Update MDC to the alpha version in order to use some extra features.
I was planning to switch to the new MaterialSwitch, but alpha03 has
this insane crippling issue with ripples that blocks such. Use alpha02
and prep the app for the addition of the switch.
Lowercase music names when hashing them to prevent drift stemming from
grouping.
The addition of the song may change the case of an artist if such
mitigations are in effect. To prevent such from invalidating music
hashes, we take the lowercase of every name hashed.
Don't take the app widget host padding into account when trying to
calculate the widget sizing.
Doing such produces bad results, at least on my devices. At least, it
does now for some reason.
Fix an issue on some devices where the delayed actions system would
not work on startup.
Apparently onStart can be called several times with the same intent.
No, I am not joking. What the actual hell. This would cause any
delayed action other than Restore to be overwritten. The only hard
rule about android is that there are no rules. I hate this platform
so much.
Set the enforce*BarContrast methods to false above API 29, so we can
use fully transparent navigation and status bars.
Previously, we used a mostly transparent color for such, but now, we
can use a completely transparent color without it being changed to
match contrast.
Update id hashing to correctly handle null artists and take discs
and durations into account.
Note that we try not to hash values only obtained with the ExoPlayer
parser, as those could feasibly change if the setting was toggled,
thus causing the playback state to wipe.
Make the search algorithm take in account the raw sort name and file
name when searching.
This allows the user to search for a particular song without typing in
a unicode/non-ideal title, instead typing in a latinized sort name or
suitable file name.
This does make Auxio's search a bit more fuzzy, but it still gets the
job done.
Resolves#184.
Related to #172.
Add a new Date class to represent both years and more fine-grained
dates extracted using the ExoPlayer metadata system.
In-app, the year is still shown, but sorting will use the new precision
when present. The MediaSession will also post an RFC 3339 formatted
date with this new precision, as the MediaSession documentation states
I should. No clue if the latter will cause any bugs with naive metadata
UIs in other apps.
Resolves#159.
Implement sort tag support in the ExoPlayer backend.
Sort tags for grouping is still derived from the templates. Album
artist sort tags are only picked if one is present. System might be
a bit buggy at the moment given that it messes with grouping/sorting
a little.
Resolves#172.
Added date added sorting to albums as well.
This implementation picks the earliest song in an album that was added
to the library and then uses that for sorting.
Prepare support for sort tags by reworking how sortName is structured.
Instead of rawName and sortName, music must now implement rawName and
rawSortName. rawSortName will be checked first, followed by will
falling back to rawName.withoutArticle. This allows hard-coded tags to
be neatly implemented while avoiding an immediate copy.
Add a "Last Added" sorting option to the home UI's song list.
I don't know if there is any demand for last added in other contexts.
That will be resolved later.
Remove the excluded directory migration code, as it causes far more
issues than it fixes.
Due to an unfixable logic bug that occurs when trying to read the
setting, Auxio will always try to migrate the database when there is
no music folders, causing a hang in some situations. Fix it by just
removing the migration.
Add an abstraction over the generation system called GenerationGuard.
This allows cheap cooperative threading to be implemented consistently
in many places.
Add automatic rescanning, for real.
This is the culmination of 6 months of work to make Auxio respond to
music updates and to research the most versatile implementation of
such. Is it the best system? No. It's actually a bit messy by necessity
in order to prevent bugs. Does it work well? Yes.
This will not be enabled by default, as it does require a battery
draining foreground service and is generally not useful for most
circumstances.
So glad to be done with this.
Resolves#72.
Move the switch from IO to Main to within Indexer itself, through
withContext.
This is much easier to work with than the previous system of a separate
"update" coroutine, which isn't really needed anymore given that I no
longer need to do IO work when sanitizing the playback state.
Edge case I thought existed did not. PlaybackService must have saved
before dying, and thus if it did not foreground after a sanitization,
the saved state would still sanitize in a similar manner.
Save, but do not read the playback state when sanitizing.
Turns out there is an edge-case where we want to save the state. Still
keep the runtime sanitization, as that greatly reduces the time it
takes to rescan the library.
Do not save the playback state when sanitizing.
After some thought, there is no situation where re-saving the playback
state is desirable. A previously saved state will be consistent with
a sanitized state, and there is no need to save when the service is
active. Thus, save on speed and reduce insane race conditions by just
sanitizing the current runtime state and not saving at all.
Remove the ripple resources and replace them with their system
counterparts.
This is to remove redundancy in-app and make the widget more consistent
with other android widgets.
Finally unify all icon activation states under the new grade modifier
in Material Symbols.
This provides similar clarify to the dot/greyed state, but without
inconsistent meanings. The shuffle icon did have to be bolded a little
more though, as the grade did not seem to do much.
Rework the UI flow for music loading to be more coherent with runtime
rescanning.
The loading progress is now shown as a card on the bottom of the
screen. This way, app use is not completely crippled when the app has
to rescan the music library, albeit the shuffle button still has to
be disabled during this period.
Revert the optional rounded corners on the bottom sheet.
It was causing too many bugs to be a sensible addition. I will only add
it if Google moves WindowInset application to layout time AND only on
devices with the stretch overscroll implemented.
Audit usages of Synchronized throughout the app to prevent deadlocks.
This is primarily composed of not making long-running work
synchronized, instead only making mutations and reads synchronized.
This does cause minor issues, such as during a sanitization event
the playback state could be feasibly saved, changed, and then restored
back to the previous state unintentionally. However, preventing
deadlocks is generally better than trying to fix those.
Use the @Synchronized annotation instead of synchronized.
Makes my ability to manage thread-safety on the shared objects much
easier. Because I don't have to think about what I should guard
and what I shouldn't.
Finally enable runtime rescanning, opening the door for a ton of new
features and automatic rescanning later on.
More work needs to be done on making the shared mutable state in-app
safer to use.
Improve the indexer callback system to be more coherent and efficient.
This delegates the old Callback role to a new singular Callback and
Controller roles. IndexerService also handles the loading process
more gracefully, reducing the amount of time music loads take.
Make notification updates entirely reliant on the MediaSession.
Android 11 and onwards automatically populate the notification with the
MediaSession state. This apparently conflicts with updating the
notification in some cases, resulting in the incorrect song being
shown. Fix this by not populating the notification from Android 11
onward and only posting it when the MediaSession state was set.
Resolves#179.
Remove WindowInsetsCompat from the project, as it is a pile of garbage.
WindowInsetsCompat refuses to actually update the correct fields when
you transform it back into a WindowInsets. This makes it impossible to
use in Auxio. The fact that Google decided to make such an
overengineered wrapper to some version-specific methods is an absolute
joke. All you needed were some extension functions, but no, let's make
an entire wrapper class filled with so many gotchas as to be unusable.
Rework the rounded covers option into a new "Round Mode" option.
This commit extends the rounded corners configuration to now the
widget, thus making the setting apply now to covers, the bar, and
the widget configuration. This makes a naming change useful.
