶

added a 'Recovery' headline section to the help and migrated the .txt recovery docs to basic markdown
the basedir 'help my client will not boot' is migrated to here
all the .txts in the db dir like 'help my db is broke.txt' are migrated to here
as planned, the static/db_files dir is removed. you no longer get a bunch of .txts in any new db folder. feel free to delete any old ones you have, but it isn't a big deal

local file parsing optimisation¶

when you drop a folder on the program, the main scan of that folder is a good bit faster than before and will scale a bit better
when you drop a folder on the program, symlink loops are now recognised and broken out of
when parsing import files from a folder, the main parse object now uses several fewer drive hits
checking for 'file is in use' requires one less drive hit

faster folder checking on startup¶

when hydrus boots, it checks for the presence of all file storage folders. on a normal client, this is 512 directory presence checks; on an advanced granularity 3 system, this is 8192. this time adds up on boot, particularly on a cold HDD. I have improved the regular test here to do just one hard drive hit per folder instead of two. also, especially for the bootup phase, these locations are now scanned for en masse with a carefully efficient/failsafe top-level scan on the main storage locations, massively reducing the number of hard drive hits required here

optimised caching tech¶

a user identified that a hacky id-to-value lookup cache used in tag and hash database modules was not working great. under certain types of strain, it would churn, leading to memory bloat and fragmentation
I have tried several solutions and figured out a fairly decent replacement (LRU cache, nothing crazy) that will not churn so much and has less overhead. there's some additional long-term work that needs to be done to solve the bloat problem fully (full weakref tracking of tags/hashes), but I'm overall happy. tag and hash fetching when you load media or do various other heavy database jobs is now a little more optimised in several ways, and in most cases causes less memory duplication and fragmentation
while I was poking around here, I also overhauled the general LRU cache used by a bunch of UI-level guys. thumbnail refetch and image zooming back and forth may be a shave faster

source environment cleanup¶

as planned a few months ago, v673 cleans up the 'running from source' setup significantly. you shouldn't have to do anything unless you run from source and use a custom script to automatically recreate your venv. I delete some old redundant scripts today, so if you happened to set an executable permission on something a long time ago, git may moan at you about being unable to pull because of your pending changes. deleting the files and then pulling again should work
the pyproject.toml file no longer has any groups. there's one setup, nice and simple. the venue to test alternate library versions is now setup_venv.py exclusively
the old basedir requirements.txt is now removed
the manual 'running from source' help is updated. you now do just pip install . for a manual, pyproject.toml based pip install, with no groups needed
the .bat/.command/.sh versions of setup_help and setup_venv and git_pull are removed--use the multiplat .py files from now on
the open_venv.bat/.ps1 scripts and auto_update_installer.bat, which were just fun experiments, are deleted. if you need some rinky-dink scripts to pull off a very custom thing like this, I recommend talking to an AI to get exactly what you need for your setup
to improve hydrus package security, all dependency versions in the pyproject.toml and setup_venv.py and the build requirement.txts are now pinned/capped to recent latest versions. anything that was >= is now <= for the version as of the 672 build. all library version updates will now be considered manually by human eyes in future builds
relatedly, the windows ffmpeg version is no longer latest but pinned at 8.1.1
deduped the basedir license files and renamed to LICENSE
wrote a very basic CONTRIBUTING.md to mention that public pulls are closed right now
for KISS, I'll switch the builds from their requirements.txts over to the pyproject.toml in the next future build test

boring cleanup¶

moved some file parsing code out of ClientGUILocalFileimports to ClientImportFileParse
jiggled some 'make this panel x characters wide' numbers after last week's character-width update. this generally meant clearing out old +2 padding hacks and shaving some 64 to 60, that sort of thing, and I fixed a couple of things that were a little out of whack or sizing the wrong widget

Version 672 ¶

misc¶

fixed a stupid error where the new-ish media playback->ffmpeg call timeout setting was not hooked up correctly on options dialog ok and was not saving! sorry for the trouble, I don't know how this slipped through testing
fixed an issue where edit-pasting a prefixed 'sha256:abcd...'-style hash into an existing 'system:hashes' would wipe out the existing hashes (issue #2015)
fixed the 'is this video rotated 90 degrees?' test in my ffmpeg output parsing for ffmpeg 8.1.x (which the windows builds moved to recently) (issue #1377)
improved the speed and precision of the core call in the 'hey roughly how wide is 16 characters for this widget?' size calculations used by stuff across the program. system:hash panel should fit better in different fonts and sizes now. this may make some tight multi-column lists (like the one in duplicate page, auto-resolution tab) go a bit wide, requiring you to shrink them a little manually to hide new horizontal scrollbars--forgive me
if a PNG file has chromaticity data but not gamma data, I now sub in a gamma of 0.45455 (which works well as a best-case fallback), and continue with the new ICC Profile-based chromaticity correction. thanks to the user who noticed this in the duplicate filter and had example files to test that rendered with a slightly different glow but were inexplicably marked, in an older version of the client, as pixel dupes

duplicates auto-resolution¶

the cog icon of the potential duplicate par search panel (which you see on dupe pages and a couple other places in auto-resolution UI) has a new 'start new potential duplicate pair search panels paused' entry. if you use a bunch of these and wrestling with the pause status is annoying, try it out
you can now use 'system:number of pixels' in the single-file 'test A or B using search terms' comparator
you can now use 'system:ratio' in the single-file 'test A or B using search terms' comparator
you can now use 'system:ratio' in the 'test A against B using file info' comparator, but the UI for it shows some default operator labels; < instead of taller than. the comparator summary label should work though
the system pred dropdown in that edit panel (this is the Metadata Conditional edit panel) collapses system:width, height, num_pixels, and ratio down to 'system:dimensions', like in a normal search page
added unit tests for the new comparators

some import options follow-up¶

added three simple examples to the top of the new import options help for 'setting up import previously deleted'; 'sending some tags elsewhere'; and 'forcing a tag redownload' as a stepping stone between 'ignore this whole system m8' and 'how to harmonically conjunct the polyhierarchic metalateral defaults'
also added an example of how to customise and clone URL Class defaults
added a 'help for this panel' button that links to the html help to the regular edit import options panel
the regular edit import options panel's favourite button has a new 'save current value as new favourite' entry in its menu, under edit/add
in the new import options system, the 'locations' import options now shows for post urls, watcher urls, and the subs defaults in simple mode

client api¶

added /client_info, accessible to all valid access keys, which provides a random hex boot_id, a float boot_time, and currently_idle bool, for tracking client restarts and throttling decisions
Client API version is now 92

boring cleanup¶

the 'fetch service id' button in review services has nicer error handling and now disables the button while it works
cleaned up how the 'set forced mimetype to these files' operation works behind the scenes
reworked how the sidebar taglist broadcasts tag changes to the current search; moving from an old pubsub to a newer Qt signal
fixed an issue with dissolving an OR predicate from the active predicates menu where the signal was being double-sent
reworked the NumberTest rendering tech to better handle custom number and operator rendering
the Number Tests across the program, which power a bunch of the newer system predicates where you can say 'width is approx 400 ± 15%' are no longer coerced to integers behind the scenes. this doesn't affect much, but in the duplicates auto-resolution system, where you can do A has height > 1.8x B, that multiplier can now result in a float. in the trivial case of B height 1, 1 * 1.8 is now less than 2, rather than being rounded up

