Migrate all esoteric adapters to the new RecyclerView framework.
One of the shortcomings with the previous RecyclerView utilities was
how more esoteric adapters with data that does not implement Item
could not use the utlities. The new system, by comparison, is capable
of taking any type of data, so we can no migrate all of the esoteric
adapters to the new system.
Fix an issue where album covers fail to load when not explicitly
specifying a size.
At some point, Coil changed some internal scaling logic to depend on
this weird pseudo-screen size value that caused the bitmap to fail
to load only on in widget creation process. Who knows why. Fix this
by explicitly specifying the size in all cases.
Spin off the data instances into their own class called BackingData.
This is to isolate the sane adapters that rely on one type of diffing
from the insane adapters that use synchronous and asynchronous diffing
simultaniously. It also allows some of the more esoteric adapters to
implement their own backing data without much trouble or leaky
abstractions.
Completely rework the base adapter class to require less boilerplate
and properly handle cases such as diffing. The major adapters have
been migrated to this system, but the other adapters have not been
changed so far.
This is only part 1 of a multi-part rework, as this is an incredibly
complex system.
Remove databinding entirely.
Databinding was a terrible idea for Auxio. I rarely leveraged it, and
when I did, it produced messy code and bloated build times. Dumpster it
for just viewbinding, which is good. This reduces building times by
nearly 2/3, and generally makes the codebase more coherent and usable.
Move all fragment instances to the new ViewBindingFragment paradigm,
which should help immensely with reducing memory leaks from list
bindings and to really alleviate the overloaded onCreate functions.
Rework the playback views to follow the new idioms I want to use for
Auxio.
This change mostly consists of flattening away some of the custom
views, removing databinding, and using a general viewbinding
fragment that I hope to extend to the entire app.
Rework the playback slide up implementation to be more straightfoward.
This is really composed of stylistic improvements, very little in
actual behavior changes. This does re-introduce a regression when
nothing is playing where the scroll position will become off when
rotating, but that desynchronization happens often so unless I were
to completely migrate both the panel and the bar to a view, I don't
really see it as an issue.
Migrate to a centralized constant table for easier management.
Previously, Auxio would tie constants to the class itself, which
led to a largely disjointed system that relied on an internal table
so that it would stay sane. This commit moves all of those constants
to a single table for easier usage and management.
Completely refactor the Sort class to be more efficient and
straightforward.
The original Sort class had *major* shortcomings, it was slow,
poorly organized, and relied on abusing compareBy to implement
special things like article sort. This rework eliminates all of
that in favor of a new system relying on custom comparators and
chaining to achieve something much faster and maintainable.
Make resolvedName the standard name to use across all music items.
Previously, the way Auxio differentiated between names in UIs was to
use name for normal items and resolvedName for parents. This was odd
and disjointed, as it muddled the meanings of the two fields and lead
to a lot of bugs. Fix this by making resolvedName *the* standard name
to use across all music items, even in ones that don't have to resolve
their name.
Switch to the spotless linter with ktfmt used as a backend instead of
ktlint.
This switch was done for two reasons:
1. ktfmt is more thorough than ktlint
2. License headers can be added more effectively with spotless than
the default Android Studio behavior.
Dump all of the changes now so I don't have to deal with it over a long
period of time. I don't care.
Disable the default long-press action in the ItemTouchHelper usages.
ItemTouchHelper provides a long-press action to start an item drag by
default. However, because Auxio adds a drag handle on top of this
action, this actually results in a conflict with the default behavior
in certain cases. Replace it with a custom version of the long-press
action within the viewholders themselves.
Fix a crash that stemmed from the use of the API 30+ database field
CD_TRACK_NUMBER.
I switched to CD_TRACK_NUMBER in 1e39ceb to fix [#88], however I did
not realize that CD_TRACK_NUMBER was actually only supported on API
30 onwards. This led to a load failure on versions before API 30.
Fix an esoteric crash with queue synchronization during the playback
restore process.
Auxio will attempt to re-synchronize the queue index whenever it is
desynchronized, however during the check for if it's desynchronized,
Auxio would do a direct index of the queue, which could result in a
crash in situations where the desynchronized index is outside of the
queue bounds.
Fix this by replacing that unprotected access with a protected access,
which not only fixes the crash but also still correctly detects
desynchronization in that case.
Resolves#89.
Turn the headset focus setting into the headset autoplay setting.
The way auxio handles headsets is...odd. Sometimes the MediaSession
handles it and Auxio could not care less, and sometimes Auxio actually
needs to handle it. As a result, the idea of being able to disable
headset focus is more or less moot because it will only apply to some
devices and not others.
