GitHub user Ma77Ball added a comment to the discussion: Proposal: Ditch Angular 
and Switch to a new frontend framework

Before we lock in React, I want to push back a bit. I think the framework 
should be driven by the *real-time* part of "real-time data science engine," 
not mainly by ecosystem size.

**R3 shouldn't be a flat "No" for Svelte or Solid.** The table treats the 
node-graph canvas as React-only, but `@xyflow/svelte` (Svelte Flow) is from the 
same xyflow team as React Flow and is actively developed. Solid has 
`solid-flow` as well. Both are less mature than React Flow today, but they're 
real libraries, so the canvas requirement isn't the clean disqualifier the 
table makes it look like.

**Performance and memory.** React's re-render model costs us most exactly where 
this app is busiest: streaming agent tokens, live console and result updates, 
operator status badges ticking during execution. Keeping that smooth in React 
means leaning hard on memoization and careful render boundaries. Solid 
(fine-grained signals) and Svelte (compile-time) skip VDOM diffing and touch 
only the DOM that actually changed. Solid in particular consistently leads the 
js-framework-benchmark results on both update speed and memory.

That memory angle matters more than it sounds for us. Texera is a large SPA 
that people keep open all day, and React holds a virtual DOM representation 
plus a heavier runtime in memory that grows with component count. On a 
long-running session with a big workflow and live data, that baseline footprint 
is real. At a time when RAM and hardware aren't cheap, pushing more memory onto 
every user's browser tab is a cost we'd be choosing to take on.

**Where React's case is genuinely strong: R2.** The AI component ecosystem 
(CopilotKit, assistant-ui, etc.) really is React-first, and that gap is real 
today. But @Xiao-zhen-Liu already noted the Yjs diff and sync logic is plain 
portable TypeScript, so the collaboration layer isn't a React advantage. It's a 
rewrite either way.

So my ask: since we're prototyping anyway, could we build the canvas prototype 
in Solid (or Svelte) alongside the React one? Same component, same features, 
with side-by-side line counts plus a memory and perf profile under live 
execution. If React Flow wins on maturity and the AI components, great, but 
let's measure it rather than rule the leaner options out on a table cell.


GitHub link: 
https://github.com/apache/texera/discussions/9387#discussioncomment-19662941

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to