Version 671 ¶

import options overhaul¶

the frontend and object storage is moved to the new import options system! there's a bunch going on here, but in brief--
the old file/tag/note import options are split into seven smaller types: prefetch, file filtering, tag filtering, locations, tags, notes, presentation
your old options have been converted to the new system and it should all work as before, with only minor logical changes. you do not have to do anything
all the defaults are now edited in file->options->import options. the system is richer and supports favourites/templates for quick-load
all the 'import options' buttons across the program use some slightly newer UI and offer quick 'load from favourites'
the 'edit subscriptions' dialog now has a column to show a summary for any custom import options you have set
the 'edit subscription' dialog now has a column to show a summary for any queries that have 'additional tags' set up
full help, now with screenshots, is here: https://hydrusnetwork.github.io/hydrus/getting_started_import_options.html

misc¶

the pretty media info lines that appear on a single media right-click flyout menu and some other places now includes a approx bitrate: 1.4MB/s line. it is pretty similar to the 'sort by approx bitrate' sort but only appears for stuff with an actual duration (the sort also puts big static images before small ones). it is just a naive size / duration and it'll do 1.4MB/s with existing display tech rather than 7,200KBps as you otherwise usually see, so I may revisit when I eventually get around to 1000/1024 byte-count presentation options and such
renamed the shortcut labels 'close page' and 'restore the most recently closed page', to 'pages: close current' and 'pages: unclose the most recently closed'
added a 'pages: rename current' shortcut to 'the main window' set
fixed the setting of referral urls to child import objects in two cases: A) when a file import produced multiple child urls with the note Found x new URLs in one post and B) when a gallery url in the search phase produced an unexpected non-gallery url because of 3XX or api url class redirection with note was redirected to a non-gallery url, which has been queued as a file import.. the referral url was not being set correctly because of a regression in v664. thank you to the enterprising user who ran this through Codex to find the solution
fixed a foolish typo that broke importing a cookies.txt (issue #2011)
when loading a cookies.txt, if the file includes some complicated utf-8 encoding that the current default locale cannot decode, I now catch the decoding failure and try to recover with a newly encoded temp file (also issue #2011)
fixed a bit of page counting logic that would turn up in some 'hey do you want to close the x pages?' labels and some 'do to current page' logic, where a page of pages was not counting child page of pages
updated the base requirements.txt to match the future build stuff we folded in recently. missed this guy by accident; he'll be deleted in v673 in lieu of the cleaner pyproject.toml that's coming up
did a quick hotfix patch to master for the curl_cffi test, which was frozen by accident at chrome/http2
added a note to the installation help, Linux specifically: if you get a crash on any keypress after booting the client, you probably need to run from source (issue #2010)
added some careful typedefs and imports to fix a deferred import that was causing certain recent database updates to fail out when trying to load the core options object

normal users do not pass this line

import options detailed changes¶

an import options manager with default settings is now created with a new db
on update from v671, your existing quiet/loud file import options, post/watcher tag/note import options, and all the url class tag/note import options are converted into a new import options manager. the old settings are deleted from your options and domain manager objects once it is all done and clear
a new button-with-a-dropdown-menu-arrow to handles the import options container of a specific import context. its arrow menu offers a detailed review of what it currently holds and supports quick 'favourites/template' loading
the 'fetch page metadata even if hash/url recognised and file appears "already in db" stuff is moved from tag import options to the new 'prefetch import options'. this is the only movement of options; everything else has been separated into the sub-panels you've seen in recent weeks
updated the new system's freeze routine to consult both the given url of the import object and any referral url, and if either is an API redirect, to look up every url class in the api redirect chain and select the first for which there is a custom import options entry. previously, referral url lookup was spotty, and it would only ever look at the final place in the api redirect chain. very sophisticated setups can now have separate per-url-class import options that nonetheless connect to the same final api url class
import folders and subscription queries now add 'additional tags' even if the file ends up 'previously deleted' (all imports generally apply new content updates to a 'previously deleted' result, so this brings these supplementary tags into normal behaviour.) I may add an option around which import statuses quality for a post-import content update commit, now everything is on the same page
if you hop skip and jump, you can now set 'additional tags' to a local file import that is entirely adjunct to the normal filename tagging system. there's probably some weird edge undesired behaviour now, because of odd new capabilities like this, so I'm interested in error reports or 'hey you should probably hide this by default here' reports
made some svgs and screenshots for the new import options help, removed the old help, integrated it all into the 'gettting started - downloading' help, and added it to the main index

import options detailed fixes¶

fixed some import situations (usually when the import object had no other tags to add) where 'additional tags' in a tag import options might not be added
if a file import results in 'ignored' but somehow gets an sha256 hash (local imports hit by file filtering rules would get this), no post-job content updates like urls or tags will be added to the file
in edit subscriptions, the 'set file/tag/note import options' stuff is merged into one and there are copy/paste/favourite buttons like in options->import options with the same clever paste-merge tech. there is also a new column to see which subs have custom import options set. just a bit easier to mass-manage custom options here now
fixed a bug with adding a watchable URL with a custom import destination (via Client API); previously, it would just to the first watchable page. now it checks and creates a new one with the desired custom location import options as needed
tag import options now present a single-line summary (they'd spill over to multiple lines when things got complicated), with some tighter syntax

client api¶

the get_client_options call no longer presents a default_file_import_options summary row
Client API version is now 91

non-hydev do not pass this line

some last-minute fixes and tweaks¶

added some repair code for if your 'global' options container set is missing or missing a sub-option type. the holes are filled in and the fixed structure is saved back
wrote some more safety code to ensure a non-full global options container cannot be set to the core options structure with the issue reported nicely to the user
I didn't like how the favourite buttons were still loading up the complicated custom-merge dialog on 'load', so I made a quick-multi-load 'load' and slow-single-load 'custom load' menus. if you just want to set x to a big selection somewhere, no problem
same deal with the new general 'import options button'--I added a submenu to pull from the favourites, and you can 'load' in one click or 'custom load' if you need something like 'hey I just need the prefetch options from this template'
in the main edit panel, updated the options summary strings, which were in the form file filtering: default/stuff to default file filtering and > file filtering: stuff. this makes the non-default stuff stand out better, but maybe we can do more here. maybe this guy could evolve into a checklist or a multi-column list
the new vertically stacked listbook now disables horizontal scrollbar too (the scrollbar was drawing over the bottom item, let's go)
fixed an issue with my migration routine where a new container would not contain prefetch import options with the 'force refetch' stuff from tag import options when the original file import options was missing/default
updated the migration routine to abstain from adding prefetch stuff if it was uninteresting
updated all my importer migration code to safely attempt to determine the already-migrated defaults for this importer before migration so the 'if we don't disagree with our parent's defaults, we can set that sub-options to default' logic can kick in. this eases the transition and ensures we don't end up with a bunch of importers with default 'tag filter: allows everything' stubs because the user only changed the parsing rules
fixed a little thing in the import options manager initialisation where the defaults set the 'quiet' presentation rules to the Client API setting--the Client API import does not publish files anywhere, so no need for that
fixed an issue where the 'panel types to load for this import caller type' CONST reference was being unintentionally updated and thus poisoning later dialog opens
ditched the 'set and check full' system as too brittle with bad error handling

