Add a new color palette and rework all site colors (both light mode and dark mode) to
use the new palette.
This ensures that colors are used consistently, from a carefully designed color palette,
instead of being chosen at random.
Before, colors in light mode were chosen on an ad-hoc basis, which resulted in a lot of
random colors and inconsistent design.
The new palette has 7 hues: red, orange, yellow, green, blue, azure (a lighter blue), and
purple. There's also a greyscale. Each hue has 10 shades of brightness, which (including
grey) gives us 80 total colors.
Colors are named like this:
var(--red-0); /* very light red */
var(--red-2); /* light red */
var(--red-5); /* medium red */
var(--red-7); /* dark red */
var(--red-9); /* very dark red */
var(--green-7); /* dark green */
var(--blue-5); /* medium blue */
var(--purple-3); /* light purple */
/* etc */
The color palette is designed to meet the following criteria:
* To have close equivalents to the main colors used in the old color scheme,
especially tag colors, so that changes to major colors are minimized.
* To produce a set of colors that can be used as as main text colors, as background
colors, and as accent colors, both in light mode and dark mode.
* To ensure that colors at the same brightness level have the same perceived brightness.
Green-4, blue-4, red-4, purple-4, etc should all have the same brightness and contrast
ratios. This way colors look balanced. This is actually a difficult problem, because human
color perception is non-linear, so you can't just scale brightness values linearly.
There's a color palette test page at https://danbooru.donmai/static/colors
Notable changes to colors in light mode:
* Username colors are the same as tag colors.
* Copyright tags are a deeper purple.
* Builders are a deeper purple (fixes#4626).
* Moderators are green.
* Gold users are orange.
* Parent borders are a darker green.
* Child borders are a darker orange.
* Unsaved notes have a thicker red border.
* Selected notes have a thicker blue (not green) border.
Bug: In Postgres 13, getting the count of a blank search underestimated
the page count by a large margin (~700,000 posts).
The query we were executing was this:
EXPLAIN (FORMAT JSON) SELECT * FROM posts ORDER BY id DESC
The `ORDER BY id DESC` clause triggered a parallel seq scan query plan
in Postgres 13, which for some reason causes Postgres to underestimate
the row count by large amount in each parallel branch.
Getting rid of the ORDER BY clause makes it do a regular seq scan, which
gives an accurate estimate.
* Refactor the paginator into a ViewComponent.
* Fix inconsistent spacing between paginator items.
* Fix a bug where the sequential paginator generated the wrong next /
previous page links in the <link rel="{next|prev}"> tags in the <head>.
* Always include the final page as a hidden html element, so that it can
be unhidden with custom CSS.
* Make it easier to change the pagination window.
Adjust input boxes to fit the screen better on mobile, and to use
consistent sizes throughout the site (no ad-hoc overrides to make
certain input boxes a bit bigger in random places).
On desktop, this makes most input boxes a bit taller and narrower. On
mobile, it makes input boxes the full width of the screen.
This most notably affects the tag edit box, the comment and forum
post box, the wiki page edit box, and the commentary box.
Refactor CSS to use standard Tailwind-style utility classes instead of
ad-hoc rules. This eliminates a lot of single-purpose rules for specific
UI elements and standardizes margins to be more consistent throughout
the site.
Utility classes are defined manually on an as-needed basis instead of
importing Tailwind as a whole. Naming conventions mostly follow
Tailwind's conventions, otherwise they follow Bootstrap.
* https://tailwindcss.com/docs/
* https://getbootstrap.com/docs/5.0/utilities/spacing/
Require the user to re-enter their password before they can view,
create, update, or delete their API keys.
This works by tracking the timestamp of the user's last password
re-entry in a `last_authenticated_at` session cookie, and redirecting
the user to a password confirmation page if they haven't re-entered
their password in the last hour.
This is modeled after Github's Sudo mode.
Track when an API key was last used, which IP address last used it, and
how many times it's been used overall.
This is so you can tell when an API key was last used, so you know if
the key is safe to delete, and so you can tell if an unrecognized IP has
used your key.
Add the ability to restrict API keys so that they can only be used with
certain IP addresses or certain API endpoints.
Restricting your key is useful to limit damage in case it gets leaked or
stolen. For example, if your key is on a remote server and it gets
hacked, or if you accidentally check-in your key to Github.
Restricting your key's API permissions is useful if a third-party app or
script wants your key, but you don't want to give full access to your
account.
If you're an app or userscript developer, and your app needs an API key
from the user, you should only request a key with the minimum
permissions needed by your app.
If you have a privileged account, and you have scripts running under
your account, you are highly encouraged to restrict your key to limit
damage in case your key gets leaked or stolen.
Regression caused by the switch from the mobile API to the Ajax API. In
the Ajax API, commentaries have /jump.php?<url> links that we have to strip out.
* fix the test for R18 posts, which are now supported
* replace the dead post for the imageMap test with an active one
* fix the ordering of images: it was possible to get the wrong image in
some cases when not using the batch bookmarklet, because of how fanbox
orders imageMap posts
* fix a bug where the sample wasn't properly replaced with the full
size
* rename test file to match the rest of the other files
Fix the Pixiv API no longer working by rewriting the Pixiv strategy to
use the Ajax API instead of the mobile API.
