These were originally used to find uploaders with high-scoring uploads
for promotion purposes. The weekly reports generated by Reportbooru are
better suited for this.
Also clean up some unused and redundant code.
Add option to allow for storing original files under an `original/`
subdir. Previously preview, crop, and sample files were stored under
subdirs but original files were stored in the top-level directory. This
is optional for backwards compatibility.
Fix bug where pruning uploads failed because the rclone binary couldn't
be found. The upload pruner runs under cron, which has a fixed default
$PATH of "/bin:/usr/bin", but in production rclone is installed under
/usr/local/bin. This caused the upload pruner to fail, which prevented
the rest of daily maintenance from running.
* Change minimum activity threshold to 30 approvals in 45 days.
* Exclude mods from losing privileges, not just admins.
* Fix undeletions of old posts not counting towards the activity threshold.
* Switch CloudflareService from HttpartyCache to Danbooru::Http.
* Purge cached urls from Cloudflare when a post is replaced and the md5
doesn't change. This happens when a corrupted image is replaced or
thumbnails are regenerated. Before we purged urls when a post was
expunged, which was unneeded because those urls can expire naturally.
It was also wrong because the subdomains were hardcoded, the urls used
http:// instead of https://, and we didn't account for tagged urls.
The twitter gem had several problems:
* It's been unmaintained for over a year.
* It pulled in a lot of dependencies, many of which were outdated. In
particular, it locked the `http` gem to version 3.3, preventing us
from upgrading to 4.2.
* It raised exceptions on normal error conditions, like for deleted
tweets or suspended users, which we really don't want.
* We had to wrap it to provide caching.
Changes:
* Fixes#4226 (Exception when creating new artists entries for suspended
Twitter accounts)
* Drop support for scraping images from summary cards. Summary cards
are the previews you get when you link to a website in a tweet. These
preview images aren't always the best image.
* B2 doesn't allow the path to start with a '/' character.
* When storing the file, we have to rewind the file pointer to make sure
we get the whole file.
Fix an error in operator precedence:
> sum + Tag.find_by_name(token[1]).try(:post_count) || 0
This was treated as `(sum + X) || 0` not `sum + (X || 0)` as intended.
This failed when X was nil.
Filter out the user's own uploads and favorites from their
recommendations.
Note that in most cases a user's top-N recommendations will be things
they've already favorited. If a user has 10,000 favorites, most of their
top 10,000 recommendations will be their own favorites, so we have to
generate a little more than 10,000 recommendations to be sure they won't
all be filtered out.
In other words, the more favorites a user has, the more recommendations
we have to generate. The upper bound is clamped to 50,000 for
performance reasons. If a user has more favorites than this we may not
be able to find any recommendations for them.