boring import options data migration log¶

all importers' storage and final pipeline: hdd imports, import folders, simple downloaders, url downloader pages; a couple of misc 'download these urls' routines used by stuff like IPFS, hydrus file repository download, gallery imports, the multiple gallery import container, watcher imports, the multiple watcher import container, client api imports, and subscriptions
subscription query 'additional tags'
migrated some network report mode stuff over to the new system as a frozen import options container is assembled with potential URL data
ancient legacy subscription update routine
subscription serialisation unit test

boring cleanup and refactoring¶

the import options manager now tracks if it is dirty and needs saving, and the main client controller consults that in the normal maintenance and shutdown cycle
removed the default file import options from options->importing and tag/note stuff from network->downloaders->default import options
deleted the old system's legacy edit panels and widgets
removed some hackery that made gallery/watcher list selection-inspection a little faster but was too beardy
the old show_downloader_options display option is replaced by direct import_options_caller_type inspection in all situations
wrote an ImportOptionsMetatype for our seven new sub-options to help some typing inference. all these guys now know their own import_options_type too, and some Set stuff is a bit simpler
moved the subscription query 'additional tags' button to a simpler single-panel-dialog solution
updated the UI for all importers to use the new import options container button
figured out some default/display fixes for the newer container edit UI panels to show the correct 'this defers to default type x' whether editing a default, global, favourite, or a specific importer
cleaned up some prefetch options juggling
improved some edge cases in the process by which paged importers figure out their best-guess current location context
moved some legacy presentation import options consultation in the gallery/watcher sidebar list menu is over to the new system
when a new url import page call wants to queue up urls with a custom import destination or, now, prefetch import options, this is merged into an import options container. as an edge-case logical change here, such jobs will no longer be added to pre-existing pages that have the same import destination and prefetch settings but other options differences. this is really in the weeds, but I may change this behaviour and just merge them into existing pages but with different import options; we'll see
a check that regularly goes 'hey are we still good to import stuff?' in all importers that reviews the validity of the destination import location (generally checking if we want to import to a local file domain that has since been deleted) has been cleaned up, refactored into one central location, and now consults the next-pending object for specific options. if someone sets all x url classes to import to 'cool places' and then deletes that, the queue now pauses ahead of time rather than spitting out a hundred errors. the actual exception raised here has an improved typedef, too, that the veto system catches easier
moved the new system's constants to their own ImportOptionsConstants.py, as IOC, so stuff can see them better
moved the new system's manager to its own ImportOptionsManager.py
moved some 'filter these pending tags this way' from the now-defunct ClientImportOptions to TagImportOptions
the various edit/custom-paste UI now obeys the 'simple mode :^)' set in options->import options
renamed the old ClientImportOptions to CheckerImportOptions, since that is all it does now
cleaned up some old rubbish subscription.ToTuple mess
cleaned up some misc bad code in hdd importers
deleted all the legacy UI code for the old system

Version 670 ¶

misc¶

when you paste queries into the edit subscription window, it now shows you the pending changes, i.e. which are new/already in/DEAD, before you say yes/no. if you have DEAD, it asks with a yesyes/no if you want to revive or not
the 'rescue off-screen window' system now tries to slide windows down and/or right before falling back to 'topleft of primary screen'. for instance, if the off-screen dialog's originally proposed bottom-left corner would be in view on a particular screen, it now just slides the guy down a bit and adds any fuzzy padding. this makes the new 'put dialog over the mouse' mode nicer to work with near screen edges, although I need to do some more work to better handle particularly short dialogs and those that span across a multi-monitor border
for the new 'put the dialog over the mouse' mode, if the dialog would appear off-screen, it no longer gives you a 'woah, that was offscreen' popup
for the new 'put the dialog over the mouse' mode, if the current mouse cursor is so invalid it is seemingly not on any current screen, it falls back to 'topleft of parent mate'
if you run from source with a normal-looking 'venv' directory, any update that includes new package versions is now going to say, 'hey, looks like you are running from source. this is a good time to update your venv' when you update. this will happen in this release
thanks to a user, the F13-24 keys are now mappable in the shortcuts system

tag suggestions¶

in 'manage tags', on the 'related tags' panel, the 'do it just for my files/all known files' flipswitch is now visible on local tag services
the 'related tags' panel now has a tag service selector, so you can say 'hey provide related suggests from the PTR' for your 'my tags' and so on
the 'related tags' panel now has a 'searching through/not searching siblings and parents' button. this involves some black magic and I don't really know how well it works, but the tech seemed to be done db side so I just wired it up
this is mostly experimental. if you use this a lot, give it a go and let me know if any of it is useful. if you find you are setting something one way or another over and over, it might be nice to have it remember the last settings
the 'related tags' panel now displays sibling and parent relationships, like the other tag suggestion panels
if you hit either of the 'related tags' search-type buttons, or the media changes, the 'related tags' search now refreshes using the same duration that was last searched (or 'quick' if none has been done yet)
on tag menus, the 'favourites' sub-menu has been overhauled. for KISS, it now only appears when you have one tag selected, and it provides add/remove options for the general favourites and, now, all the 'most used' tag suggestion columns that appear in the manage tags dialog. ten points to anyone who can come up with a better names than 'most used tags' and 'favourite tags' that represents what they do while not confusing one for the other
on tag menus, an empty 'search' sub-menu no longer appears in contexts without any search-page stuff (e.g. manage tags was doing this)

curl_cffi test¶

a user pointed me at the curl_cffi library this week, which provides http/2 and http/3 support with a requests-like interface. I have written a very small and basic shim to test this out with our network engine, and a new test mode is under help->debug->network actions->curl_cffi test mode. it asks for some test details and then switches the whole network engine over.
very advanced source users can now add curl_cffi to their venvs and give it a whirl. if we discover it solves some problems that requests at http/1 can't handle, I'll formalise the test and plan out how to integrate this into a future domain manager so we can shape things to particular network contexts as needed
I cleaned a bunch of session code and figured out some nicer session switching/reinitialising-tech and cookie-migration tech. I feel even more optimistic about also playing with httpx, which is a comprehensive requests-like replacement that offers http/2

boring import options work¶

wrote out a draft of a help document for import options at https://hydrusnetwork.github.io/hydrus/getting_started_import_options.html. I'll populate this with nice diagrams and screenshots and insert it properly as this becomes real. if you have been following along here, feel free to have a look and let me know where I'm making mistakes

