Danbooru 2 upgrade on Feb 16, 2013

ghostrigger said:
*messed up old post; deleted*

EDIT: artists.xml seemed to be responding, but how to you access the artist's url fields which seemed to be missing?

EDIT2: since github is accessible again, i was reminded the correct way of accessing notes and comments were notes.xml?group_by=note and comments.xml?group_by=comment respectively. however, how to access the wiki and correct way of accessing the artist api parameters (urls)?

Artist URLs should be fixed. By the way, all API endpoints also return JSON results (artists.json in this case) which I think is much easier to parse.

The wiki pages are accessible at /wiki_pages.json.

The Danbooru downloader should still work.

I've fixed the complaints about the paginator. The last page is now linked provided it's less than 1,000 pages.

Could you populate the help:API page the Site Map links to?

I have to admit, I have not had time to pay much attention to Danbooru 2 before now, but looking at it now it seems pretty straightforward. I rather like the tag list calling out art/copy/char tags before the general tags too.

Question about the batch alias/implication import. Before, it would redirect me to the jobs page and show me what I added, but this time there's no redirect. I went to the jobs page manually and the aliases I submitted weren't there, and aren't in the aliases list either. What's the best way for me to quickly tell if aliases or implication I've submitted went through?

If the wiki is going to be an alternate to the posts instead of tags shouldn't the whole article be displayed?

jxh2154 said:
Question about the batch alias/implication import. Before, it would redirect me to the jobs page and show me what I added, but this time there's no redirect. I went to the jobs page manually and the aliases I submitted weren't there, and aren't in the aliases list either. What's the best way for me to quickly tell if aliases or implication I've submitted went through?

Yeah, sorry that's not clear. Your aliases were created and they do create jobs, but the queue is much faster now and it will start processing almost instantly. I guess the best way is to just search for the alias or implication. If the job errored out it will say so in the job listing.

Well I tried deploying and it's clear there are performance issues. I'll have to spend some time profiling it I guess. In the mean time I've reverted to the old code base and you can upload again.

albert said: Yeah, sorry that's not clear. Your aliases were created and they do create jobs, but the queue is much faster now and it will start processing almost instantly. I guess the best way is to just search for the alias or implication. If the job errored out it will say so in the job listing.

One of the main reasons I like to get the feedback on the next page is because one major issue I run into is the existence of conflicting aliases. so for example, A → B exists, and I don't know that, and try aliasing A → C. In the current/old codebase, it doesn't tell me that alias failed (due to a conflict existing) but I can sometimes tell by eyeballing the job task page that comes up and seeing, "Oh hey one is missing". Though I still miss it sometimes. Manually searching for each alias after submitting to confirm could also get to be a problem when I'm doing a run through the thread backlog and perhaps submitting 50+ total.

Well, I'll play around with it a bit more when the upgrade is done.

Any estimate on the next release date?

Performance profiling is looking good. I fixed a bunch of degenerate cases where a page was taking 4+ seconds to load.

I'm running a version of it on this server at http://danbooru2.donmai.us (pictures don't work because they're stored in different paths). There's performance monitoring available at http://danbooru2.donmai.us/newrelic

Most pages should ideally take less than 500ms.

Comments : Listing is a fairly slow to load.

Yep, comments:listing is particularly slow to load, and comments:search results in "something broke".
I don't like 'group_by' argument in the link. If SQL query really uses full select + group by, then, AFAIK, MYSQL can play some nasty tricks, like not using the index.

Type-kun said:
I don't like 'group_by' argument in the link. If SQL query really uses full select + group by, then, AFAIK, MYSQL can play some nasty tricks, like not using the index

Danbooru uses PostgreSQL, not MySQL. And the site's source code is available on Github if anyone wants to examine it and/or submit patches.

Bapabooiee said:
Danbooru uses PostgreSQL, not MySQL. And the site's source code is available on Github if anyone wants to examine it and/or submit patches.

Oh, never knew it, thanks for pointing that out.

I'm gonna try upgrading again this weekend and see how it pans out. Once again the site will go into read only mode this Saturday and will probably last until Sunday evening.

albert said:
I'm gonna try upgrading again this weekend and see how it pans out. Once again the site will go into read only mode this Saturday and will probably last until Sunday evening.

Hope it goes well, been looking forward to it for awhile now.

Well, so far the interface is not nearly as user friendly.

Well, so far the interface is not nearly as user friendly.

Which part specifically?

There are some clunky parts with dup detection when uploading that I'll get around to fixing.

Is it just me or did all the RSS feeds get messed up? In Google Reader I'm seeing a whole bunch of feeds with 20 pics, none of them with the tag for that feed and apparently almost all the same 20 pics.

Can't access the mod queue.

Also everything relating to the forums is ugly as hell.

One minute ago ago.

For some reason I keep getting a "An error occurred: error: ArgumentError - expected a file to upload". I've uploaded jpgs just fine but the file I'm getting this error on is a png. Tried uploading from pixiv and directly from my harddrive.