* Attempt to speed up sidebar menu for mobile (#283)
* Exclude default homepage data at compile time
The youtuber IDs alone is pretty huge, and is unused in the `CUSTOM_HOMEPAGE=true` configuration.
* Remove Desktop items and other cleanup
- Moved constants out of the component.
- Remove SIMPLE_SITE check.
- Remove Desktop-only items
* Sidebar: limit subscription and tag section
Too slow for huge lists
Limit to 10 initially, and load everything on "Show more"
* Fix makeSelectThumbnailForUri
- Fix memo
- Expose function to extract directly from claim if client already have it.
* Fix and optimize makeSelectIsSubscribed (#273)
- It will not return true if the uri provided is canonical, because the compared subscription uri is in permanent form. This was causing certain elements like the Heart to not appear in claim tiles.
- It is super slow for large subscriptions not just because of the array size + being a hot selector, but also because it is looking up the claim twice (not memo'd) and also calling `parseURI` to determine if it's a channel, which is unnecessary if you already have the claim.
- Optimize the selector to only look up the claim once, and make operations using already-obtained info.
* Simplify makeSelectTitleForUri
No need to memo given no transformation.
* Simplify makeSelectIsUriResolving
- Memo not required. `resolvingUris` is very dynamic and is a short array anyways.
- Changeg from using `indexOf` to `includes`, which is more concise.
* Cost Info selector fixes
- no memo required since they are just directly accessing the store.
Co-authored-by: infinite-persistence <64950861+infinite-persistence@users.noreply.github.com>
Co-authored-by: infinite-persistence <inf.persistence@gmail.com>
Frequently used; top in perf profile
Most of the time, you already have the claim object in the current context. `selectClaimIsMineForUri` will retrieve the claim again, which is wasteful, even if it is memoized (looking up the cache still takes time).
Break apart the logic and added the alternative `selectClaimIsMine` for faster lookup.
Co-authored-by: infinite-persistence <inf.persistence@gmail.com>
`makeSelectDataForUri` always returns a new reference, so `ClaimPreview` was constantly being rendered. It's pretty expensive since `ClaimPreview`'s rendering checks against a huge blocklist, which is another issue on it's own.
- This commit tests the usage of `re-reselect` as the solution to the multi-instance memoization problem (https://github.com/toomuchdesign/re-reselect/blob/master/examples/1-join-selectors.md)
* Add option to pass in url-search params.
Impetus: allow linked comment ID and setting the discussion tab when clicking on the `ClaimPreview`.
* comment.list: fix typos and renamed variables
- Switch from 'author' to 'creator' to disambiguate between comment author and content author. For comment author, we'll use 'commenter' from now on.
- Corrected 'commenterClaimId' to 'creatorClaimId' (just a typo, no functional change).
* doCommentReset: change param from uri to claimId
This reduces one lookup as clients will always have the claimID ready, but might not have the full URI.
It was using URI previously just to match the other APIs.
* Add doCommentListOwn -- command to fetch own comments
Since the redux slice is set up based on content or channel ID (for Channel Discussion page), re-use the channel ID for the case of "own comments". We always clear each ID when fetching page-0, so no worries of conflict when actually browsing the Channel Discussion page.
* Comment: add option to hide the actions section
* Implement own-comments page
* Use new param to remove sort-pins-first.
comment.List currently always pushes pins to the top to support pagination. This new param removes this behavior.
* ❌ Remove old method of displaying active livestreams
Completely remove it for now to make the commit deltas clearer.
We'll replace it with the new method at the end.
* Fetch and store active-livestream info in redux
* Tiles can now query active-livestream state from redux instead of getting from parent.
* ⏪ ClaimTilesDiscover: revert and cleanup
## Simplify
- Simplify to just `uris` instead of having multiple arrays (`uris`, `modifiedUris`, `prevUris`)
- The `prevUris` is for CLS prevention. With this removal, the CLS issue is back, but we'll handle it differently later.
- Temporarily disable the view-count fetching. Code is left there so that I don't forget.
## Fix
- `shouldPerformSearch` was never true when `prefixUris` is present. Corrected the logic.
- Aside: prefix and pin is so similar in function. Hm ....
* ClaimTilesDiscover: factor out options
## Change
Move the `option` code outside and passed in as a pre-calculated prop.
## Reason
To skip rendering while waiting for `claim_search`, we need to add `React.memo(areEqual)`. However, the flag that determines if we are fetching `claim_search` (fetchingClaimSearchByQuery[]) depends on the derived options as the key.
Instead of calculating `options` twice, we moved it to the props so both sides can use it.
It also makes the component a bit more readable.
The downside is that the prop-passing might not be clear.
* ClaimTilesDiscover: reduce ~17 renders at startup to just 2.
* ClaimTilesDiscover: fill with placeholder while waiting for claim_search
## Issue
Livestream claims are fetched seperately, so they might already exists. While claim_search is running, the list only consists of livestreams (collapsed).
## Fix
Fill up the space with placeholders to prevent layout shift.
* Add 'useFetchViewCount' to handle fetching from lists
This effect also stashes fetched uris, so that we won't re-fetch the same uris during the same instance (e.g. during infinite scroll).
* ⏪ ClaimListDiscover: revert and cleanup
## Revert
- Removed the 'finalUris' stuff that was meant to "pause" visual changes when fetching. I think it'll be cleaner to use React.memo to achieve that.
## Alterations
- Added `renderUri` to make it clear which array that this component will render.