boring cleanup¶

added a little note to the 'running from source' help regarding ffmpeg on LTS Linux flavours--you probably have an old system ffmpeg, for good reason, so I now say how to find a newer one for hydrus if you like
added a small section to the backing up help on how to set up a wineprefix for ToDoList. if you try it, let me know if you have a different mfc experience
cleaned up how mouse coordinates are fetched across the program; everything now goes through one place. Wayland has some issues with this (generally, in Wayland, I only get the last place the mouse was seen over my UI panels), so I may end up inserting some DEBUG recoveries for when this gives crazy answers
decoupled a bunch of session and cookie code. the way cookies are fetched, inspected, and edited is a good bit cleaner and more centralised and abstract now
'session' cookies are now cleared faster and each session maintains its connection pool less redundantly
misc refactoring to tidy up the initial import as the main controller boots. still lots to do
pulled ClientMedia apart and cleaned up some inheritance mess. this whole thing is still awful, but this was a good step
cleared out more linting issues with client media and friends, and in continuing a very long rewrite, a bunch of MediaSingle stuff is replaced with MediaResult, and a bunch of edge-case 'we wanted to show a collection, but it just lost its last member' bugs are fixed
fixed 'copy thumbnail bmp' for collections, which wasn't falling back to a nice default behaviour before

future build committed¶

This release commits the changes tested with the recent future build. The test went well, and there are no special instructions for the update. Source users are encouraged to rebuild their venvs this week. Update as normal, and you will get--
requests from 2.32.5 to 2.33.1 (security fix)
chardet from chardet>=3.0.4,<6 to chardet>=3.0.4,<8 (requests dependency issue resolved in the new version)
OpenCV from 4.11.0.86 to 4.12.0.88 (normal update)
dateparser from 1.2.1 to 1.4.0 (had a build problem before)
adds tldextract 5.3.1, which provides for more sensible '.co.jp' top level domain recognition
setuptools no longer pinned at 78.1.1 for the builds (an old hack removed), seems to now be ~82.x
pyinstaller 6.14.1 to 6.16.0 for the builds (normal update, and by the hand of fate it worked)
test version of OpenCV from 4.13.0.90 to 4.13.0.92 (build dependency fix that'll hit us later)
test version of PySide6 (Qt) from 6.10.1 to 6.10.3 (normal update, but we discovered a fun bug yesterday in 6.10.1 that broke several lists and widgets)
Windows mpv dll from 2023-08-20 to 2024-08-18 (cautious update; we've had many issues trying newer mpv dlls over the years)
action-gh-release from v2 (Node 20) to v3 (Node 24) in the Linux and Windows build scripts
thanks to a user, the Docker package build files are also now Node 24 compatible

Version 669 ¶

misc¶

thanks to a user, the Paper Dark QSSes now have colours to stand in for the stuff in options->colours
thanks to a user, fixed some bad shortcut enums that probably broke 'media-next' and 'volume-up/down' as mappable keys
the 'manage times' dialog now has copy/paste buttons beside the file modified time, archived time, and last viewed times. works the same as the copy/paste buttons inside the smaller edit dialogs. they'll copy a float like 1776805484.252 but will eat pretty much anything
the rich copy button menu at the bottom of on 'edit times' now lets you copy just the: file modified time; archived time; last viewed times; and web domain times
fixed a traceback when hitting certain 'move thumbnail' shortcuts on an empty page/selection
added info to 'big updates' and 'running from source' help regarding how to checkout a particular tag with git and how to discard and stash changes when pulling
added a --no_qt_multimedia debug launch argument to disallow any attempt to import QtMultimedia, which drives the QtMediaPlayer. in certain environments, this guy will cause a segfault, either on boot or when opening file->options and my newer audio device check hits it, let's go

some window options¶

the 'regular_center_dialog' entry in options->gui is now split into quick_select_dialog, quick_yesno_dialog, and quick_entry_dialog. all defaulting to 'center on parent window'
more of the 'select from a list of buttons'-style dialogs now have definitions, usually the new quick_select_dialog
regular_center_dialog is removed, and also, if you have it, a deeply_nested_dialogs, note the plural, stub that was never used
in options->gui you can now set that a non-remember-position window spawn centered on the current mouse cursor position. I daresay this is kino for mouse-centric users on the new quick-dialog entries, but try it for yourself
gave the list in options->gui a label brush-up. hope it is easier to pick out what is going on now

boring import options overhaul¶

I made it more pleasant to load and paste import options around the upcoming UI (which you can preview under options->import options while in advanced mode)--
wrote a 'custom paste' dialog that shows you the source and the incoming paste as two checkbox lists and a clear preview of the pending result. this dialog sets up a merge-paste by default but has buttons to quickly set up fill-in-gaps-paste or replace-paste. it has special labels for when the incoming paste makes no changes and also a couple commentary labels if the paste makes no changes at all or if the source is 'global' and thus has special rules
added 'custom paste' to the top of the paste button menus around here as the default way users should handle pastes while still allowing quicker mass-pastes with the old entries
the old entries in the paste menus now have simple labels to reflect what the 'custom paste' dialog uses, pushing spurious explanation to tooltips. also, the safe and 'what users want to happen most of the time' merge-paste is now the top of the three
hooked up a new 'load favourite' menu off the star button to the custom paste dialog, so you can now load without having to juggle via your clipboard and you get a nice clean preview of what is about to happen
added a star button to the url class list too
the individual import options panel now has a copy button, so you can just copy a single set of options nice and quick
the rules around pasting to 'global' are tightened up
when pasting an import options container into an 'edit import options container' panel that is in 'simple mode' and has a limited set of viewable options, the paste object is now filtered to that sub-selection of available import options types

removal of UPnP¶

as has been long-planned, hydrus no longer does anything UPnP related. this is old port-forwarding tech that could apply to the Client API or a Server service if you jumped through several hoops and put 'upnpc' in the install_dir/bin dir, but for most users it never fired. I removed the exe from regular installs years ago, removed a upnp mapping dialog two months ago, and am clearing out the last server port forwarding tech today. if you want to set up port forwarding, run your own solution particular to your hardware, not my old garbage!
specifically--
the client api service and serverside services no longer offer a 'upnp port' value to edit in their UI panels
upnp readme is removed from install_dir/bin
the primary controllers of the client and server no longer spin up a upnp manager/daemon to serve these jobs
removed HydrusNATPunch and TestHydrusNATPunch

other boring cleanup¶

the way service names are fetched has a new safe variant that recovers from a missing (e.g. recently deleted) service properly, and every call for a service name is now safe across the program. this fixes an issue with deleting a tag service that is currently in focus in a search page (and likely several other similar issues, and some general skipped summaries and such for missing services. now it'll explicitly say "unknown service" universally when this happens)
cleaned up the copy/paste logic in 'manage times' dialog. it was all good before, but a little ugly in how it decided whether to grab/push times wrt non-visible widgets
decoupled some datetime copy/paste code so different datetime widgets can share the timestamp parsing
KISSed a domain-umbrella checking trick that helps us migrate to tldextract
did a touch of misc linting

