* Remember paused state for autoplay videos
* Update cached playback info
Update cached playback info instead of keeping track of it with a local variable
* Add missing null check
* Move logic into play()
Gallery can set aspect ratio to -1 which would result in negative measured height. It seem that Android treats dimensions as unsigned numbers, so negative numbers actually become huge positive numbers resulting in "infinite" height.
It should be noted that this change allows to disable aspect ratio logic even when some dimension is set to `wrap_content`.
Low effort/high impact case. Simple change that removes Retrofit instance creation from startup path because it is used only after receiving response from Reddit. Eliminates Retrofit instance creation completely if it is not used. Also improves type safety by replacing common Retrofit type with specific api type.
* Allow importing theme as any text
Browsers copy content as text/html and it caused theme import to fail
* Coerce clipboard data to text
Clipboard can have mime type of text even when it doesn't actually contain text. For example it can contain uri to text content provider. Using `coerceToText` automatically handles all cases and returns best text representation of content.
Unified branches, extracted repeatedly used expressions.
Fortunately both branches had the same logic, except for placeholder position hint calculation.
* Fix placeholder position check
Because of the missed negation `placeholderPosition` could actually point to a different comment. As a result loaded comments would be displayed at a wrong position which could result in duplicated comments.
* Extract placeholder search code
Simple refactoring. The only notable change is that now technically `placeholderPosition` can be -1. In practice that should never happen, but I added checks anyways
* Separate network dependency injection module
- Moved network-related dependencies into a separate module
- Consolidated common dependencies to save resources constructing a http client/retrofit
* Separate construction of access token interceptor
* Create providers for Context and Application to be injectable
* Refactor AppModule and AppComponent
- Use component builder to store application context and provide to modules
- Optimise AppModule providers
* Use component factory to add component dependencies
* Updated network dependencies to singleton.
Add missing OAuth base url for oauth request
Co-authored-by: Kurian Vithayathil <no.reply@github.com>
Co-authored-by: Docile-Alligator <25734209+Docile-Alligator@users.noreply.github.com>
* Use and store sort type as enum for comments
* Use and store sort type as enum for posts
* Use sort type enum for search
* Remove unused reddit api methods
These were missed in the first "sort type" pr
* Remove unused isHiddenInRecyclerView Post field
Was used only for reads and was always false
* Remove unused isHiddenManuallyByUser Post field
Was only written to and read for serialization. Never used for any logic
* Show the current image index in gallery in PostFragment.
* Fixed issue where filters applied to same feed were behaving incorrectly (#1172)
* Fix ItemTouchHelper and gallery swipe gesture fighting each other.
* Show correct image in the gallery in ViewRedditGalleryActivity.
* Swipe to view images in a gallery in Card Layout 2.
* Fix gallery layout issues.
* Remove `commentData` argument that is always an empty ArrayList
* Return top level comments in addition to expanded comments after parsing
Since 0f1c4d loading more comments loads not only first level replies, but also deeper comments. Because of this `expandedComments` can contain those deep replies if `expandChildren` is true. Adding `expandedComments` to parent causes a bug because parent's children are supposed to be only next level replies. Because of previously mentioned changes that is not true.
Now expanding parent comment results in duplicate comments: one of them correctly comes from the parent of duplicated comment. The other one is shown because it is incorrectly stored in the parent of "load more comments" button.
This comment separates top level comments (fist level replies) and expanded comments. `expandedComments` are still used for display, but only first level replies are added to the parent
* Add debug assertion for children depth
Co-authored-by: Docile-Alligator <25734209+Docile-Alligator@users.noreply.github.com>
Co-authored-by: Aidan223 <110802888+Aidan223@users.noreply.github.com>
* Move mAccessToken and mAccountName initialization upper
* Marked INSTANCE as a volatile to be DCL compliant
* fix a typo in a method name and remove redundant switch
* remove redundant ifs
* removed DCL on RedditDataRoomDatabase as its controlled by dagger
* remove redundant ifs in PostGalleryActivity as nColumns can be either 2 or 3 only
* replace MediaStore.Video.Media.IS_PENDING with MediaStore.MediaColumns.IS_PENDING
* reverting simplified ifs as it going to be used in the future
* reverting HistoryPostViewModel
* Create and add to Retrofit SortTypeConverterFactory
* Clean up SortType class
Mark fields as final, annotate nullability
* Use SortType enums as query arguments
* Use optional query arguments to simplify code
Retrofit ignores null values, so there is no need to remove them from arguments manually
* Load more comments from /morechildren endpoint
Previous implementation requested comments from /api/info which returned the comments themselves but did not include any information about their children. Also /api/info does not allow to specify sort type. On the other hand /morechildren supports sort type and it will be added in a later commit.
I am not proud of this implementation, but I had to fight with both Reddit api response and existing code. The problem with api response is that it is a flat list of comments, not a tree structure. So the tree has to be rebuilt on our end. And the problem with existing code is that it merges "more children" node into its parent but then adds a placeholder anyways.
The code relies on the fact that parent comment will be located before any of its children in the response. The code sequentially processes comments, tries to find their parents and either adds them to the tree or puts in a "top-level" array which will be handled by outside code.
One possible problem is the removal of `depth` argument from parsing as I couldn't find a way to fit it in the new logic. However I also didn't experience any problems with it during my testing and the response seems to always contain depth key. Moreover current depth handling logic in ParseComment#parseCommentRecursion is broken because it does not increment depth when making a recursive call.
* Store moreChildren ids instead of fullnames
/morechildren endpoint accepts ids instead of fullnames so there is no point in converting ids to fullnames and back
* Send all comment ids to Reddit so it can select what to display itself
Sending all ids allows Reddit to sort them properly. Since the number of comments can be very bing, it requires using POST request instead of GET.
This commit changes the meaning of Comment#moreChildrenIds field, now it stores only ids of comments that are not yet loaded. This simplifies the code and removes the need for Comment#moreChildrenStartingIndex
* Fetch more comments with current sort type
It used position of viewholder during binding which could get outdated by the time user clicks on the button. This would result in retrieving wrong comment or even null. Replaced with getting comment based on current position of vieholder.
When extracting sort type loading logic the shared preferences that are used to load sort type got accidentally changed to the wrong ones. This resulted in always using the default value which is displayed as Best.
Fortunately the saving code was not changed so only reading has to be fixed.