Before we could authenticate in the mobile API by using the OAuth 2.0
grant_type=password authentication flow. This no longer works. Now it
requires logging in through a HTML page, which is protected by Google
reCaptcha. This makes using the mobile API infeasible.
Instead we switch to the Ajax API, which only needs a PHPSESSID to
authenticate. This can be obtained by logging in manually and using the
devtools to extract the cookie.
This also temporarily removes support for Pixiv novels. This should be
moved to a separate source strategy.
Pixiv API client is currently broken. Temporarily disable Pixiv source
strategy so direct image Pixiv uploads can still go through. The
posts.pixiv_id field will need to be backfilled later.
The artist name is supposed to be the display name according to the
base file, however the artist name was treated like the tag name
instead. This commit renames all instances of "artist_name" to
"tag_name" and then adds an "artist_name" function that uses the
Twitter display name if available.
This refactors Pundit policies to only rely on the current user, not on
the current user and the current HTTP request. In retrospect, it was a
bad idea to include the current request in the Pundit context. It bleeds
out everywhere and there are many contexts (in tests and models) where
we only have the current user, not the current request. The previous
commit got rid of the only two places where we used it.
Refactor the post preview html to use the ViewComponent framework. This
lets us encapsulate all the HTML, CSS, and helper methods for a UI
component in a single place.
See https://viewcomponent.org.
When a user does a tag search, log a few more things, including the normalized
search string, the number of tags in the search string, and the number of results.
Refactor page limits to a) be explicitly listed in the User class (not
hidden away in the Danbooru config) and b) explicitly depend on the
CurrentUser (not implicitly by way of Danbooru.config.max_numbered_pages).
Optimize autocomplete to ignore various types of bogus input that will
never match anything. It turns out it's not uncommon for people to do
things like paste random URLs into autocomplete, or hold down keys, or
enter long strings of gibberish text (sometimes in other languages).
Some things, like autocorrect and slash abbreviations, become
pathologically slow when fed certain types of bad input.
Autocomplete will abort and return nothing in the following situations:
* Searching for URLs (tags that start with http:// or https://).
* Overly long tags (strings longer than the 170 char tag name limit).
* Slash abbreviations longer than 10 chars (e.g. typing `/qwoijqoiqogirqewgoi`).
* Slash abbreviations that aren't alphanumeric (e.g. typing `/////////`).
* Autocorrect input that contains too much punctuation and not enough actual letters.
Optimize searches for non-English phrases in autocomplete. These
searches were pretty slow, and could sometimes cause sitewide lag spikes
when users typed long strings of non-English text into the search box
and caused an unintentional DoS.
The trick is to use an `array_to_tsvector(other_names) USING gin` index
on other_names. This supports fast string prefix matching against all
elements of the array. The downside is that it doesn't allow infix or
suffix matches, so we can't support wildcards in general. Wildcards
didn't quite work anyway, since artist and wiki other names can contain
literal '*' characters.
Fix the `normalize` and `array_attribute` macros conflicting with each
other on the WikiPage model. This meant code like
`wiki_page.other_names = "foo bar"` didn't work. Both macros defined a
`other_names=` method, but one method overrode the other.
The fix is to use anonymous modules and prepend so we can chain method
calls with super.
Add tracking of certain important user actions. These events include:
* Logins
* Logouts
* Failed login attempts
* Account creations
* Account deletions
* Password reset requests
* Password changes
* Email address changes
This is similar to the mod actions log, except for account activity
related to a single user.
The information tracked includes the user, the event type (login,
logout, etc), the timestamp, the user's IP address, IP geolocation
information, the user's browser user agent, and the user's session ID
from their session cookie. This information is visible to mods only.
This is done with three models. The UserEvent model tracks the event
type (login, logout, password change, etc) and the user. The UserEvent
is tied to a UserSession, which contains the user's IP address and
browser metadata. Finally, the IpGeolocation model contains the
geolocation information for IPs, including the city, country, ISP, and
whether the IP is a proxy.
This tracking will be used for a few purposes:
* Letting users view their account history, to detect things like logins
from unrecognized IPs, failed logins attempts, password changes, etc.
* Rate limiting failed login attempts.
* Detecting sockpuppet accounts using their login history.
* Detecting unauthorized account sharing.
Add a Restricted user level. Restricted users are level 10, below
Members. New users start out as Restricted if they sign up from a proxy
or an IP recently used by another user.
Restricted users can't update or edit any public content on the site
until they verify their email address, at which point they're promoted
to Member. Restricted users are only allowed to do personal actions
like keep favorites, keep favgroups and saved searches, mark dmails as
read or deleted, or mark forum posts as read.
The restricted state already existed before, the only change here is
that now it's an actual user level instead of a hidden state. Before it
was based on two hidden flags on the user, the `requires_verification`
flag (set when a user signs up from a proxy, etc), and the `is_verified`
flag (set after the user verifies their email). Making it a user level
means that now the Restricted status will be shown publicly.
Introducing a new level below Member means that we have to change every
`is_member?` check to `!is_anonymous` for every place where we used
`is_member?` to check that the current user is logged in.
