-
Notifications
You must be signed in to change notification settings - Fork 127
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Assign view id paths by parents #9
Conversation
I'm still wrapping my head around this one, and am not yet really seeing the advantage, but I want to fully understand it before making a decision. One observation: a reason to make the id mutable in In any case, thanks for these. I do want to explore the neighborhood of the original Xilem draft and iterate toward things that are actually improvements. |
|
I appreciate some of the reduced boilerplate but this seems to trade a simple, common |
It seems like the use cases for ids have goals that pull in conflicting directions: you want ids to be static for the life of each component, and also support dynamic allocation under heavy component construction-destruction load. Maybe something like a generational index 1 2 would be useful here? 3 4 Footnotes |
@dfrg Dynamic containers could just use the existing strategy of a global It's possible a generational index would be an optimization that dynamic containers could use for better event dispatch. |
I've got my head around this - thanks again for the proposal and discussion. It is a viable alternative way to organize id paths. However, I'm going to close it, as I think it's a sideways move. Yes, the signatures of build and rebuild become smaller, and the dispatch of events by a tuple to a child is more direct, but I'm also seeing the downsides, which to me outweigh the benefits: async resolution is by the Just a suggestion, but a blog post on the design neighborhood of the Xilem architecture could make a good read - this proposal as well as the one that's generic over widget trait so can support both native and web from the same app logic. |
A major motivation is to optimize event dispatch, which is the mechanism used by async resolution. It relies on
I'd expect |
Alternative to #218 The version of viewsequence added in #205 is robust for all properties supported in that version. However, it would be incompatible with future expansions to the Xilem model, in particular async wiring. In this PR, I propose a system to resolve this, which uses generational indexes in the view id path (for the items where this is relevant). Some other historical record: #9 This brings back the `View::State` associated type (renamed to `ViewState` and `SeqState` depending on the trait), in order to perform this wiring.
New tracy image:  New log tracing file: <details><summary>An overview of the new logs</summary> <p> ``` 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}: masonry::passes::update: RootWidget received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}: masonry::passes::update: SizedBox received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}:Flex{id=#3}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}:Flex{id=#3}:Prose{id=#1}: masonry::passes::update: Prose received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}:Flex{id=#3}:Label{id=#2}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}:Flex{id=#5}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#8}:SizedBox{id=#7}:Flex{id=#6}:Flex{id=#5}:VariableLabel{id=#4}: masonry::passes::update: VariableLabel received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#10}: masonry::passes::update: Button received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#10}:Label{id=#9}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#12}: masonry::passes::update: Button received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#12}:Label{id=#11}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#14}: masonry::passes::update: Button received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#14}:Label{id=#13}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#16}: masonry::passes::update: Button received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Flex{id=#17}:Button{id=#16}:Label{id=#15}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}: masonry::passes::update: Portal received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}: masonry::passes::update: SizedBox received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}:Flex{id=#20}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}:Flex{id=#20}:Prose{id=#18}: masonry::passes::update: Prose received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}:Flex{id=#20}:Label{id=#19}: masonry::passes::update: Label received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}:Flex{id=#22}: masonry::passes::update: Flex received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#24}:Flex{id=#23}:Flex{id=#22}:VariableLabel{id=#21}: masonry::passes::update: VariableLabel received Update::WidgetAdded 14:37:40.365Z TRACE update_new_widgets:RootWidget{id=#165}:Flex{id=#164}:Portal{id=#163}:Flex{id=#160}:SizedBox{id=#31}: masonry::passes::update: SizedBox received Update::WidgetAdded ``` </p> </details> This was originally an experiment into caching spans, but I determined that was non-viable due to the pass names.
This changes
Id
allocation from using a global unique id per view to letting containers pick uniqueId
segments for their children. TheIdPath
when thebuild
method on a view is called uniquely identifies an view. This results in slightly cleaner code. Views likeVStack
can process event dispatch with a jump table by hardcoding children ids instead of using an if/else tree.A further optimization would be to change
IdPath
toVec<u8>
and let containers only use the number of bytes required, reducingIdPath
sizes overall.