* attempt some sort of "usurping" system
* polling-based active video approach
* split into inner component again
* click to steal active video
* disable findAndActivateVideo on native
* new intersectionobserver approach - wip
* fix types
* disable perf optimisation to allow overflow
* make active player indicator subtler, clean up video utils
* partially fix double-playing
* start working on controls
* fullscreen API
* get buttons working somewhat
* rm source from where it shouldn't be
* use video elem as source of truth
* fix keyboard nav + mute state
* new icons, add fullscreen + time + fix play
* unmount when far offscreen + round 2dp
* listen globally to clicks rather than blur event
* move controls to new file
* reduce quality when not active
* add hover state to buttons
* stop propagation of videoplayer click
* move around autoplay effects
* increase background contrast
* add subtitles button
* add stopPropagation to root of video player
* clean up VideoWebControls
* fix chrome
* change quality based on focused state
* use autoLevelCapping instead of nextLevel
* get subtitle track from stream
* always use hlsjs
* rework hls into a ref
* render player earlier, allowing preload
* add error boundary
* clean up component structure and organisation
* rework fullscreen API
* disable fullscreen on iPhone
* don't play when ready on pause
* debounce buffering
* simplify giant list of event listeners
* update pref
* reduce prop drilling
* minimise rerenders in `ActiveViewContext`
* restore prop drilling
---------
Co-authored-by: Samuel Newman <10959775+mozzius@users.noreply.github.com>
Co-authored-by: Hailey <me@haileyok.com>
* Broadcast the update in the same tick
The motivation for the original code is unclear. I was not able to reproduce the described behavior and have not seen it mentioned on the web. I'll assume that this was a misunderstanding.
* Remove defensive programming
The only places in this code that we can expect to throw are schema.parse(), JSON.parse(), JSON.stringify(), and localStorage.getItem/setItem/removeItem. Let's push try/catch'es where we expect them to be necessary.
* Don't write or clobber defaults
Writing defaults to local storage is unnecessary. We would write them as a part of next update anyway. So I'm removing that to reduce the number of moving pieces.
However, we do need to be wary of _state being set to defaults. Because _state gets mutated on write. We don't want to mutate the defaults object. To avoid having to think about this, let's copy on write. We don't write to this object very often.
* Refactor: extract tryParse
* Refactor: move string parsing into tryParse
* Extract tryStringify, split logging by platform
Shared data parsing/stringification errors are always logged. Storage errors are only logged on native because we trust the web APIs to work.
* Add a layer of caching to readFromStorage to web
We're going to be doing a read on every write so let's add a fast path that avoids parsing and validating.
* Fix the race condition causing clobbered writes between tabs
* Delete logic for legacy storage
* Delete superfluous tests
At this point these tests aren't testing anything useful, let's just get rid of them.
* Inline store.ts methods into persisted/index.ts
* Fork persisted/index.ts into index.web.ts
* Remove non-essential code and comments from both forks
* Remove async/await from web fork of persisted/index.ts
* Remove unused return
* Enforce that forked types match
* Don't reconstruct threads from separate posts
* Remove post-level dedupe for now
* Change repost dedupe condition to look just at length
* Delete unused isThread
* Delete another isThread field
It is now meaningless because there's nothing special about author threads.
* Narrow down slice item shape so it does not need reply
* Consolidate slice validation criteria in one place
* Show replies in context
* Make fallback marker work
* Remove misleading and now-unused property
It was called rootUri but it was actually the leaf URI. Regardless, it's not used anymore.
* Add by-thread dedupe to non-author feeds
* Add post-level dedupe
* Always count from the start
This is easier to think about.
* Only tuner state need to be untouched on dry run
* Account for threads in reply filtering
* Remove repost deduping
This is already being taken care of by item-level deduping. It's also now wrong and removing too much (since it wasn't filtering for reposts directly).
* Calculate rootUri correctly
* Apply Following settings to all lists
* Don't dedupe intentional reposts by thread
* Show reply parent when ambiguous
* Explicitly remove orphaned replies from following/lists
* Fix thread dedupe to work across pages
* Mark grandparent-blocked as orphaned
* Guard tuner state change by dryRun
* Remove dead code
* Don't dedupe feedgen threads
* Revert "Apply Following settings to all lists"
This reverts commit aff86be6d37b60cc5d0ac38f22c31a4808342cf4.
Let's not do this yet and have a bit more discussion. This is a chunky change already.
* Reason belongs to a slice, not item
* Logically feedContext belongs to the slice
* Update comment to reflect latest behavior
* Prep
* Pass in optional moderation to FeedCard
* Compute moderation decision, filter contentList contexts, pass into card
* Let's go a different route
* Filter from within search queries
* Use same search query for starter packs
* Filter lists from profile tabs
* Cleanup
* Filter from profile feeds
* Moderate post embeds
* Memoize
* Use ScreenHider on lists
* Hide both list types
* Fix crash on iOS in screen hider, fix lineheight
* Memoize renderItem
* Reuse objects to prevent re-renders
* state for video uploads
* get upload working
* add a debug log
* add post progress
* progress
* fetch data
* add some progress info, web uploads
* post on finished uploading (wip)
* add a note
* add some todos
* clear video
* merge some stuff
* convert to `createUploadTask`
* patch expo modules core
* working native upload progress
* platform fork
* upload progress for web
* cleanup
* cleanup
* more tweaks
* simplify
* fix type errors
---------
Co-authored-by: Samuel Newman <10959775+mozzius@users.noreply.github.com>
* Copy FeedViewPost into FeedSliceItem
* Explicitly construct feed slice items by copying known fields
* Type rootItem as FeedViewPost for now
Mergefeed logic relies on that.
* Unify reason and __source for slice items
* Move feedContext out of FeedSliceItem
* Remove slice.isFlattenedReply
* Remove unnused slice.ts
* Inline slice.isFullThread
* Refactor condition for clarity
* Extract slice.includesThreadRoot
* Encapsulate more usages of slice.rootItem into slice
* Rename slice.rootItem so semi-private slice._feedPost
* Move reason into slice
* Simplify slice ctor argument
* Reorder getters to reduce diff
* Make feedContext a getter to reduce diff
* add ffmpeg-kit-react-native
* get select video button + compression working
* up res to 1080p
* add progress component
* move logic out of compressVideo
* (WIP) add lonestar compression
* rework web compression a bit
* mess around with adding a thumbnail
* 3mbps
* replace
* use 3mbps
* add expo-video
* remove unnecessary try/catch
* rm ToastAndroid
* fix web
* wrap lazy component in suspense
* gate video select button
* rm web compression
* flip sign
* remove expo-video from web
* review nits
* add video picker permissions + rm temp buttons
* add ffmpeg-kit-react-native
* replace
* hls-capable player
* start trying to hoist up video player instance
* hoist video player and move things around
* always show native controls
* fix controls on expo video android
* gate temp video player in feed
* rm IS_DEV, doesn't do what I thought it did
* use __DEV__ instead
---------
Co-authored-by: Samuel Newman <10959775+mozzius@users.noreply.github.com>
Co-authored-by: Hailey <me@haileyok.com>
Eric Bailey
dan
Minseo Lee
Hailey
Samuel Newman
bnewbold
GSMT
Dmitrii Kartashev