Version 668 ¶

mistakes¶

HOTFIX: I made a v667a last week to fix a regression where subscription queries with outstanding file queues were downloading those files even if the individual query was paused. The 'finish jobs even if you are DEAD' check change was too lenient first time around. Sorry for the trouble!
HOTFIX 2: I then made a v667b since the first one didn't get everything. individually paused queries that had a due-to-work 'do gallery search' check were 100% CPUing the subscription daemon. I fixed it in the same way as the file check and added additional guards for other 'cannot work' error states and calls. the run logic in individual subscription queries was not clean, and a little change revealed the faults. I regret it and apologise to those hit with inconvenience and frustration here
I have added unit tests for query pause, DEAD, 'has outstanding file work', 'has a due sync', 'is checking now', and 'is expecting to work in the future' states. this specific error will not happen again

new media viewer shortcuts¶

thanks to a user, we have some more window and media viewer options, including some clever debug stuff for those who get annoying 'rescued from offscreen' behaviour--
you can now map 'always on top: on, off, flip on/off' in the media viewer's shortcuts set
you can now map 'frameless: flip on/off' in the media viewer's shortcuts set
in options->gui, you can now control whether an offscreen-rescued window gets the (default 40, 40) top-left safety padding
you can also only remove the safety padding only when doing the 'resize to media' job for the media viewer
and you can set the fuzzy 40px amount to something else

misc¶

renamed a handful of shortcuts action labels from 'verb thing' to 'thing: verb' for better sorting
ADVANCED: added settings to options->importing to expose the until-now hardcoded bottlenecks on how many particular import queues can run at once (e.g. 'gallery files'). this is the thing that handles a 'pending' vs 'working' state in import page UI. it has an 'ADVANCED' warning label

client api¶

I chickened out of removing/default-off-ing some deprecated calls in the Client API as I had planned. it is not the season for causing fresh headaches, and nothing will break today
since we'll indefinitely live with some mess, I decided to make a quick update to The Services Object. I hate that I made the service_key the key of the Objects involved here since in some JSON libraries it is a real pain to navigate. thus, anywhere I add the old Services Object under services, I also now add services_v2, which has the services as a list. same format except they now have a service_key field. I will not remove services any time soon, so no need to change anything you already have
the Client API docs now refer to the services_v2 structure so that new guys and projects are steered towards something non-weird
I concede this is spammy and inelegant, but it resolves and KISSes an early mistake. the cruft is intensifying to the point where we may not be able to contain it, but why contain it? let it spill over into the add-ons and the MCPs, let the scripts pile up in the streets. in the end, they will beg us for a Rust conversion
Client API version is now 90

db dir build stuff¶

I reworked how the 'help my db is broke.txt' help files (and for Windows, 'sqlite3.exe') end up in a db dir. these files now primarily live in static/db_files. any .txt files in there are mirrored into a new database folder on db creation. I strongly considered updating these files on any update, but I went back on it, preferring not to auto-fiddle with the db dir overmuch--I'll make a manual command for it
a CONTENT_DB_PATH variable that had caused some hassle for patched source installs like the AUR release is now gone
I will be removing the 'db' dir from the main repo in the future. I'll do more planning and testing around .gitignore just to be certain doing that will be smooth. we obviously want this to be absolutely failsafe
I also expect to eventually migrate these .txt files to proper .md help files, with formatting and everything, at which point this will scale back significantly, probably just to a single .txt that says 'if everything is busted, go here in the help'
I am finally ready to plan moving the default location of the db outside of the install dir. we'll have a first-start wizard to select/find location and a small file in your user folder to track available profiles or whatever if the client is run without -d, or something along those lines. this transition will take careful planning, so your general thoughts are appreciated

boring import options overhaul¶

every single importer across the program now converts the legacy file, tag, and note import options to the new 'import options container' before passing it to the main file import object. the whole file import object pipeline below the importer level is now converted to the new options structure!
cleared 'file import options' out of some of the parsing pipeline where it wasn't used any more. this seems to have cleared all import options out of the gallery pipeline
wrote out a modern 'set up a default set of options' call following the existing loud/quiet dichotomy and 'post urls get tags; watchers don't' ruleset
my in-testing 'import options' panel gets a 'reset to defaults' button that will reset the main defaults, all the custom url classes, or the favourites/templates
fiddled with some of the wording in this UI. defaulting the defaults back to 'use defaults' etc.. can make for unclear tooltips
fleshed out the container object to offer up different import options in a more convenient and nicely typed way in the real pipelines
in the file import, the 'should fetch metadata even if url/hash recognised...' advanced setting is now strictly any metadata. previously this checkbox would only kick in if you actually had some tags set up to parse and import somewhere. after the migration, these checkboxes will be moving to an independant 'prefetch import options' guy, away from tags completely
certain non-already-in-db imports that have no tags or notes are now one db hit cheaper
certain already-in-db imports are now one db hit cheaper

other boring stuff¶

fixed a possible count bug in the thread slot system where in certain late-initialisation systems it could undercount the number of active jobs in a slot
the 'prefer system ffmpeg' setting no longer needs a client restart. the call that discerns the ffmpeg path to use is also a little more sensible and failsafe, and cycling the options dialog now resets some state so you can put an ffmpeg in your hydrus 'bin' folder, ok the options, and it'll redo the 'is there one there?' test
updated crash.log to /db/hydrus_crash.log in .gitignore file

Version 667 ¶

misc¶

fixed an issue with subs where a sub with all DEAD queries was considered 'no work remaining', and thus subs in that situation with nonetheless outstanding file downloads would not clear those outstanding file downloads if the work spilled over to future sub runs
all thumbnail pair lists (which are in the duplicates auto-resolution UI) now have a menu that says 'show x rows in a new page' so can you can review things in more detail. let me know if this swallows deleted thumbnails anywhere
the file storage granularisation system can now handle cross-filesystem granularisation. this can happen on 3-to-2 when some of the 3-granular subfolders have been migrated to a second location and need to be collected in a central source that might be on a different partition. it seems to also occur in 2-to-3 when the filesystem has overlapping subfolder orphans from a previous borked migration or similar and things are tidied as part of the regranularisation. the previous very fast atomic 'file rename' was raising an error here; now I do some efficient device checking of subfolders and move to a 'copy2+unlink' combo instead when the device differs
the granularisation panel now warns those with weird symlinked storage to stay away
the routine that counts up the durations of a gif's frames and the hydrus native renderer now fall back to best-guess defaults if A) the file changes resolution mid-parse or B) any frame appears to have a width or height bigger than 16k pixels. we encountered a malformed gif this week that went from 640x480 to a fake 28579x93263 and caused a memory explosion. should handle this situation better now
continuing the 'clean up PIL Images better' work, native animation renders that error out are now explicitly closed before reinitialisation
thanks to a user, the 'Paper_Dark' QSS files have fixed, visible ▲/▼ button labels. relatedly, the objectNames I added for this guy last week are renamed to HydrusCollapseArrowUp/Down. they were Expanded/Collapsed, but since the arrows reverse depending on an upwards or downwards expand, we realised we'd want something more specific to the actual label being shown if we ever wanted custom icons/labels