On the other end, the way Auxio automatically begins playback once a
headset is plugged in is also quite weird. It only works on wired
headsets, and when it does, it overrides all other apps that might
also be playing audio. It's not to say that it's a bad feature, but
it's also one that I don't want to make the defualt. Auxio should
still play along within the confines of Android's expectations, after
all.
Replacing the existing "Headset focus" setting with a new "Headset
autoplay" setting solves both of these issues, as it prevents a
mis-guided disabling of the setting that doesn't actually disable
the feature and it relegates the quirky autoplay behavior to an
setting not enabled by default.
Fix an issue where headset focus would result in unexpected playback
whenever the service started.
AudioManager.ACTION_HEADSET_PLUG seems to always fire when the initial
BroadcastReceiver is set up, which results in a weird bug where if a
wired headset is connected while PlaybackService is started, playback
will start immediately, which was not user friendly.
I fear that this may result in an edge case that results in headset
focus not firing in an unrelated situation, which in that case I would
be forced to remove headset autoplay entirely (or at least relegate it
to a quirk option).
Add even more checks and guards to the music loading process to ensure
proper metadata loading.
Auxio has always had issues handling track numbers and years, mostly
with ensuring their validity. This change resolves them by more or less
surrounding them with gobs of null-checks and fallible parsing
routines, which should help avoid frustrating bugs and crashes with
metadata in the future.
Resolves#88.
Resolves#84.
Revert the removal of the song/artist values from list items for now.
The current plan of removing extraneous values from songs and albums
only works if I add multi-select and a dedicated menu button, but that
is still in the air, and in general I want feedback before I go ahead.
Aside from reverting the previous changes, this actually standardizes
all item descriptions under a single bullet point formatting system,
instead of the mix of commas and bullet points that existed beforehand.
Remove the album name from the main song item in favor of a duration
value.
Auxio has traditionally used "Artist . Album" for song items. However,
this had some shortcomings:
1. The way the artist and album names are packed probably results in
truncation on small screens, which I doubt is very helpful.
2. All other items in Auxio have only one specific "subject" per line,
which makes the song items a bit of an outlier since it has two (the
artist and the album)
3. The empty space available could be used for another UI element,
such as a duration or maybe a menu button in the future.
For consistency, this also removes the song count from all album items
as well.
Remove all usages of memberBinding from the app.
For some reason, certain devices running Android 10 and lower will
have a lifecycle race condition whenever the theme is mis-matched.
This ends up resulting in an invalid state whenever memberBinder was
used. Since we can't replace memberBinder with a better solution,
just dumpster the whole thing. This platform is so god damn broken,
jesus christ.
Resolves#80.
Change createMosaic to use SquareFrameTransform.
This should help fix a theoretical bug where non-1:1 album covers will
be oddly stretched in a generated mosaic, as all album covers will be
forcefully cropped to a 1:1 ratio.
Create a custom view for rounded images, making them more nuanced in
the process.
The previous method for applying rounded images in-app was generally
clunky and fragile. Introduce a new custom view that actually takes a
cornerRadius attribute from the ImageView itself that then applies it
whenever the user enables the setting. This also allows rounded images
to be more nuanced, as typical 8dp elevation can be used for small
views and a more fitting 16dp radius can be used for large views.
Rework how the fab's adaptive functionality works.
Handle fab adaptiveness by using a style instead of a fully custom
view. This helps eliminate a usage of private resources.
Rework logging to be clearer and more standardized.
Rework all usages of lossing to follow a single unified style,
introducing a new "warn" option alongside this.
Crop all images to a 1:1 aspect ratio.
Sometimes cover art will not be a square. I never realized this because
all covers in my music library are square. This led to many strange bugs
that I would rather avoid, so create a new transformation that crops the
image to a 1:1 aspect ratio.
The reason why we use a transformation is that it won't be dependent on
the actual state of the ImageView, which is what would happen if we
used ScaleType.CENTER_CROP. In the case that we use a very weird hacky
ImageView like in the widgets, doing such would result in any cropping
not actually working correctly. It's better to do it in the
Transformation stage simply to ensure consistency.
Allow the material library to handle dynamic colors.
Turns out I was mis-understanding how DynamicColor themes were meant to
be used by the material library. Turns out you can just inherit from it
and it will work perfectly fine. Refactor the V31 styles to remove our
insane shims and finally fix the slapdash dynamic color usage across
the main theme and widgets.
