Persist book.subjects and add_author when form validation fails.
This does not resolve the problem of cover image uploads being dropped because this is a broader problem and is covered separately in #2760
(we should investigate the plugin ` django-file-resubmit`)
Previously when the deduplicate_book_data script tried to merge an
edition that was on a shelf or in a list then it would fail because when
the canonical book was added to the shelf or the list then it wouldn’t
set the extra fields of the linking table for the “through” model of the
field. These would end up defaulting to NULL, but that is not valid for
some of the fields in ShelfItem and ListItem so postgres wouldn’t accept
it.
To fix that, this patch makes it skip updating fields that have a
non-autogenerated linking table. The linking table would appear as a
separate model anyway so the book will be moved via that instead.
Fixes: #2817
GoToSocial sends 'tag' values as a single object if there is only one
user mentioned, rather than an array with an object inside it.
This causes Bookwyrm to reject the tag since it comes through as a
dict rather than a list.
This commit fixes this at the point the incoming AP object is transformed
so that "mention" tags are turned into a mention_user.
Docker is removing support for docker-compose, and it doesn't appear to
be possible to install it anymore. Instead, it has been replaced with
compose V2 which is a docker plugin called with 'docker compose' (no
hyphen). See https://docs.docker.com/compose/compose-v2/
Thanks to @Neriderc for noticing this in #2781.
Bookwyrm keyIds are at `userpath/#main-key`, however when signing AP objects we have claimed in the headers that the keyId is at `userpath#main-key`.
This is incorrect, and makes GoToSocial's strict checking break.
Simply updating the signatures to use the correct KeyId breaks legacy Bookwyrm's signature checks, becuase it assumes that the keyId path is the same as the user path plus a fragment.
This commit allows for either option, by sending the request a second time with the incorrect keyId if sending with the correct one causes an error.
Previously the 'tag' value in an activitypub object was assumed to be a List (array).
Some AP software sends 'tag' as a Dict (object) if there is only a single tag value.
It's somewhat debatable whether this is spec compliant but we should aim to be robust.
This commit puts an individual mention tag inside a list if necessary.
Fixes#2801
Related to #2794
It is legitimate to use any url for the user's key id. We have been assuming this id is the user id plus a fragment (#key-id) but this is not always the case, notably in the case of GoToSocial it is at /key-id. This commit instead checks the remote user's information to see if the key id listed matches the key id of the message allegedly received from them.
Whilst troubleshooting this it also became apparent that there is a mismatch between Bookwyrm users' keyId and the KeyId we claim to be using in signed requests (there is a forward slash missing). Since everything after the slash is a fragment, this usually slips through but we should be consistent so I updated that.
The series link needs an author so if it doesn't have one, instead of
showing a server error let's just show the series details as plain text
without a link.
Fixes: #2797
This is essentially a revert of 9cbff312a. The commit was at the advice
of the Celery docs for optimization, but I've since decided that the
downsides in terms of making things harder to debug (it makes Flower
nearly useless, for instance) are bigger than the upsides in performance
gain (which seem extremely small in practice, given how long our tasks
take, and the number of tasks we have).
This avoids filtering for the user that made the post in the same query
as we use for other things, which should allow for better use of indices
in all cases. Previously, #2723 did some work on this that only worked
for some cases in HomeStream, but this code should work for all cases.
Related: #2720
