I yanked out the parseURI part in a prior commit ... the comment was misleading me to think it was redundant. But it had another hidden function, which is to handle abandoned claims which `claim` will be `null`.
* apiCall: add option to not send the auth header
## Why
Want an option to make un-authenticated `resolve` calls where appropriate, to improve caching.
## How
All `apiCall`s are authenticated by default, but when clients add NO_AUTH to the params, `apiCall` will exclude the X_LBRY_AUTH_TOKEN. It will also strip NO_AUTH from the param object before sending it out.
* Add hook for 'resolve' and 'claim_search' to check and skip auth...
... if the params does not contain anything that requires the wallet.
* doResolveUri, doClaimSearch: let clients decide when to include_my_output
- No more hardcoding 'include_purchase_receipt' and 'include_is_my_output'
- doResolveUri: include these params when opening a file page. This was the only place that was doing that prior to this PR.
* is_my_output: use the signing_channel as alternative
## Notes
`is_my_output` is more expensive to resolve, so it is not being requested all the time.
## Change
Looking at the signing channel as the additional fallback, on top of `myClaimIds`.
## Aside
I think using `myClaimIds` here is redundant, as it is usually populated from `is_my_ouput`. But leaving as is for now...
## Why
- No memo required (no transformation).
- `makeSelect*` is an incorrect pattern.
## Changes
- Replaced makeSelectClientSetting with selectClientSetting.
- Remove unused selectShowRepostedContent.
## Issue
One of the bottlenecks of livestream page.
The component probably needs a re-design:
- Don't perpetually mount -- only mount when activated by the user through "@". This would avoid the heavy processing entirely.
- Better way of resolving uris (too many arrays, too many loops).
- Tom also mentioned that we should not be resolving every commenter as we see encounter them in a livestream. This is currently the case because the component is always mounted.
## Changes
Until the re-design occurs, attempt to cache the heavy processing. Also, trimmed down the amount of loops.
For the case of livestreams, the comments are added incrementally via websocket. The selector returns everything, which grows as a user watches the livestream.
We could even make it a bit more efficient by passing in `maxCount` to `filterComments`, and do a `for` loop there, but decided to keep things readable by not changing the `filter` usage.
- Memo not required. `resolvingUris` is very dynamic and is a short array anyways.
- Changeg from using `indexOf` to `includes`, which is more concise.
* 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
## Issue
Too slow for huge lists
## Change
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.
## Issues with `makeSelectIsSubscribed`
- 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.
## Changes
- Optimize the selector to only look up the claim once, and make operations using already-obtained info.
## Why
Frequently used; top in perf profile
## Changes
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.
* Publish button: use spinner instead of "Publishing..."
Looks better, plus the preview could take a while sometimes.
* Refactor `doPublish`. No functional change
This is to allow `doPublish` to accept a custom payload as an input (for resuming uploads), instead of always resolving it from the redux data.
* Add doPublishResume
* Support resume-able upload via tus
## Issue
38 Handle resumable file upload
## Notes
Since we can't serialize a File object, we'll need to the user to re-select the file to resume.
* Exclude "modified date" for Firefox/Android
## Issue
It appears that the modification date of the Android file changes when selected, so that file was deemed "different" when trying to resume upload.
## Change
Exclude modification date for now. Let's assume a smart user.
* Move 'currentUploads' to 'publish' reducer
`publish` is currently rehydrated, so we can ride on that and don't need to store the `currentUploads` in `localStorage` for persistence. This would allow us to store Markdown Post data too, as `localStorage` has a 5MB limit per app.
We could have also made `webReducer` rehydrate, but in this repo, there is no need to split it to another reducer. It also makes more sense to be part of publish anyway (at least to me).
This change is mostly moving items between files, with the exception of
1. An additional REHYDRATE in the publish reducer to clean up the tusUploader.
2. Not clearing `currentUploads` in CLEAR_PUBLISH.
* Restore v1 code for livestream replay, etc.
v2 (tus) does not handle `remote_url`, so the app still needs v1 for that. Since we'll still have v1 code, use v1 for previews as well.
- memo not required.
- start to not use the confusing and wrongly-named 'selectCommentsByUri' (per comment from Sean); use the existing 'selectClaimIdForUri' instead. This works because currently we do fetch any comments without first visiting a claim/uri, so we'll always have fetched the required claim, and can be queried in 'selectClaimIdForUri'.
## Issue
`normalizeUri` | `parseURI` is expensive and has been causing sluggish operations when called repeatedly or within a loop.
## Change
Since I'm not confident enough to remove the call entirely from makeSelectClaimIsMine (although I've yet to find a scenario that the uri is not already normalized), we'll try caching the calls instead.
## Results
- in a simple test of toggling between 2 category pages, we saved 20ms from `parseURI` calls alone.
- in a test of opening all categories one time, the memory usage remained similar. This makes sense since we removed a `makeSelect*` (which creates a selector for each call + not memoizing), and replaced that with a cached selector that's actually memoizing.
## Issue
- selectMyActiveClaims memo problem -- being recalculated on every click -- high workload for wallet with large uploads.
- Mistake in handling abandoning IDs (it was trying to extract keys from an array)
## Changes
- selectAbandoningIds: never use `state` as an input selector. Breaks memo.
- Don't use selectMyClaimsRaw and then reduce it back to IDs. Use selectMyClaimIdsRaw instead.
