Hydrus Network

I had a great week. The 'busy time' optimisation is better, the human-readable file metadata filtering is cleaner, there is an UI test for advanced users, and there are several library updates to fold in from the future build.

windows

zip

exe

linux

tar.zst

I had a good week. There are some fixes, some experiments, and a future build test for advanced users to try out.

full changelog

highlights

I may have improved the CPU thrashing we have seen during some heavy work periods. If archive/delete commit and the duplicate system are suddenly faster for you, let me know. There's more I will do here.

The system tray icon is now available for non-advanced-mode Linux users (an old safety check I forgot about), and there are a couple new BUGFIX checkboxes in the options for those who have trouble with minimising to system tray.

If you are not English locale, try out the new 'TEST: Use your locale for integer rendering' under options->gui. It will try to render most '5,000' across the program as '5 000' or '5.000' or whatever your system prefers. It seems to break a little layout sizing here and there, and system predicate parsing, but I'll keep working on this. Let me know how it goes for you.

A user noticed a library I recently integrated to do better domain parsing was fetching the latest domain definition from publicsuffix.org the first time it was ever called. I regret missing that it did this. I have put the file the library wants in the main hydrus install dir and instructed it to strictly use that local copy only.

future build

Only for advanced users!

I am making another future build this week. This is a special build with new libraries that I would like advanced users to test out so I know they are safe to fold into the normal release. More info in the post here:

https://github.com/hydrusnetwork/hydrus/releases/tag/v680-future-01

next week

I had a good week. I fixed several things, added a new locale number test, and prepped a hefty future build for advanced users to try out.

windows

zip

exe

linux

tar.zst

I had a good week. I fixed some things, sped up some UI, improved the system tray icon, and cleaned up some file metadata parsing.

full changelog

highlights

Some recent changes seem to have slowed down the manage tags dialog and perhaps some other media dialogs. I cleaned up some newer code this week and I think I found the core cause of the lag. I also optimised some older things that were working bad in manage tags. Let me know how it feels now!

I made the system tray simpler. Left-click now cycles between 'bring to front' and 'minimise', and middle-click does the standard show/hide tech. Window state is preserved through a hide/show better, and I've added the basic system tray options to its menu so we can avoid options-dialog softlocks if something breaks.

The 'human-readable embedded text metadata' (which you can see on many images by clicking the little document icon up top of the media viewer) is now purged of some spammy technical fields. Some stuff like DPI and 'this is a jfif' stuff is moved down to an expanded 'extra info' section, and the actual human-readable section should more often have something interesting in it. There's still a bit more to do here, but when I am happy, I'll trigger a big regen of this data and 'system:has embedded metadata' will become more useful.

Let me know if anything related to archive/delete or the duplicates system works faster this week.

two client api projects

There are two new Client API projects added to the help links this week:

First, 'hydit', which is a lightweight, feature-rich client hydrus client for Android. https://github.com/BashCooler/hydit

Second, 'kaimen', which uses some clever FUSE mount tech to show hydrus searches in your file explorer. https://github.com/Dry-Leaf/kaimen

next week

I had success in better integrating one of the UI experiments this week. The new code unfortunately revealed that I need to do a ton of cleanup on some of my ancient page code, but I did some of that and it was all to the better. I'll keep at it; doing some small jobs and clearing some larger background cleanup in prep for new UI tech.

I had a good week. I fixed some things, sped up some UI, improved the system tray icon, and cleaned up some file metadata parsing.

windows

zip

exe

linux

tar.zst

I had a pretty good week mostly fixing little things. I also added thumbs for audio files!

full changelog

highlights

If you have audio files with embedded images, they ''should'' get those images as thumbnails soon after you update. If you have any mp3s or flacs or whatever that do have an image but fail to generate a nice thumb, please send them in and I'll check them out! There's a secondary issue of loading these images during actual playback, btw, which I think works in some situations but not others, but I'll see what I can do there too.

I cleaned up a bunch of 'talking to ffmpeg' code this week. The ffmpeg timeout setting that was recently added now works better and some audio, video, and image stream parsing is improved. For Linux users, I am starting to recommend they get a new version of ffmpeg since the versions available on most package managers, it turns out, are super old. Bit of new help here under the Linux tab: https://hydrusnetwork.github.io/hydrus/getting_started_installing.html#installing but I'll probably soon roll a specific ffmpeg into the Linux build like I do the Windows ones, so you can, alternately, just wait a few weeks.

