mpv added HDR support for Wayland. https://www.phoronix.com/news/MPV-0.40-Released

Hi! I'm having this rare but still intermittent issue where the api just stops responding. The issue persists until restart. Windows platform. Usually it happens after I try to access videos via the api in quick succession. How would I even go about finding the reason for this? The "api test" option in the debug menu has no effect until the restart as well.

>>17484 Ah, I thought I had broken something myself, glad I checked the thread.

For duplicate auto-resolution rules, I hope to see fuzzy filesize comparison (A's filesize is within x% of B's filesize), comparison of image resolution and comparison of aspect ratios. For JPEGs, also quality/compression profile. I think having these types of rules available would make a lot of work with variant sets automatable, and even be able to reliably detect obvious better/worse pairs.

>>17488 >removes "creator:sound warning" because it's a trash tag and changes "creator:third-party edit" to "meta:third-party edit" via regex substitution. Why would you do that? If I wanted those changes I could do them non-destructively with display filters and siblings.

>>17501 Because I cannot imagine a single situation where having "creator:sound warning" as a tag would be useful. In fact, the tag is far more likely to be wrong than the built-in "has audio" file property.

>>17502 I don't care about your lack of imagination.

>>17502 it's useful if you don't want your ears blasted by a sudden loud noise. what's hard to get about that?

>>17503 Then shut the fuck up and remove it from the parser filter for yourself. Here, I'll even give the exact path: Network -> downloader components -> manage parsers... -> "e621 file page parser" -> Edit -> content parsers -> creator tags -> Edit formula -> Big string processor button -> select "TAG FILTER: ..." -> X - > Yes -> Apply everything. >>17504 Then use the "system:has audio" predicate? Which does a better job? Out of 394 files with "creator:sound warning" in tags only downloaded from e621, 4 of them don't actually have any audio at all, and out of the remaining 527 files with audio, there's many that could be considered having "obnoxiously loud sound". I'd take it as a good, valuable tag if it were mechanically applied, like if the e621 website detected any sound past 30dB (arbitrary) on upload and added the tag (which shouldn't be under 'artist' anyway) but it's clearly not. To be fair, on the same logic I should also filter out the other various file tags like "meta:sound" or "meta:webm" because they're all useless, I actually forgot they even existed.

>>17506 If you're posting something on the internet for people to use, try to be user-friendly instead of enforcing your own arbitrary biases. Imagine if I posted a e621 parser that got rid of the species namespace and just moved them to unnamespaced, and then said "hurr durr i cannot imagine a situation where you would want the species namespace". You can see how some people might be annoyed by that, right? Especially in the hydrus thread were people are autistic about tags. I just don't see any reason to arbitrarily change the results to be different from what's actually on the page.

>>17504 If that's what sound warning was actually used for I'd agree.

>>17488 I'm dumb, explain this. To fix e621 tags not working I need to remove creator sound warning? I checked the catbox as well but it's just an image...

>>17504 >it's useful Sure, but only for you. Imagine if much of these moot feature requests are fulfilled, the software would be so bloated and slow that anons would be demanding for a Hydrus Light version. What about if you try the Mute icon before you load the videos instead?

>>17511 >Imagine if much of these moot feature requests are fulfilled what feature request? the reason they have the tag on e621 is to let people know that the video has a sudden audio spike. nothing about hydrus here so I don't know what you're responding to >What about if you try the Mute icon before you load the videos instead? because that just... mutes the audio?

>>17511 >there shouldn't be a jumpscare tag in any context >just like, don't play the video

>>17510 The image has the downloader data inside of it. Download the picture and put it somewhere you can find. Go to hydrus>network>downloaders>import downloaders. Click on Lain's face and find the image, it will tell you what you are importing, click yes and now the parser's been added. Now go to network>downloader components>url class links, find e621 file page, edit and set it to the new one, which should be e621 file page parser (1), matching parsers should be at the top.

In case anyone used the e621 with notes variant I updated it to work with the new parsing thing. I didn't change the sound warning stuff because I didn't feel like it. No image uploads so here it is: https://files.catbox.moe/t6jw3w.png

>>17512 >because that just... mutes the audio? Exactly, then you adjust the volume located right on top of the Mute icon just before to click on a video. Not a big deal, I think.

>>17514 >>17522 >>17488 Thanks

any ways to make some images children/parents of others? i'm seeing the children/parent tag thing but it doesn't seem to correlate

>>17526 No, not yet, you can set them as alternates*, which just indicates they are related, it does not have any sort of hierarchical info. I think parent/child images might be planned but I might be making that up. *select 2 or more images>manage>file relationships>set as alternates.

>>17447 Great idea, thanks. I'd still like to completely rewrite how export filenames are created. We want something richer than the current template string, and we need better fallback logic for when some data is missing. I'll write this down and think about it. >>17448 That's an interesting idea, thanks. I'll write it down. >>17458 We do it for cbz now, although my cbz detection algorithm is a bit shaky, so if your cbzs are read as zip, let me know. I think it would be neat for normal archives to pick a decent thumb, although I know some users don't want it because many results will be unhelpful (imagine if you were storing lots of technical brush zips and the thumb every time was just a 20x20 white square or other weirdness), so I'd introduce it as optional for arbitrary archives, I think. We have zip reading right now (python comes with it), but I don't think we have any tech to read rars or 7z yet. I think it is fairly easy to add, but I haven't looked into it properly yet. I'd like thumbs for cbr and is it called 'cb7'? I generally want more archive-parsing tech in future. You mention a .thumb file--is that something you see often? Is that a standard in a particular .zip format you use? Are there any others I could look for? My general approach here has been to list all the .jpgs and .pngs and just pick the one with the earliest filename, since that works for cbz. As I (slowly) roll this stuff out, let me know how it works for you! >>17459 Thanks, I am glad you like it. If you are a new user, then, in a few weeks, once you are comfortable with things, let me know what has been difficult and easy to learn. Keeping the help guides updated and overall UI easy for new users to pick up is a constant battle. >>17462 Thanks. I am afraid 'rating' file sort does not mix with system:limit yet. I added a tooltip to the file sort menu button a couple weeks ago so it actually tells you now which do and don't work. That said, rating shouldn't be too difficult to add, so I'll look into it.

>>17467 I would like to help you but I'm afraid I don't work on the Docker stuff, and I just have no Docker experience myself. Suika is on the discord if you want to talk to him directly. >>17468 Thanks--sorry for the trouble! I wish these guys made it more obvious which behaviour they don't like, but I guess they keep those cards close to their chest precisely because they don't want the bad guys knowing. For a fun story, we once got an anti-virus false positive because the scanner sandbox ran the hydrus installer exe in some automatic mode, and we used to have a 'open hydrus getting started guide' checkbox on the last page. The sandbox clicked this somehow, even I think when we set it to default false, and that opened Internet Explorer, which triggered a Windows Update network call on the sandbox, and that triggered the false positive since the installer was doing weird network stuff, hooray. >>17480 Sorry for the frustration! I'll put some time into the UI and help to make this less obtuse. >>17484 >>17488 I've got a job to fix this in the defaults as well, btw. Looks like they changed their HTML recently. Either me or another user is this week going to look at an API solution since it turns out their API provides tag namespaces. I'll figure out new downloaders for e926 and e6ai too. >>17489 I am no Linux expert, but it seems recent Wayland changes don't play well with hydrus. There's some help here, under the Linux tab: https://hydrusnetwork.github.io/hydrus/getting_started_installing.html#installing Adding "QT_QPA_PLATFORM=xcb" env variable seems to be the simple solution; it causes hydrus to launch with X11 instead. Running from source (which I think the flathub does, I don't know?) generally improves compatibility too. It is easy to set up yourself now: https://hydrusnetwork.github.io/hydrus/running_from_source.html

>>17492 Thank you for your feedback. You have a lot of dupes, so you have a special perspective. I've had similar thoughts from some other users today, basically that they'd like this system to be a touch less automatic and fire-and-forget, particularly as we push from easy pixel duplicates to more fuzzy situations. The preview panel is good, but it is too tucked away and not obvious (as you found too). I am going to think about this seriously this week. Maybe the rules get new pause modes or something where they do their calculations but for the final action and then you can review that. Or perhaps I boost the preview panel to be more user-friendly and do a better job of showing upcoming outcomes. This is tricky, but I think I put real time into it. I built this system to be completely automatic, so my preference is to KISS and keep it that way, but several users want more control, so I'm thinking about what I can do without breaking the bank. Another thing is firming up our tools to make sure these fuzzy choices are less uncertain. If I add some tech that does 'A is 99.7% similar to B', to arbitrary precision, then we will be able to filter out slight artist corrections as 'exact match - distance 0' is currently unable. Another option is to rewrite my perceptual hash system to use a longer hash so we can go finer than the current precision on the search end. An alternative is plugging the new 'comparator' tech the auto-resolution system has into the normal duplicate filter. I could provide the queue of what a rule would work into the duplicate filter and arrange the A-B in the same way, and then you'd be processing it like an archive/delete filter maybe. I don't know, I'm thinking about my options now. >would it be possible to do something like that will the current duplicate filter? maybe as a companion to the large full image view? Yeah I think it could. I thought making that 'thumbnails-in-a-panel' part would be the most difficult of the whole system, which is why I left it basically for last, but in the end it wasn't too bad. I feel better about having a sort of carousel for the media viewer in general now; and the duplicate filter could definitely have a carousel of two. >>17497 Thanks, interesting. Presumably this 0.40 will percolate down to libmpv in several versions of Linux. I'm expecting to do another .dll test on the Windows side soon as well. We'll see what it fixes and breaks. >>17498 Thank you for this report. Please tell the Client API in manage services to 'log requests', and turn on help->debug->profiling->profile mode. Then see if you can trigger the fault, and we'll see if anything stands out. I wouldn't pastebin your whole log here, and the profile will be huge, but if you email me or hit me up on discord, we can figure out a secury transfer, or you can look yourself and just cut and paste the last stuff that happens before it breaks. I think the log will say how long each Client API job takes. Maybe there will be some errors about saturated connections or something in the log. (view your log and new profile by hitting file->open->database directory)

