I'm using the number of milliseconds since midnight as a way to randomize the choice of image to display on the top. But the big question mark has always been how granular that was.
Used to be, back in the days of Win95, it actually changed at 30 hz. So it would increment by 33 (or 34). But is that still true? Today I decided to find out. And I need your help, below the fold.
UPDATE: Chooser version 28, 20121108. Total number of images remains the same at 1772. But the way it chooses the image to display is different, and it should give you a wider variety of pictures. Subjectively it seems to for me, but that's not really scientific.
Also, I used to have a place you could click to get a new image, but it seemed to have a slow memory leak so I took it out. I figured out where the leak was and fixed it. So now, if you mouse-hover the upper left corner, you get the info screen. If you click it, you get a new image.
Apparently the clock on XP runs 60 hz. I expected some version would do that. I wonder whether this was a Vista change, or a Win7 change? Since no one in their right mind runs Vista any longer, I probably can't find out.
I'm not actually certain the clock rate was 30 hz on Win 95; it may have been 60 back then, too.
Posted by: Steven Den Beste at November 08, 2012 11:29 AM (+rSRq)
In my WinXP virtual machine on my Win7 computer, it seems the clock is 100 hz. It changes by 10 milliseconds.
Win7-32 (Saten) also changes by 16, i.e. 60 hz.
Posted by: Steven Den Beste at November 08, 2012 11:33 AM (+rSRq)
4
Mac OS 10.6.8, Safari or Firefox: always 1 1 1 1 1
Same machine, Chrome: most common result by far is 15 followed by a bunch of 10/11, with very occasional variations like "13 11 11 9 8".
-j
Posted by: J Greely at November 08, 2012 12:00 PM (fpXGN)
5
Firefox 16 on 64-bit Intel Linux: 1 1 1 1 1 (consistent across multiple reloads of the page)
Chrome 23.0.1271.64 on the same machine: plenty of variation. I've seen 4 6 6 6 150, 7 7 6 5 16, 5 5 97 6 6, 5 5 5 4 6, and 7 6 5 6 5 in five (re)loads of the page. Overall things seem to cluster around 5 and 6.
Posted by: Chris Siebenmann at November 08, 2012 12:09 PM (YmdZq)
And, yeah, when I saw that up there, I thought you'd be happy.
Posted by: sqa at November 08, 2012 02:25 PM (iWwXY)
13
When I click the image to refresh I get no image just a plain gray field with the Chizumatic title and show title. If I refresh a second time a new image will appear. This is consistent, only get an image every other time. Running 64 bit Win7 with Firefox 16.0.2
Posted by: Doyen at November 08, 2012 02:33 PM (nV9J4)
Doyen, sorry. That appears to be a Firefox strangeness and I have no idea what the problem is or how to fix it.
That function isn't exactly critical, and I don't intend to chase it.
Posted by: Steven Den Beste at November 08, 2012 03:24 PM (+rSRq)
15
SQA, I'm about a third done now. It took a while for the swarm to start working, what with few seeds and many leeches, but it's a lot more reasonable now.
Posted by: Steven Den Beste at November 08, 2012 03:26 PM (+rSRq)
16
BTW, Chrome/Win7-64 I'm seeing that clicking to get a new image also only works every other click. The top right updates to refer to the new image, and the info hover works too, but the picture doesn't actually load. It does NOT do that in IE.
Posted by: RickC at November 08, 2012 05:41 PM (WQ6Vb)
Because it was a hassle for Pixy, I used to accumulate a really big update before making changes to the top rotation. But now that I can upload zip files and use them, I don't have to cause him trouble any longer, and I'm going to go with a lot of smaller changes.
I have 67 pictures from Upotte ready, but that's not enough. So I was crawling through my fansub archive, and noticed something I really should have remembered:
26 possibles through the middle of the 5th episode. As usual, the big problem is finding decent shots of the women without any men, and when it comes to Charlotte that's a problem because she's always hanging onto Teppei. An additional problem is "no nudity". Charlotte likes to wear wispy stuff when she's alone at home, which would otherwise be nearly ideal were it not that we can see through it. Maybe I should download a copy of the (censored) broadcast version to work with instead of the BD rip...
