repl(options?: UserRunOptions): Task<ReplSession, ReplFailure>
Opens a REPL against a live browser page and hands back the session.
The verb the CLI's qunitx repl is built on, and useful on its own wherever a page needs to be
asked questions rather than tested: a scripted browser probe, a documentation example that must
really run, a test for something that only exists in a real DOM.
inputs are preloaded modules, not a test target — each one's exports land on the page's
globalThis, and any tests it registers run once as the session opens. Nothing else is loaded:
unlike run(), package.json#qunitx.inputs is not a default here.
Lazy, like every verb in this API: nothing launches until the Task is awaited. Nothing
is printed either, unless a reporter is asked for — tests typed at the prompt are reported
through exactly the same reporters a run uses.
import { repl } from './repl.ts'; // Defined, not invoked: launches a browser. async function measure() { await using session = await repl({ inputs: ['test/helpers.ts'] }); const { output } = await session.eval('document.querySelector("#qunit-fixture").tagName'); return output; // "'DIV'" }
options: UserRunOptions
repl(preload: string | string[],options?: UserRunOptions): Task<ReplSession, ReplFailure>
The same session, with the preload named positionally: repl('test/helpers.ts', { port: 4000 }).
The shape every other verb takes, offered here so they read alike. The positional preload wins
over any inputs in the options object.
options: UserRunOptions