sidecars¶

sidecars gets an easy one-click template that covers simple stuff. the 'edit metadata migration routers' dialog, which the main guy in sidecar editing, now has a star button. when exporting to sidecars, it offers an 'easy one-click JSON that covers the basics' template that spams out definitions for a single 'filename.jpg.json' JSON sidecar that gets your notes, urls, tags for each tag service by storage (no siblings/parents) and display (with siblings/parents), and timestamps for archived time, main import time, and media/preview viewer 'last viewed time'. when importing from sidecars, the buton offers the same 'easy one-click' option to eat from a JSON using the same format and will line up any tag services that have the same service name
if you try to export notes to a .txt sidecar using \n as the separator, or conversely import notes from a \n .txt sidecar, the 'edit sidecar stuff' dialog now moans at you on ok, asking if you are totally sure that is correct. since notes often include newlines, something like |||| is more appropriate
when dragging and dropping individual files onto the client, it can now give a pretty 'sidecar' result when you drop individual files. this only worked if you dropped directories previously, with individual file drops giving 'PROBLEM: filetype unsupported' for the sidecars. this is an aesthetic change only; it doesn't matter that sidecars are not parsed ok at this step, but it is nice to have the correct 'won't try to import' label
in the 'add metadata before we import' sub-dialog of the manual file import, the 'sidecars' list now shows better metadata context around the sidecar content preview. the 'notes' now compresses multiple notes better, with a total count and elided preview; the 'tags' now says to which tag service they are going; the 'URLs' now says how many URLs and elides better; the 'timestamp' now converts to a nice datestring and says what the destination is for that time (previously it just said the UNIX timestamp flat wew)

Client API and setting archived time¶

for sidecars and the Client API, if you push an archived time on a file in the inbox, the file is now auto-archived! the UI and Client API docs now mention this
previously, the data was (buggily) recorded but generally swallowed and not reported anywhere
the Client API 'file metadata' call now gives a time_archived if the file is not in the inbox and there is a known timestamp
sidecar routers, when harvesting timestamps to send to a sidecar file, will no longer suck an orphaned archive timestamp from a file that is in the inbox
Client API version is now 89

QtMediaPlayer transparency¶

the QtMediaPlayer should now support transparency properly! previously, if you put up an animated gif with transparency, the transparent areas would be black with ugly fuzzy edges. now it does a checkerboard or greenscreen using the same settings as a static image. on some files there is still fuzz, but most of the time it looks great
the QtMediaPlayer should also show the normal background colour while it waits for a non-transparent media to load. it probably flashed black before, so should cut down on some load flicker
the normal static image checkerboard and greenscreen transparency uses the same, updated tech here. rather than drawing a bunch of boxes, we now have a textured pixmap brush, and in fact in all cases we just set a brush and do an eraseRect, nice and clean. Qt is great
the 'show transparency and checkerboard in duplicate filter' setting in duplicates is now its own thing and applies independently to the one in media playback. previously it had some whack 'only applies if the media playback guy is unchecked', which is the kind of thing you can convince yourself is useful and technically correct but violates KISS

ingesting serialised objects¶

the 'import downloaders' panel now accepts drag-and-drop or pasted JSON files
added a help->debug->what is this object? debug review panel to figure out mysterious bits of json or png or whatever

boring stuff¶

improved error handling for setting QtMediaPlayer null audio device on media load
improved how 'show menu' events are handled, Qt-side, for the: colour picker widget; preview viewer; navigable media viewer; taglist
cleaned up some ugly menu generation/show for the main thumbnail grid
decoupled the 'import downloaders' serialised object ingestion routine and brushed it up a bit with nicer error reporting and so on
all serialisable objects can now describe themselves with regards to the serialisation system, with special hooks for named objects and the special list, dict, and bytes dict
fixed up several places that do a fast 'is it a jpeg/png' mime check, which was accidentally skipping the png side and doing the slower check
fixed a bad layout flag in the 'select from a list of checkboxes now mate' dialog
fixed some close-curly-brace typos in the example Client API Services Object JSON

☠⛤ Version 666 ⛤☠¶

misc¶

when you create a new subscription with 'add' button, it now asks you for the downloader up-front. this was easy to forget to set, previously; now you can't miss it
the problem where media could sometimes not receive any content updates (rating changes, new tags, archive state, whatever) after a backup, lock, vacuum, or certain database transaction error recovery until a page was refreshed is fixed. a cache that the in-session pages refer to was held by a database module being reinitialised through these processes and thus, after cache reset, the connection to send the update signal on media changes wasn't getting through to UI
added ^(?!...$).+$ to the regex input menu's quick selections for 'anything except this exact phrase'. I have to search stackexchange to figure this out every time I need it; now it is just there. also fixed a formatting typo in this menu for the 'any of these' entry

media players¶

improved the guard around the QtMultimedia module for those Qts where it is not available or otherwise fails to import
if your options say to show something with QtMediaPlayer but the module isn't available, it now falls back to an 'open externally' button and does a little popup; the same handling for missing mpv
improved the popup text when you boot help->about with an unusual module load error to report
rejiggered the help->about 'Optional Libraries' list a little and moved mpv over there
added another mpv 'null' audio error text, 'Could not open device', to our new cascade of 'calm down, but set the null audio device now' error handling
added a KDE and QtMediaPlayer note to the 'installing - linux' help. installing mit-krb5 may help

qss¶

the e621, e621_redux, and Paper_Dark qss stylesheets get 'built_release' alternates to handle the new 'lib' directory structure of the built release
updated the guidance on external asset paths in options->style

import options test panel¶

booting the options dialog in 'advanced mode' now adds a 'PREVIEW: import options' panel. this is the new central place that we'll be customising import options after my rewrite here is done
this panel saves nothing; it is only for playing around with. it takes your current settings and migrates them to the new system as the expected pending migration will
I am happy with the technical side of things here, but this thing is complicated and I would like to throw advanced users in the deep end and get feedback on where the roughest parts are. I'll sand down the roughest corners with better UI and consider it all as I revise the help docs
april fools this year is yet more hellish UI from hydev

removing the default downloaders¶

a fresh hydrus client database no longer starts with any downloader definitions! I am cleaning some legacy things and making the program more neutral from the start. users will now add whichever downloaders they want from third-party sources like the user-run downloader repo. not an april fools--sorry for the trouble!
existing users keep what they have; this only affects new installs
this also frees me from the maintenance burden, which I was failing at. several of the defaults were breaking and I was not keeping up with the issues
so, the defaults are removed from the static dir and there are no 'add from defaults' buttons in the edit panels. the UI now supports and starts with a 'null' state of 'no downloader selected' everywhere
users without any downloaders are pointed to the network->downloaders menu and the user-run repository
I updated the help documentation all over regarding this change. it may need a revisit to make new-user onboarding smoother

import options boring work¶

added a 'paste-fill' paste type to the options panel, for filling in gaps without overwriting
wrote paste buttons for individual options panels and the import options container panel
wrote copy buttons for the container panel
fleshed out the valueChanged signals for all the new options panels: prefetch; file filtering; tag filtering; locations; tags; notes; presentation
wrote a favourites button and adjusted the UI to handle it in the appropriate locations
wrote some basic default favourites
polished the UI, swapping to icon buttons, bunch of stuff
fixed some little logic and label bugs in the new import options stuff