1
This show must not take place in a normal anime universe. Even the smaller ones seem to be respecting gravity...
Posted by: Mikeski at November 03, 2012 05:19 PM (1bPWv)
2
That's not the least of the strange things in this show. In the last few episodes we see
girls getting shot. Holes in their clothes. No blood, but they fall down and stop moving.
We also see the teacher get shot, and he falls down in a pool of blood. And then the pool vanishes, and we learn that if a human get shot by a gun that isn't wielded by a human, it can't harm them.
Which doesn't explain why the teacher had to spend time in the hospital after FNC hosed him... they're not even consistent.
Posted by: Steven Den Beste at November 03, 2012 06:23 PM (+rSRq)
I just uploaded 144 new images from Dog Days 2 for the top rotation and updated "chooser", the flash that displays it all. You will probably need to dump your browser cache in order to load the new version. To be sure you've got it, mouse over the upper left corner. It should say "Version 25" and the total number of images should be 1578.
By the way, the images from Dog Days 2 are in a single zip file, and chooser is accessing them using Pixy's new mechanism, which works great.
I'm inclined not to. For one thing I have a huge number of pictures of Yukki already. For another, it's weird if you know the story behind it. (It approaches squick, frankly, at least for me!)
I've finished 11 episodes and I have 135 shots I'm going to use. Of those, Yukki is in about 40 and there will be more before I'm through. So losing this one won't really hurt all that much.
I'd probably cut it. The outfit just looks rather odd on her/him.
Posted by: sqa at October 16, 2012 12:17 AM (R4ww4)
3
I would think that with the number of images you have, if you have doubts over an image, just say "No" and move on. No need to agonize over it when there are dozens more.
Posted by: Boviate at October 16, 2012 07:19 AM (L1IVj)
4
I'm not agonizing, but it made for a post when I didn't have anything else to say.
Posted by: Steven Den Beste at October 16, 2012 08:46 AM (+rSRq)
I wanted to try an experiment with the new zip-file feature. So I'm going through Dog Days 2 collecting candidates for the top rotation. It's a particularly fertile source; in the first three episodes I got 51 candidates and I'll be able to use almost all of those. Hope you folks don't mind looking at catgirls and doggy girls and squirrel girls.
The idea is that once I'm through, I'll package them all into a zipfile and upload it, and then add an entry to the flash to use them. If it works, it'll mean I can incrementally upgrade instead of accumulating a huge number of pictures to add all at once. (Which is what I was doing because I didn't want to bother Pixy constantly with small uploads.)
I used to trim out older pictures when I added new ones, because I was trying to maintain the total in the range of 700-800. But with this new approach (the flash) I'm not going to do that any more. New pictures will get added but I won't delete any older ones.
UPDATE: Hope you folks don't mind looking at catgirls and doggy girls and squirrel girls...
Posted by: sqa at October 14, 2012 09:53 PM (R4ww4)
2
Might even be a convenient way to add back anything you reluctantly purged.
Posted by: Mauser at October 15, 2012 12:12 AM (cZPoz)
3
What all kinds of girls are there?
There's dog girls, cat girls, wolf girls, squirrel girls, fox girls... What am I missing?
Posted by: tellu541 at October 15, 2012 10:57 AM (G0RNj)
4
IMHO, there are only two kinds of girls that matter: those that find me attractive, and those that do not. In other words, fictional and nonfictional...
Posted by: Siergen at October 15, 2012 11:50 AM (Bv5ty)
Well, goat and sheep girls, in episode 5. And two ape girls are working as heroes.
As to the "cat girls" there's quite a variety there. Jaune is a tiger. Noir is a black panther. Leo is a white lion.
And Couverte is a chipmunk. I'm not sure what kind of squirrels the others are; every squirrel I've seen IRL had gray fur, but all the ones in this show have brown tails.
Of course, Eclair and Millefiore have fur color which doesn't occur in nature, too. And if we think too hard about this we'll depopulate Galette.
Posted by: Steven Den Beste at October 15, 2012 12:12 PM (+rSRq)
And I should mention that I wouldn't be at all surprised to learn that Brioche is a wolf.
Posted by: Steven Den Beste at October 15, 2012 12:14 PM (+rSRq)
7
Also Vert is a bunnygirl. Galette seems to be a pretty varied menagerie.
Posted by: Steven Den Beste at October 15, 2012 12:19 PM (+rSRq)
8
Brioche once said (in S1): "but I'm a human". Yeah, right. Lemme look closer at those ears... Yes, I understand that she meant the contrast with Yukki, who is not even a biological entity.
Siberian chipmunk (burunduk) has brown tail. It may live in Hokkaido and Honshyu, too.
Posted by: Pete Zaitcev at October 15, 2012 03:11 PM (RqRa5)
Posted by: Steven Den Beste at October 13, 2012 04:48 PM (+rSRq)
3
I use 7Zip as well. It's fast, reliable, and supports all the main formats. And it's open source.
Posted by: Pixy Misa at October 13, 2012 07:07 PM (PiXy!)
4
Whee! It might change my workflow a bit, but that would probably be an improvement.
It does mean changing out individual images might be trickier.
The one other thing that would be very helpful is a visual preview before you make a post live.
Posted by: Mauser at October 13, 2012 07:31 PM (cZPoz)
5
Definitely has a performance advantage against the other programs - takes advantage of multiple cores to work faster. We zip a LOT of stuff at work (right now I'm zipping, even) and 7zip regularly gets those multi-GB jobs done much faster.
Though it's also been the case that, every so often, we'll get a zip that 7zip just can't open up, and we'll need to drop back to WinRAR...
Posted by: Avatar_exADV at October 13, 2012 07:35 PM (pWQz4)
It does mean changing out individual images might be trickier.
Not really. You swap the new image into the zip file, and then upload it again. It's slightly more complicated, but not really.
Posted by: Steven Den Beste at October 13, 2012 07:36 PM (+rSRq)
7
The next release will make zip files act like virtual folders, so instead of /images/zip_file_name.zip:image_file_name.jpg, you can just have /images/zip_file_name/image_file_name.jpg. Because the handling of file paths in Minx is kind of funky, it was easier to put in the zipfile:imagefile syntax for now. I'll have a little converter script in place when the time comes to cut over.
Posted by: Pixy Misa at October 14, 2012 12:03 AM (PiXy!)
This should work when the pictures are invoked out of a flash, too, right? Which means I should be able to use this in the top rotation, so that I don't have to bug you any more to unpack zip files for me?
Posted by: Steven Den Beste at October 14, 2012 09:52 AM (+rSRq)
It makes me want to watch the movie, which should arrive here week from this coming Friday. Even without a translation I think it'll be good to watch. With a higher budget and a longer production time I expect the animation to be excellent.
Posted by: Steven Den Beste at October 14, 2012 02:41 PM (+rSRq)
That's the basic comment entry box. First thing we do is type part or all of our comment. (It isn't necessary to write it all; links can be added at any time.)
Then we highlight the text we want to make into a link.
We click the magic "make a link" button:
It looks like chain links, get it? In response we get a popup:
We enter our URL in the appropriate box:
And we hit "Insert".
The box automagically changes some things around. Don't let that bother you; it doesn't matter. Our last step is to hit "OK".
And our link is now complete.
And now you guys have no excuses for putting ugly URLs directly inline. Please don't do it from now on.
It would be nice if the box made both kinds of links look the same.
When you enter a raw URL, how would it know what the link text should be? Making it prompt for it would be a headache -- especially if there were several raw URLs being fixed.
Posted by: Steven Den Beste at September 07, 2012 06:08 PM (+rSRq)
8
"When you enter a raw URL, how would it know what the link text should be?"
Well, BBCode and the like have been around for ages. It's not a big deal to me, I'll try to do it the right way. The editor is quirky for me, though.
Posted by: RickC at September 07, 2012 08:17 PM (WQ6Vb)
Update to "chooser"
I had an idea today for a way to make the top rotation chooser do a better job of mixing up the pictures, and tweaked the code just a hair. The latest version is 24, and to get it you'll probably have to dump your browser cache. (You can mouse-hover over the upper left corner of the top rotation to find out which version you're running.)