- selectAbandoningIds is already in array form, so don't run Object.keys.
- Fix abandoningById never clearing when succeeded.
* Don't update 'pendingById' if no changes.
'pendingById' isn't frequently updated, but using it as a proof-of-concept to fix how reducers should be written to avoid unnecessary updates.
ImmutableJS apparently does all of this for us, but there are cons to using it as well, so using own wrappers for now.
* Don't update 'byId' if no changes + add 'selectClaimWithId'
## Ticket
116 Claim store optimization ideas (reducing unnecessary renders)
## Changes
- Ignore things like `confirmations` so that already-fetched claims aren't invalidated and causes re-rendering. The `stringify` might look expensive, but the amount of avoided re-renders outweighs it. There might be faster ways to compare, though.
- With `byId[claimId]` references more stable now, memoized selectors can now use 'selectClaimWithId' to pick a specific claim to depend on, instead of 'byId' which changes on every update.
* Fix memo: selectMyChannelClaims, selectActiveChannelClaim
## Issue
These should never recalculate after `channel_list` has been fetched, but they do because of poor selector dependency.
## Change
With the `byId` changes from the previous commit, we are now able to memoize these selectors correctly.
- selectMyChannelClaims depends on `byId`, which currently is always invalidated per update, so it is not memoized.
- Most of the use-cases just needs the ID or the length of the array anyways, so avoid generating a Claim array (in selectMyChannelClaims) unnecessarily -- the client need to reduce it back down to IDs again :/
- The simpler boolean also removes the need to memoize the selector, which saves a bit of memory.
`selectMyActiveClaims` includes `byId`, which gets invalidated on each resolve. Having this as an input selector breaks memoization.
For the case of comment-filtering, we don't really care about pending or abandoned own claims (I think), so just grab the raw IDs.
* Prevent multiple parseURI calls
## Ticket
129
## Issue
Code was shortened to use `isURIValid` during the consolidation. `isURIValid` calls `normalizeURI`, which calls another `parseURI`.
`parseURI` is pretty expensive.
## Approach
- Add optional parameter to `isURIValid` to skip the normalization.
- Set those that were converted during the consolidation to skip the normalization. Also covered a few other instances where it is obvious to me that normalization is not required.
- For the rest, I can't tell for sure if it's safe to remove the normalization, so the default `normalize=true` will leave things as is.
The whole `parseURI` probably needs a refactoring, or a few lighter version for specific needs.
* Simplify isURIEqual
## Issue
`parseURI` is too expensive to be used in a loop, plus `normalizeURI` itself is calling `parseURI`.
## Approach
Not sure if it covers all cases, but just try convert colons to hashes before comparing.
## Issue
We previously automatically reload when there is a chunk error. This works fine if it's the case of new code was pushed recently while the user was active. But if the failure was caused by other things like network problems or the file IS actually missing, we end up in an infinite loop of refreshes.
## New approach
Tell the user to reload instead of automatically doing it.
followedTags:
- Moved the filtering to the reducer side, so that we don't do it every time. We can't rely on `createSelector` because the store will be invalidated on each `USER_STATE_POPULATE`, unfortunately.
tags:
- Memoize via re-reselect for the "ForUri" selector.
## Issue
`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.
## Changes
- 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)
## Issue
When comments are refreshed, each `Comment` gets rendered 4-5 times due to reference invalidation for `othersReacts` (the data didn't actually change).
## Change
For selectors without transformation, there is no need to memoize using `createSelector` -- just access it directly. Also, don't do things like `return a[id] || {}` in a reducer, because the reference to the empty object will be different on each call.
Always return directly from the state so that the same reference is returned.
This simple change avoided the wasted resources needed for `createSelector`, and reduced to render to just 2 (initial render, and when reactions are fetched).
## Issue
Components render unnecessarily due to reference invalidation from `selectMutedChannels` selector.
## Notes
`selectMutedChannels` run and return a new reference each time the app gains focus. `createSelector` will not help in this case, because we are indeed invalidating the data in the store in `USER_STATE_POPULATE`.
## Changes
- Don't update the state if the array is identical in content.
- Fixed `selectMutedChannels` to return the reference from the store, so `createSelector` is not needed.
- Also, the filtering is not needed because we've already done it in the reducer.
## Comments
I've done some profiling on large blocklists. The time needed for the array comparison is still an order magnitude lower than the time needed to render all the Components that got incorrectly marked by this.
The ideal solution is for the sync code to return a hash or timestamp of the array, so that we can compare that instead of the array.
I think this the best solution so far, at the expense of a slight delay in scrolling if the network call stalls.
- Added "fetching by ID" state so that we don't need to use the ugly N-retries method.
- `scrollIntoView` doesn't work if the element is already in the viewport, and the `scrollBy` adjustment doesn't take into account the y-position restoration that we perform on certain type of pages. Use `window.scrollTo` instead and taking into account current scroll position.
## Issue
.../archives/C02FQBM00Q0/p1633044695010600
## Changes
When querying a search key, it has to be an exact match. This was broken by the insertion of `free_only` in the fetch.
Added a function to generate the options, so that all clients stay in sync.
It was being recalculated repeatedly.
This memoizes it, although it still re-calculates occasionally despite none of the source arrays changed. I think it is due to the state change in the Preference Sync.
Note: input selectors to `createSelector` needs to be extractions-only (i.e. must not have transformations). I think most of our `makeSelect*` selectors violate this and broke memoization.
