Convert `MetadataHandler` to an abstract base class and add an
`is_enabled` class method that allows every metadata handler to
independently report whether it is enabled based on its configuration.
This avoids the need for global variables in the config module, allowing
us to change the enabled state of a metadata handler at runtime if
needed.
Add a new scheduled task that syncs RetroAchievements progress for all
users with a RetroAchievements username.
Environment variables:
- `ENABLE_SCHEDULED_RETROACHIEVEMENTS_PROGRESS_SYNC`: Enable or disable
the task (default: `false`)
- `SCHEDULED_RETROACHIEVEMENTS_PROGRESS_SYNC_CRON`: Cron string to
schedule the task (default: "0 4 * * *" - daily at 4 AM)
The cache for fixture files was being set without a TTL, which meant
that updates to the fixture files were not reflected in the cache.
This change saves the MD5 hash of the fixture file in a different key,
and compares it to the current hash of the file before deciding whether
to update the cache.
Fixes#2347
Add a new service adapter for the IGDB API, to separate concerns with
RomM's handler for metadata. This adapter is agnostic to the handler and
only provides methods to interact with the API, and correctly return
typed responses.
The API authorization was also improved to not rely on decorating each
method that makes requests, but instead using an `aiohttp` middleware
to automatically add the required headers to each request.
Utils `mark_expanded` and `mark_list_expanded` where added to help
narrow the types of IGDB's expandable fields when we know they are
expanded, for `mypy` type checking.
Guarantee that cache is initialized during startup, and only once,
instead of every time a `MetadataHandler` object is instantiated.
Also, improve logic to determine `fixtures` paths.
When retrieving the related screenshot for a `Save` or `State`, we were
retrieving a very heavy representation of the associated `Rom` object,
only to iterate through its screenshots to find the one we needed.
This change modifies the `Save` and `State` models to directly query the
`Screenshot` model, which is much faster and more efficient. The
`DBScreenshotsHandler` has been updated to include a new `filter` method
that will simplify building queries using SQLAlchemy, something we can
extend to other handlers in the future.
Fixes#1925.