We've been playing around with a per-media mute/unmute setting recently (useful for people who load up more than one media window at a time). I cleaned up the logic of this this week, and a user sent in some improvements (and some cool new modular UI tech that we are going to experiment with), including shortcut actions to control this. Let me know how it feels, if you use this, and then let me know what you would like the volume slider widget to look like. Having three mute buttons feels too crazy, but what would work?

next week

I had a pretty good week. I fixed some bugs and figured out thumbnails for audio files with embedded images.

windows

zip

exe

linux

tar.zst

I had a good week mostly fixing bugs.

full changelog

highlights

A recent rework of some downloader code broke some specialised downloaders. I have fixed the problem. If a downloader you use stopped adding tags to files in v676, please queue those URLs up again in a new url downloader page, and it'll work this time.

I updated a bunch of old video parsing code. Some video parsing tech is fixed for very new versions of ffmpeg, and the whole video import process uses a bit less CPU now. If you have AVIFS or other unusual new animation formats, they will be rescanned on update to fix some 'num_frames=1' issues.

Some files that had human-readable metadata were not parsing correctly in recent versions. This is fixed and the affected files will be rescanned on update.

I optimised some thread scheduling code. If you have noticed the client lagging during very heavy work periods, for instance particularly in the moments after an archive/delete filter finishes, let me know if it feels any better.

next week

I had a good week mostly fixing bugs. I fixed some buggy video handling, missing human-readable metadata, and some types of downloaders that were not attaching tags.

windows

zip

exe

linux

tar.zst

I had a good week working on misc improvements, and there are several more submissions from users.

full changelog

more UI updates

A user has sent in more media viewer UI improvements. You can now mute on a per-viewer basis from the viewer menu; you can similarly set per-viewer slideshow shuffle and play-once-through settings.

You can also set what 'zoom switch' switches between via a new dropdown under options->viewer viewer hovers.

There is a new type of media shortcut action, an 'interactable' type, that you give only a tag or rating service name--no fixed content. When you hit the shortcut, you get a popup dialog asking which tag or rating you would like to set. Try it out!

client api new page

A user has written a comprehensive '/manage_pages/new_page' command for the Client API. If you are a Client API guy, check it out--it can do pretty much everything.

other things

The '-d' launch parameter now supports userdir ~ expansion, so '-d=~/hydrus' will now work properly. Same deal for the optional -v parameter in setup_venv.py, which source users use. setup_venv.py also supports a non-interactive mode for automated installs, either '-i=s' for simple or '-i=a' for 'advanced - all test/yes'.

The stylesheet asset test from last week went well, so if you are a stylesheet creator, check the changelog and the qss readme.txt. Users who were on a default 'built_release' variant stylesheet will be migrated to the original files this week, as the underlying asset path issue is fixed.

Some complicated downloaders were adding too many tags: particularly, tags parsed in certain early gallery sweeps, often when mixed with subsidiary parsers, were being passed down to subsequent gallery page fetches. Thanks to the users who helped investigate this--I think we have it fixed.

next week

I had a good week clearing a bunch of misc jobs and adding user submissions. There's some fixes, quality of life improvements, new media viewer options, and a new client api command for opening new pages.

windows

zip

exe

linux

tar.zst

I had a good week getting back to it. I fixed some small things, and there are some contributions from users.

full changelog

highlights

There was some old check that said 'do not do any repository sync maintenance work while network traffic is paused'. This was stopping PTR processing from naturally starting in idle mode when the network was paused. I have undone the check, so users who regularly have the network paused for long periods should see some movement now.

The command palette has this annoying thing where if your mouse is over where it appears, any result item under the mouse is immediately selected, messing up keyboard muscle memory. I've added a little hook that ignores the mouse when it first shows like this; let me know how it feels now.

A user has written a ton of hydrus stylesheets called Nereid, with a nice repo here: https://github.com/6788-00/nereid-theme-hydrus. I have integrated these into the defaults. Check them out! Users who create stylesheets should also check the changelog this week--I have a test I'd like you to run so we can fix an annoying path thing we've been dealing with.

