* Replace the .category-N CSS classes on tags with .tag-type-N. Before
we were inconsistent about whether tag colors were indicated with
.category-N or .tag-type-N. Now it's always .tag-type-N.
* Fix various places to not use Tag.category_for. Tag.category_for does
one Redis call per tag lookup, which leads to N Redis calls on many
pages. This was inefficient because usually we either already had the
tags from the database, or we could fetch them easily.
Change the title of the post index page to look like this:
"Danbooru: Anime Image Board" (for the front page)
"Kantai Collection Art | Danbooru" (for a tag search)
Change the meta description of the front page to look like this:
Danbooru is the original anime image 'booru. Find over 3.75 million
anime pictures categorized by over 100 million tags.
Change the meta description for a tag search to look like this:
Find over 37,168 Azur Lane images on Danbooru. Azur Lane (碧蓝航线)
(벽람항로) is a Chinese shipgirl-themed side-scrolling shoot 'em up
mobile game developed by Shanghai Manjuu and Xiamen Yongshi...
* Rename 'privacy mode' to 'private favorites'.
* Make the private favorites setting only hide favorites, not favgroups
and not the user's uploads on their profile page.
* Make the favgroup is_public flag default to true instead of false and
fix existing favgroups to be public if the user didn't have privacy mode
enabled before.
* List _all_ public favgroups on the /favorite_groups index, not just
favgroups belonging to the current user.
* Add a /users/<id>/favorite_groups endpoint.
This was an alternate frontpage that contained a list of previews of the
most popular tags. This page was never linked from anywhere and it was
unknown by most users.
If Post.fast_count returns nil, then the link_to call becomes
link_to(nil, <posts_path>) so the link text becomes the url itself,
which isn't what we want.
Instead the count is displayed as a '?' mark to indicate that we don't
know the true count.
* Add favorite count beneath recommended posts. Clicking the favcount
loads more recommended posts like that post.
* Increase number of recommendations shown on post show page.
* Pick the largest character or copyright tags by post count. Previously
we picked the tags with the longest names, which was nonsensical.
* Remove tag cateogory logic from config file. We can't avoid hardcoding
some knowledge about tag categories here, so there's no point in trying.
This affects tab titles on post show pages as well as filenames in
downloaded images.
Normally thumbnails have a fixed size of 154x154, but that's not always
desirable outside of the posts index because it creates empty gaps
around thumbnails.
* Add the source (twitter, pixiv, etc) and upload date ("X minutes ago")
to iqdb thumbnails.
* Link the filesize to the full file so you can compare files in new tabs.
* Link the similarity to a iqdb search so you can pivot your search to other posts.
This was used to discourage crawlers from crawling certain pages we
didn't want them to crawl, primarily post searches.
Remove because there are better ways to control crawling. Some of these
links weren't even visible to crawlers anyway. This lets us be
consistent about only applying rel="nofollow" to external links.
* Use underscores instead of spaces for tags in inline tag lists (upload
tags report, tooltips, modqueue, comments page).
* Allow long tags to word wrap. Fixes long sources not wrapping in the
uploads tag report. Also fixes very long tags that don't have
underscores not wrapping in the sidebar (ex: kuouzumiaiginsusutakeizumonokamimeichoujin_mika).
Set the thumbnail height to auto instead of 154px so that there's not a
big empty gap beneath thumbnails that aren't 150px high.
This requires setting dimensions on the <img> tag itself so that the
image height is known before loading, otherwise the tooltip will be
mispositioned after the image loads in.
We set the min-width and min-height instead of the <img> width and
height properties because our calculated dimensions are sometimes
off-by-one compared to the actual dimensions. I'm not sure how libvips
calculates the thumbnail dimensions, but this avoids forcing a slightly
wrong aspect ratio, which produces ugly resizing artifacts.
Make /favorites redirect to a ordfav:<user> search instead of having a
separate view just for favorites. This duplicated a lot of code for no
good reason.
Rewrite the implementation of related tags to be simpler, faster, and
more accurate:
* The related tags are now calculated by taking a random sample of 1000
posts, finding the top 250 most frequent tags among those posts, then
ordering those tags by cosine similarity.
* Related tags can generally be calculated in 50-300ms at these sample
sizes. Very high sample sizes (25000+ posts) are still relatively fast
(1-3 seconds), but generally they don't improve accuracy much.
* Related tags are now cached in redis rather than in the tags table.
The related_tags column in the tags table is no longer used.
* Only the related tags in the search taglist are cached. The related
tags returned by the 'Related tags' button are not cached.
* The cache lifetime is a fixed 4 hours.
* The 'Related tags' button now works with metatags.
* The /related_tag page now works with metatags and multitag searches.
Fixes#4134, #4146.
Changes:
* Drop Users.id_to_name.
* Don't cache Users.name_to_id.
* Replace calls to name_to_id with find_by_name when possible.
* Don't autodefine creator_name in belongs_to_creator.
* Don't autodefine updater_name in belongs_to_updater.
* Instead manually define creator_name / updater_name only on models that need
to return these fields in the api.
id_to_name was cached to reduce the impact of N+1 query patterns in
certain places, especially in api responses that return creator_name /
updater_name fields. But it still meant we were doing N calls to
memcache. Using `includes` to prefetch users avoids this N+1 pattern.
name_to_id had no need be cached, it was never used in any performance-
sensitive contexts.
Avoiding caching also avoids the need to keep these caches consistent.
In multi-tag searches, we calculated the tags in the sidebar by sampling
300 random posts from within the search and finding the most frequently
used tags. This meant we were effectively doing two searches on every
page load, one for the actual search and one for the sidebar. This is
not so bad for fast searches, but very bad for slow searches.
Instead, now we calculate the related tags from the current page of
results. This is much faster, at the cost of slightly lower accuracy and
the tag list changing slightly as you browse between pages.
We could use caching here, which would help when browsing between pages,
but we would still have to calculate the tags on the first page load,
which can be very slow in the worst case.