little stuff and cleanup¶

the little ▲/▼ button you see across the program on static boxes and the popup toaster now uses a special shared class that uses nicer Qt tech to signal things and sets two new object names for QSS devs to hook into: HydrusCollapseArrowCollapsed and HydrusCollapseArrowExpanded. note that these guys use reversed labels depending on whether they expand up or down (I just realised this, working on it now), so if you want to put a new label or icon in, I think we might want to do some more here. let me know how it goes
I think I fixed an issue where the 'hey your page is super mega tall, careful' message was showing more than once if the session includes multiple pages in this state
updated some new async ui code, the stuff that does new unified callBack/errBack panel restoration after job reset, so that it better handles the error of its parent window being closed early, and a program shutdown call. this was causing some harmless error spam in some things like a delayed subscription popup files button press/dismiss
fixed a stupid error in the updated Linux .desktop file deployment script that broke the Icon with quotes around the path
made the various widgets, options, and importers ok with having no downloaders available. the first time a GUG comes in via the drag and drop system, it is set as the default
amidst the downloader work, cleaned up how subscriptions do some edit panel Set stuff, removing some legacy jank
fixed up an old hardcoded 'downloader updated' callable with a nicer Qt Signal
improved the error messages around invalid/missing/null GUGs at different stages
I also removed some legacy downloader db update code that would have fallen to bitrot here
fixed a couple finally/return syntax issues
'manage login scripts' now has some red text up top warning that it is an ancient system that does not work well
updated the 'adding new downloaders' help to be more clear about what is going on with the png files. it now has an example png for artbooru.com
updated a bunch of ancient screenshots in the help, particularly the parsing system
fixed duplicated 'index to fetch' label in html parsing formula panel
replaced some whackadoodle ListBook fixed sizing with a proper sizeHint and some live rowheight checking and framewidth correction
cleaned up some subscription unit tests
the media result cache is now a singleton instance that persists through a database restart, and the unit tests now do some careful clearing on test database resetting here
fixed a missing definition typo in the filetype labels

Version 665 ¶

misc¶

fixed an issue where a duplicates page would not re-enable the 'launch the filter' button sometimes when it previously had a count of 0 but got a pair-discovery update in the background that added new pairs (issue #1988)
the 'help: random 403 errors' entry is now moved down in the 'retry' buttonlist. it will disappear in a few weeks, leaving just the network->downloaders menu item, once people hitting this menu have had more chance to see it
fixed an issue where if you selected some files with a subset of trash and then said 'move/copy n files to blah local file domain', it would try and move the trashed files and throw an error. it now filters those files out of the actual operation as you'd expect
fixed an issue in numerical rating rendering after deleting a numerical rating service (bad error handling on the missing service)
added a section to the 'contact' help page about how to send broken files to hydev (zipping them up explanation etc..)
updated the io.github.hydrusnetwork.hydrus.desktop file with easy mode help comments on how to edit things and added StartupWMClass to help taskbar grouping

more audio device stuff¶

applying the options dialog now updates all open mpv players to use the specified audio device
applying the options dialog now updates all open QtMediaPlayers to use the specified audio device
the ao/xxxxxx: There are no playback devices available mpv error, which until now has sparked the 'hey things are looking unstable with this file, so unloading it' response, now triggers a new 'hey, set all mpv windows to null audio device'. if you get this stuff when you unplug your headphones or something, let me know how it goes
when a new QtMediaPlayer initialises, if the desired audio device is invalid, it falls back to auto. if the auto device is invalid, it resets to null
new DEBUG checkboxes in options->media playback allow you to auto-set mpv or the QtMediaPlayer to 'null' audio device when playing silent media like animated gifs. I know we've had some issues around this over the years with mpv on Linux in particular, so let's see how it goes. this used to be default behaviour for the QtMediaPlayer btw; now it isn't

some boring cleanup¶

improved how media objects determine if they are in 'combined local file domains'
a database routine used in repository sync and 'are we mostly caught up to this repo?' that fetches missing update hashes is now significantly faster
the QtMediaPlayer no longer leaves the video output disconnected until the first non-audio file is loaded. things are more KISS
the mpv and QtMediaPlayers recognise better their current audio device and can trigger update calls only on changes
cleaned up some more QtMediaPlayer output setting generally
updated some critical error text when trying to boot into a database that was created--and failed to initialise--on the last program boot
removed some OpenRaster debug statements

removed Qt5 gubbins¶

Qt5 (which for us means PySide2 or PyQt5) is well behind us now, so any lingering support is just getting in the way. I removed it all this week. if you are struggling on a hyperpatched Win7 machine, forgive me but it is time to move to Linux
removed Qt5 initialisation code
removed debug code that tests for Qt5 support
removed old thumbnail UI scaling Qt5 hack
removed Qt5 stylesheet hacks
removed Qt5 media panel swap hacks
removed About Window Qt5 stuff
removed an old Qt5 qtpy init hack
removed Qt5 patches for mouse and drop events
removed Qt5 QPDF check
removed Qt5 QKeySequence conversion

nicer PIL memory cleanup¶

PIL images are now closed promptly (freeing up memory better and faster) in more locations: the visual tuning suite; jpeg quality estimation; icc profile inspection; embedded metadata inspection; exif inspection; decompression bomb testing; 'show file metadata' window; import metadata gen; icc profile inspection on load; on forced PIL loads that error out on conversion to numpy; on another standard method to load images; variable frame duration fetching, 'get number of times to play animation'; 'animation has valid duration'; serialisable object import; ugoira rendering; ugoira API rendering; ugoira property fetch; ugoira json property fetch; Ugoira thumb gen; PSD rendering; PSD thumb gen; the native PIL animation renderer; on EXIF rotation conversion; some weird dequantization; resolution fetch; when some thumbnail stuff errors out; some animation property fetch; OpenXML thumbnail gen; Paint.NET thumb gen; Krita thumb gen; Krita rendering; openraster thumb gen; openraster rendering
I went overkill here and yet there are still some gaps. I got all the file loads though, I think, which is the main stuff here that I think was lagging. some of it is also a little ugly. we'll see if this improves some lazy memory cleanup during heavy import. if it helps, I may revisit to clean up

import options overhaul¶

wrote migration code that takes the old file/tag/note import options and produces a new populated ImportOptionsContainer
updated the prefetch import options to track the 'fetch metadata even...' checkboxes, although they will do nothing until migration
updated 'note import options' to a legacy object that can deliver a trivially shucked 'note import options' that has no 'is default' properly and works in the new system. the edit notes dialog, Client API add-notes call, duplicate content merge options, notes sidecar exporter, and unit tests now use the new object
wrote an edit panel for the new note import options
the legacy note import options now holds its shucked version inside it, now only taking responsibility for the 'is default' property otherwise, and all import contexts now consult the new object for work
juggled around my options stack preference again for simplicity and added 'import folder' and 'client api' options import contexts
brushed up the UI significantly with new labels, better options summaries, better help, a KISSer workflow that filters out overengineered options by default, a little description label for each import type, and another for each options type
made the stack description and display clearer and added it to the url class section
fixed some default options display
brushed up all the import options summary statements and rearranged them all into single-line
updated some tag filter label grammar. in some cases it was saying 'tags taken: allowing all tags', which comes from internal permissions language, when just a 'tags taken: all tags' was a better fit