Another user has written a KDE plasma widget that does a slideshow, right here!: https://github.com/apampurin/hydrus-slideshow / https://www.opendesktop.org/p/2362439/

And another user who has been working on a complete overhaul of the thumbnail grid delivered their work to me just recently. I have now integrated the first version as an experimental test, and it looks very promising. I still have to do lots of work before we can try out wider testing, but it looks like we'll have easier support for masonry layouts, vertical grids, more dynamic thumbnail sizes, all sorts of new display tech.

AUR package attack

Unfortunately, hydrus was touched by the AUR package attack this week. The 'hydrus-git' package, which I had never heard of and seems completely abandoned by the person who originally made it, was one of the thousand-odd packages taken over and compromised. I understand the situation has been resolved now, but if you were somehow running this package from the AUR and updated just recently, you need to check your system.

I do not maintain anything hydrus related on the AUR, but a user I know does, and this semi-official package is just called 'hydrus', here: https://aur.archlinux.org/packages/hydrus

I still generally recommend that users who want to run from source use my 'running from source' guide for a manual setup, since the AUR's need to use the newest available python packages tends to cause trouble for hydrus every now and then, but I can appreciate it is convenient.

In any case, I hope this wave of attacks gets figured out. The package manager situation is crazy.

next week

I moved to a new dev machine over my vacation. I'm now on a nice and faster box, finally moved to Linux, and all the important stuff is migrated over ok. I took some time to clean up my messy and old environment too, and I'm hoping to stay on top of a clean desktop and todo and so on going forward.

I had a good week getting back to it. I was a little short on work time, but I have fixed some small issues and there are new stylesheets from a user.

windows

zip

exe

linux

tar.zst

I had a good week doing some misc cleanup before my break.

full changelog

highlights

I've rewritten the logic of how mouse wheel events propagate from the media viewer taglist to the underlying canvas to trigger previous/next navigation events. We've had some different rules here over the years because it can get annoying to scroll a long taglist and suddenly start changing media. The new system now only propagates if you have not recently used the taglist's scrollbar. I hope it feels natural and helpful. If you would like to change this behaviour, check out the new options under options->media viewer hovers.

I fixed some bugs in cookies.txt import. Session cookies (i.e. those with no expiry) should now parse properly.

I cleaned up some repository account sync code and gave the repository panel in review services a quick pass. It should give some nicer errors when things aren't working and catch it early if you try to do 'refresh account' while network traffic is paused.

If you are an advanced user who has been playing with the curl_cffi test, I have moved it to a permanent option under options->connection. You no longer choose the http version, just the browser name. Also, the setup_venv.py routine now asks if you want the library. Let me know how it goes!

next week

I had a good week doing simple cleanup and improving some UI quality of life.

windows

zip

exe

linux

tar.zst

I had a great week working on a mix of stuff, mostly optimisations.

full changelog

highlights

A user discovered one of the database caches that accelerates tag and file load was not working great. When under pressure, it would churn a bunch and lead to memory bloat and fragmentation. I have written a nicer cache that mitigates some of the issues. Some core database functions should be a little faster and save some memory over long-term busy use. There is more I can do here in future.

File parsing when you manually import files is also optimised. The folder parse under file->import files along with the file mime and size fetching should all require fewer drive hits.

And clients now scan their file storage on boot a little quicker too. I hope you'll notice it when booting up on a cold hdd, especially if you are an advanced user in granularity 3 mode.

The database->file history chart now has an editable Y axis range.

I moved my scrappy .txt recovery text files to the help and gave them a brief markdown polish. They are all under the new 'Recovery' section starting here: https://hydrusnetwork.github.io/hydrus/help_my_client_will_not_boot.html

custom source environment is simpler

You only really need to care about this if you run from source and use an automated script to rebuild your venv.

About ten weeks ago, I planned to simplify our new pyproject.toml file and remove some old redundant scripts, and the week for that change is here.

If you use the pyproject.toml, please note that it is simpler this week. There are no more dependency groups; just one nice clean template. The future venue to test out alternate library versions is going to be my custom setup_venv.py. The 'running from source - doing it yourself' help is updated to reflect this--basically, any manual setup lines are now simpler since there are no groups to talk about.

Also, several old test scripts and .bat/.sh/.command duplicates are cleared out of the base install dir in favour of the unified and multiplat .py files introduced back in ~v663. The base program directory is much cleaner now.

If you have a script that automatically builds your venv or otherwise is involved here, check the changelog to see if you have any changes.

next week

I had a great week working on a variety of things. Some caching tech is smarter and faster, and some file/folder parsing is also faster, and some help docs are improved.