>>17500 Thanks--I would like to roll out exactly this over the nearish future. I'm increasingly thinking I'm going to need a fine 'A is 99.7% pixel-similar to B' so we can exclude colour changes and tiny artist changes and stuff that slip through 'exact match - distance 0' search, but we'll see how it all shakes out. Please let me know how it goes for you as I roll it out. >>17526 >>17527 Yeah, set them as alternate for now. That's just a safe landing zone to hold those pairs for now. In the future when we have a better grapple on duplicates, I'll be writing a large expansion of the file relationships system to let us define non-duplicate file relationships like 'WIP' and "costume alternate' and 'messy/clean' and so on.

the sort by color features are so fun!

>>17536 It's pretty neat, but I think sort by color balance is slightly flawed, as it ends up with a ton of black and white images in the middle, while sort by hue places images with low chromatic magnitude at the end, outside the rainbow. Since you can also sort by chromatic magnitude, which places mostly black and white images at one end, I think it would be feasible to filter most of these out of color balance like they are for hue sorting. In fact, I think it might not be too hard to implement a system:color (assuming such a function can't be folded into a different system predicate) that allows you to search above or below certain color balance, hue, chromatic magnitude, and lightness thresholds. The system already fetches all the information necessary to perform such a search in order to perform these sorts, so it should be mostly just be making another UI window if it were to be added. I don't think it'd be particularly useful outside of fetching images to make those color collage images like pic related, but it would be neat.

>>17538 Also, it's personally not much of an issue to me, as I already tag images as "black and white", which lets me exclude them from color sorts. The chromatic magnitude option really helps find stray black and white images I've missed though.

>>17530 in terms of the fuzzyness, like you mentioned 99.7, would it be possible to have a "most likely compression noise" or when there is an areas where the details are quite a bit more firm "most likely same set with slight difference" I think that those two would be the... well I don't want to say easy, but probably the easiest to make a filter set know as for the window with thumbs, it would be really nice to have on the side a few buttons [a is better] [b is better] [related alturnaties] [false positive] we can give a quick scrub through of the thumbnails that come up and pick out the obvious, and what's not obvious, goes to full on opening the images and doing them one at a time while a/b ing them hell, even the 'is better' isn't necessarily true, its just that im keeping one of them and not the other. for me where things are going to get fuzzy is png's and jpeg's where I take the pngs out of hydrus, downscale/reencode to jpeg and re import, I could honestly see my biggest auto resolve being png vs jpeg where they are 99.5% the same but the jpeg is half the size or less. I have already gone though around 200gb of files like this that I got rid of, not to mention the ones I kept to be processed later on. I would find auto'ing this to be immensely helpful if they are held for review, I personally see that as the main thing I would change/add, automatically process but don't automatically delete/move to trash. on a side note, I think it would be nice if you were able to open a window of all the resolved pairs, this is more of a me problem, but I had a number of files already in trash, this dumped 19.5k files into the trash along with what was already there. it made looking though it a bit harder than it needed to be, though you can see by scope im probably a bit of an outlier in that regard.

>>17541 >on a side note, I think it would be nice if you were able to open a window of all the resolved pairs, this is more of a me problem, but I had a number of files already in trash, this dumped 19.5k files into the trash along with what was already there. it made looking though it a bit harder than it needed to be, though you can see by scope im probably a bit of an outlier in that regard. Not sure if im understanding correctly, but i think it is the case that if you open the trash domain and then 'sort by time:import time', you will get the files sorted by the time they got actually imported into trash rather than imported into hydrus, which would be useful for you wouldn't it? So everything already there would be at the beginning or end of the thumbnail grid depending if you chose newest/oldest first. With the system predicate system:time you can search then for stuff older/younger than a day or stuff like that.

Is there a way to change the design of this board back to what we had recently?

>>17543 The dropdown at the top.

>>17544 Thanks! In another browser i can see it, not in my old not updated one though :p

I have no idea what this tag data is and would rather clear than upload it. Forget however does nothing A search for "exclude current tags" + "has tags" on the PTR gives no results. Any idea?

I had a great week. Last week's duplicates auto-resolution test suggested we needed some new tech for semi-automatic behaviour and an audit log, and I managed to get it done for a second test. I also have some misc bug fixes and a new e621 downloader that works a lot better and fixes the recent broken tags. The release should be as normal tomorrow. >>17548 Did you earlier do a 'commit' of your normal pending mappings, and this was 'left over'? If so, this could be a miscount--can you try database->regenerate->total pending count, in the pending menu? If this is not a miscount, it could be pending siblings, parents, or deleted (petitioned) mappings. Check the tags->siblings/parents dialogs to see if there are any there, although I presume you'd know if you pended a couple thousand siblings. Petitioned mappings are difficult to search for arbitrarily, but if you already flushed your normal pending mappings, the petitioned should have gone at the same time (same for siblings and parents--I think this is a miscount).

>>17549 Thank you, the regenerate did it