Learn / Errors and failures

Testing failures

A failure path is a path like any other: drive it, and assert what the reader sees.

Without a browser, Trace.run from foldkit-remote/testing runs the real update and Subscriptions and holds every request; each Respond takes { fail } to answer with an error. Trace.regressions names every step where a read that showed a value stopped showing it.

import { RemoteReadError } from 'foldkit-remote'
import { Respond, Trace } from 'foldkit-remote/testing'

const trace = await Trace.run(Data, {
  start: loaded, // a Model showing the project
  update,
  active: actives,
  client: server, // a fake RemoteClient, or the server's in-process one
  steps: [
    Message.ClickedTryAgain(),
    Respond.read(Project, 'p1', { fail: new RemoteReadError({ message: 'The server is down' }) }),
  ],
})
expect(Trace.regressions(trace, { project: Data.get(ProjectName, 'p1') })).toEqual([])

A refusal made on the client is held to the server's by Law.check (foldkit-remote/testing): it runs one declared write through Data.mutate and through the server, and compares what the client predicted, a refusal included, with what the server then holds. See Testing that a prediction is the truth.

Diagnostics are asserted with Diagnostics.collect(): provide its report to the Effect under test (here program), then assert on reported, [] included.

const { report, reported } = Diagnostics.collect()
await Effect.runPromise(Effect.provideService(program, Diagnostics.Reporter, report))
expect(reported).toEqual([])

On a running page, foldkit-plus/testing/playwright's Behaviors watches the page as drawn:

import { Behaviors } from 'foldkit-plus/testing/playwright'

const errors = Behaviors.consoleErrors(page) // console.error and uncaught exceptions, from now on
await Behaviors.neverShows(page, {
  text: 'Could not load',
  during: async () => {
    await page.reload()
    await page.getByRole('heading').waitFor()
  },
})
expect(errors()).toEqual([])

consoleErrors lists every error diagnostic the default reporter logs, so this test also fails on a MountFailed or a PageContained. A warning is logged with console.log and is not listed: to fail on warnings too, give the page a reporter that calls console.error for them.

Where to read more#

Where to read more