New Test Ideas

[AlexanderKaran]

• Pre-rendered or static content runtimes tests
• SPA Mode. Frameworks that support a full SPA mode not just a route
• caching repeated navigations is also a big performance thing, it might be nice for the benchmark to include a repeated navigation somewhere to demonstrate and measure the reuse of cached data (from TanStack Team)
• Add a navigation-after-hover benchmark: hover for 100 ms, click, and measure time from click to destination content painted across all frameworks. (from TanStack Team)

[delucis]

It would be interesting to test hydrated components with server rendered markup in addition to the current tests for client rendering. Client rendering in React for example should be largely equivalent across different frameworks that support React, so the tests are likely mostly measuring React rather than the framework. The bigger differences may be where the frameworks provide unique takes on the server–client boundary to optimize performance (islands vs React Server Components etc.)

[dreyfus92]

nice list, two things i would add are:

• a cpu throttled variant, everything runs unthrottled desktop right now which hides js cost. something like 4x would show where hydration actually hurts
• hmr latency next to dev server startup, the time from a file edit until the browser shows it. startup is measured but the edit loop is what devs feel all day