Version 664 ¶

misc¶

the media 'delete' menu is no longer a flyout if there is only one deletion option (you should see 'delete from my files' more often)
the preview viewer now has the same style of delete menu as the canvas and thumbnail
the system tray options are no longer disabled on non-advanced non-Windows. this stuff works better these days

audio devices and tracks¶

you can now select the output audio device for QtMediaPlayers under options->media playback. default is 'default' (issue #1985)
there's also a "DEBUG: null" choice to say 'never load any audio output device to QtMediaPlayer'. I know we've had some users who have had trouble with this
I then did the same for mpv. it works a little different, so you hit a button to fetch the available options and then select from there, or type in manually if you know otherwise. similarly, you can select a 'DEBUG: null' option
QtMediaPlayers now show their audio tracks in the player right-click menu. it says title, language, and codec, depending on what is available. you can select another track and it changes instantly!
QtMediaPlayers also now show their video track(s)! I added the same 'switch video track' call as the audio stuff. if anyone has a multi-video-track example vid, I'd love to see it for my testing purposes, but the audio side was a dream so I assume it just werkz
some this stuff is very slightly hacky, so let me know how this works for you

file import object inheritance cleanup¶

tl;dr: some downloaders save modified time better
when a gallery parse or a file post parse creates multiple child file import objects, I have cleaned up how the parent gives the children data--
source time is now only propagated if the parent source time is A) sensible and B) the child doesn't already have an older timestamp. all instances of file import object source-time-setting now follow this rule, which is how modified times are generally updated elsewhere in hydrus. 'older is generally more useful and trustworthy, unless it is new year's day 1970 etc..' also, file import objects now clip to thirty seconds ago when given a timestamp from the future (happens with timezone fun sometimes). thanks, particularly, to the user who identified and chased this down (issue #1984)
referral url is similarly now propagated more softly; only inherited if it isn't set beforehand (was previously a forced overwrite in all cases. not sure it actually matters, but it might in future)

some boring cleanup¶

some critical image rendering sections now clean up their memory quickly and explicitly rather than waiting for the garbage collector to handle it later. more to come here in future
cleaned up how AsyncQtJobs do UI restoration after an error, harmonising how the callback and errback restore the UI
decoupled how some exception stuff is caught and processed and rendered for the user, and fixed some error-reporting pathways that were not rendering nicely

boring import options overhaul¶

I juggled some more pending import options stuff around, giving the wilder stuff a KISS pass
wrote a panel to handle editing the new defaults. I simplified things a good bit and moved it all to the options dialog. it is hidden for now but I feel fairly good about it all
filled in a bunch of holes and fixed some display bugs as I stitched it all together
improved how import options and their containers present for network vs local imports
got the import options editing dialog to remember the last selected options type
I've now got to write some favourites UI, polish this all, write some migration tech, and then update the import pipeline to handle it all. feels doable

Version 663 ¶

misc¶

the hide and anchor mouse cursor during media viewer drags setting under options->media viewer is now split into the 'hide' and 'anchor' parts, to add flexibility for trickier situations. some window managers aren't happy about mouse warping. the hide logic also now kicks in faster and sticks better
added checkboxes to options->files and trash to control whether trash maintenance and then deferred file delete can happen in 'normal' time
when you ok the 'manage tags' dialog, the commit to db now occurs in a thread and will no longer block the UI. if the job looks big or otherwise takes longer than a second, you'll get a progress popup, which is now cancellable. let me know how this feels on something big like the PTR (issue #1980)
a variety of videos that have silent audio channels were registered incorrectly in the database as 'unknown silence' and were not returning with system:no audio searches. the typo in the file parsing code that caused this is fixed, and all affected videos will be queued for a rescan on update to v663 (issue #1977)
fixed a recent typo that was causing the 'retry 403/404/blacklisted' choices to instead retry all ignored. sorry!

client api¶

added a link to HydrusTools at https://github.com/GiovanH/HydrusTools to the Client API help page. this is a toolset for a variety of metadata management and organisation tasks, actively being worked on

chardet build issue¶

changed chardet>=3.0.4 to chardet>=3.0.4,<6 in the pyproject.toml and requests.txts to rewind us to 5.2.0, as we were a few weeks ago. this handles a requests version issue and I think also a PyInstaller incompatibility that was causing chardet to not load in the recent builds. this thing is 'character detection' and helps with website decoding

boring cleanup¶

fixed a harmless but spammy log error when the client booted with a session that included an OR predicate with certain service-based system predicates
used the same 'commit content updates and make a popup if it takes a while' routine I wrote for manage tags for media viewer delete files. I don't think this guy is going to spawn ten super slow popups any time, but if it does, they'll now be more visible if the user closes the viewer immediately after a delete during busy times etc..
when telling a 'file log' to 'retry these previously deleted entries, and yes clear the deleted record', the database clear action is now asynchronous. the panel disables while it works
archive/delete async commit block size is now 10 files, down from 64, to reduce latency as it works
all the multi-column lists in the program now have the ability to change height to exactly contain their contents (like the gallery downloader does), and almost all of them now have a defined range for this tech. most are in the 6-12 or 12-24 range, depending on the type of panel or dialog they sit in. almost all of them are happy to be a smaller minimum size, and the minimum size math here is less crazy. lists should just size vertically a bit better now
started a nicer and cleaner core layout call in the new ClientGUILayout. a years-old hack from the wx days is being replaced with nicer Qt code. the core idea is finished, and one real place uses it and nothing blew up, so the next few months will have more pushes on this and a bunch of long-term layout issues will be incrementally fixed
brushed up the error handling around stylesheet loading
brushed up the 'you don't need the server' help document
brushed up the 'running from source' help regarding venvs and different versions of python
updated some help/readmes about custom assets under your db/static folder. I now mean to recommend this in all cases--don't edit the install dir, make a custom folder under your db

boring import options work¶

moved some import options panel code around. the existing dialog and button are moved to Legacy files and will be deleted after the transition
fleshed out some of the options here to differentiate between subs, downloader pages, and all 'post file' work
reworked the options so specific url class options trigger at the correct layer in our stack of swiss cheese
added some tools to the main manager so he can give nice human descriptions on his inner workings during editing
'full' import options containers can now describe where they got each import options (e.g. 'from subscriptions default')
wrote out failsafe url class type and name labelling into this
wrote most of a panel to edit a new ImportOptionsContainer object. I'm generally happy with it so far. it'll be a vertical listbook with the list of options types above switching edit panels below, and each line in the options list saying 'tag import options: does x' or 'tag import options: uses file posts default'. this is a complicated thing that I want to end up being clear and user-friendly, albeit sophisticated. I think we are getting there
next will be a dialog to handle the defaults, and some favourites management UI, and then updating the workflow. lots still to do