- Re-do the way we fetch view counts now that 'finalUris' is gone. Not the best method, but at least correct for now.
* ClaimListDiscover: add prefixUris, similar to ClaimTilesDiscover
This will be initially used to append livestreams at the top.
* ✅ Re-enable active livestream tiles using the new method
* doFetchActiveLivestreams: add interval check
- Added a default minimum of 5 minutes between fetches. Clients can bypass this through `forceFetch` if needed.
* doFetchActiveLivestreams: add option check
We'll need to support different 'orderBy', so adding an "options check" when determining if we just made the same fetch.
* WildWest: limit livestream tiles + add ability to show more
Most likely this behavior will change in the future, so we'll leave `ClaimListDiscover` untouched and handle the logic at the page level.
This solution uses 2 `ClaimListDiscover` -- if the reduced livestream list is visible, it handles the header; else the normal list handles the header.
* Use better tile-count on larger screens.
Used the same method as how the homepage does it.
## Issue
6776 Missing follow button on channel reposts
- Missing:
- Fix: For reposts, `parseURI` will fail to determine if it's a channel because the URL doesn't have enough info. Determine from the claim object instead.
- Not working in Channel Page:
- Fix: The repost uri doesn't link back to the source channel, so we'll need to handle it.
* recsys wip
better logging
fix floating player popout playing uri bug with recsys
lint
add empty entries to create
use beacon; fire on visibilitychange
cleanup, not record recs if not seen
ifweb recsys beacon
recsys handle embeds, cleanup
use history.listen to trigger events
fix recsys embed bug
bugfix
more default data
cleaner
cleaner
* remove tentative
* disable recsys debug logging
## Issue
6366 Recsys Evaluation Telemetry
The recommended list from lighthouse is obtained from `makeSelectRecommendedContentForUri`. This list is further tweaked by the GUI (e.g. move autoplay next item to top, remove blocked content, etc.). Recsys wants the final recommendation list and the clicked index (in exact order), so we need pass these info to the videojs recsys plugin somehow. Also, Recsys wants a recommendation list ID as well as the parent (referrer) ID, we so need to track the clicks and navigation.
## General Approach
- It seems easiest to just spew back the final (displayed) list and all the required info to Redux, and the recsys plugin (or anyone else in the future) can grab it.
- Try to touch few files as possible. The dirty work should all reside in `<RecommendedContent>` only.
## Changes
- `ClaimPreview`: add optional parameters to store an ID of the container that it is in (for this case, it is `ClaimList`) as well as the index within the container.
- When clicked, we store the container ID in the navigation history `state` object.
- For general cases, anyone can check this state from `history.location.state` to know which container referred/navigated to the current page. For the recsys use case, we can use this as the `parentUUID`.
- `ClaimList`: just relay `onClick` and set IDs.
- `RecommendedContent`: now handles the uuid generation (for both parent and child) and stores the data in Redux.
Pop up Menu Fixes and Improvements
Improve re-directs
Fix file deletion
Improve code, and more
Fix List Layout collection popup
Allow to edit list
Fix blocking when not logged or no channel
Fix Edit and Delete showing for not owned Lists
Fix actions breaking when not logged in
Fix List options not showing
Lint
Shorten logic
Lint
add watch later hover action
replace watch later popup item for favorites
lint
styling for watch_later overlay
Add label
Use just claim, add requiresAuth
Add list icon
Tone down text
Turn WL Hover Button into component
Change WL hover icons
small revert
Keep watch later in the menu
- [x] (6332) The IntersectionObserver method of lazy-loading loads cached images visibly late on slower devices. Previously, it was also showing the "broken image" icon briefly, which we mended by placing a dummy transparent image as the initial src.
- Reverted that ugly transparent image fix.
- Use the browser's built-in `loading="lazy"` instead. Sorry, Safari.
- [x] Size-optimization did not take "device pixel ratio" into account.
- When resizing an image through the CDN, we can't just take the dimensions of the tag in pixels directly -- we need to take zooming into account, otherwise the image ends up blurry.
- Previously, we quickly disabled optimization for the channel avatar in the Channel Page because of this. Now that we know the root-cause, the change was reverted and we now go through the CDN with appropriate sizes. This also improves our Web Vital scores.
- [x] Size-optimization wasn't really implemented for all ChannelThumbnail instances.
- The CDN-optimized size was hardcoded to the largest instance, so small images like sidebar thumbnails are still loading images that are unnecessarily larger.
- There's a little-bit of hardcoding of values from CSS here, but I think it's a ok compromise (not something we change often). It also doesn't need to be exact -- the "device pixel ratio" calculate will ensure it's slightly larger than what we need.
- [x] Set `width` and `height` of `<img>` to improve CLS.
- Addresses Ligthhouse complaints, although technically the shifting was addressed at the `ClaimPreviewTile` level (sub-container dimensions are well defined).
- Notes: the values don't need to be the final CSS-adjusted sizes. It just needs to be in the right aspect ratio to help the browser pre-allocate space to avoid shifts.
- [x] Add option to disable lazy-load Channel Thumbnails
- The guidelines mentioned that items that are already in the viewport should not enable `loading="lazy"`.
- We have a few areas where it doesn't make sense to lazy-load (e.g. thumbnail in Header, channel selector dropdown, publish preview, etc.).
While Lighthouse allows looser searching like "Tom from LBRY", it doesn't show the expected results when direct channel name with partial ID is entered to disambiguate.
