Apparently, this code was from back when we had a special "extensions"
directory for Composer packages marked as Flarum extensions.
While we're at it, we now inject the Paths instance instead of using one
of the global helpers (which I am trying to get rid of).
Refs #2055.
- Stop trying to implement Laravel's Application contract, which
has no value for us.
- Stop inheriting from the Container, injecting one works equally
well and does not clutter up the interfaces.
- Inject the Paths collection instead of unwrapping it again, for
better encapsulation.
This brings us one step closer toward upgrading our Laravel
components (#2055), because we no longer need to adopt the changes
to the Application contract.
This ensures that default values, date attributes and relationships are properly inherited, when we have deeper model class hierarchies.
This also adds test cases to ensure that inheritance order is honored for relationship and default attribute extender. As there's no way to remove date attributes, the order of evaluation there doesn't matter.
This lets us or anyone modify the path from where dependencies (usually
installed into /vendor by Composer) are loaded. We need to be able to
tweak this in our integration tests, where the application code under
test needs access to certain dependencies.
Refactored to use the Capsule Database manager for setting up the
Flarum (mysql) connection.
This will introduce the reconnector automatically, fixing #1740
These are not necessary to be available so broadly. In fact, they
seem to make it too easy to use them, which has lead to some sub-
optimal architecture decisions.
Their equivalents have been moved to the classes where used.
Updating the Migration::addPermission helper table name means we need
to move the seed migration to after the table rename migration. We also
add a sanity check for each permission's group since the foreign key
will fail if the group doesn't exist. Of course, the only way to make
sure groups are seeded before permissions is to move them into another
migration.
* Make filenames and order more consistent
* Split foreign keys into their own migrations, add statements to ensure
data integrity prior to adding them
* Add renameColumns helper, use other helpers where possible
Mostly, we only need a database connection, instead of one of
Laravel's "connection resolvers".
Again, this makes our life easier during installation, where
we already instantiate a database connection. We can now use
that to instantiate our own Migrator class, instead of using
the IoC container to build one.
This removes the funky auto-injection capability from migration
closures. While technically removing a feature, this means we do
not need a fully-wired IoC container e.g. during installation.
Instead, all migration closures simply receive a schema builder
object (which is what most of them were already doing anyway).
* Overhaul the way model visibility scoping works
- Previously post visibility scoping required concrete knowledge of the
parent discussion, ie. you needed a Discussion model on which you
would call `postsVisibleTo($actor)`. This meant that to fetch posts
from different discussions (eg. when listing user posts), it was a
convoluted process, ultimately causing #1333.
Now posts behave like any other model in terms of visibility scoping,
and you simply call `whereVisibleTo($actor)` on a Post query. This
scope will automatically apply a WHERE EXISTS clause that scopes the
query to only include posts whose discussions are visible too. Thus,
fetching posts from multiple discussions can now be done in a single
query, simplifying things greatly and fixing #1333.
- As such, the ScopePostVisibility event has been removed. Also, the
rest of the "Scope" events have been consolidated into a single event,
ScopeModelVisibility. This event is called whenever a user must have
a certain $ability in order to see a set of discussions. Typically
this ability is just "view". But in the case of discussions which have
been marked as `is_private`, it is "viewPrivate". And in the case of
discussions which have been hidden, it is "hide". etc.
The relevant API on AbstractPolicy has been refined, now providing
`find`, `findPrivate`, `findEmpty`, and `findWithPermission` methods.
This could probably do with further refinement and we can re-address
it once we get around to implementing more Extenders.
- An additional change is that Discussion::comments() (the relation
used to calculate the cached number of replies) now yields "comments
that are not private", where before it meant "comments that are
visible to Guests". This was flawed because eg. comments in non-public
tags are technically not visible to Guests.
Consequently, the Approval extension must adopt usage of `is_private`,
so that posts which are not approved are not included in the replies
count. Fundamentally, `is_private` now indicates that a discussion/
post should be hidden by default and should only be visible if it
meets certain criteria. This is in comparison to non-is_private
entities, which are visible by default and may be hidden if they don't
meet certain criteria.
Note that these changes have not been extensively tested, but I have
been over the logic multiple times and it seems to check out.
* Add event to determine whether a discussion `is_private`
See https://github.com/flarum/core/pull/1153#issuecomment-292693624
* Don't include hidden posts in the comments count
* Apply fixes from StyleCI (#1350)
