Skip to content

test(bench): add reproducible framework comparisons - #57

Merged
Marcisbee merged 2 commits into
masterfrom
kode/FXd2WZgy-benchmark-radi-against-react-preact-vue-
Oct 22, 2026
Merged

Marcisbee merged 2 commits into
masterfrom
kode/FXd2WZgy-benchmark-radi-against-react-preact-vue-

Conversation

@Marcisbee

Copy link
Copy Markdown
Owner

Radi's existing benchmarks omit Vue and do not provide a consistent comparison across performance dimensions. Add an isolated, locked Node/Playwright suite comparing local Radi with React, Preact, Vue, and vanilla JavaScript in production Chromium bundles.

The command reports bundle sizes, startup, twelve rendering/update/keyed-list/event/cleanup workloads, frame latency, and post-GC heap measurements. It saves raw samples and a Markdown comparison with versions and host metadata. Every operation checks DOM contents, selection, order, and retained keyed identity before its sample is accepted. Documentation explains scheduler differences and limits of these synthetic measurements.

Validation: clean dependency installation, five-framework smoke run, full comparison with ten samples and three warmups per operation, and a deliberately broken selection adapter rejected by the validator. Radi's 43 existing suites passed all 304 tests on rerun; an initial run had one intermittent suspense timing assertion failure. No Radi runtime code changes.

@Marcisbee
Marcisbee marked this pull request as ready for review October 22, 2026 03:01
@Marcisbee Marcisbee changed the title Add reproducible browser framework comparison benchmarks test(bench): add reproducible framework comparisons Oct 22, 2026
@Marcisbee
Marcisbee merged commit 64562be into master Oct 22, 2026
3 checks passed
@Marcisbee
Marcisbee deleted the kode/FXd2WZgy-benchmark-radi-against-react-preact-vue- branch October 22, 2026 12:40
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant