Testing

Skill: databricks-app-appkit

What You Can Build

You can test AppKit apps at two levels: unit tests with Vitest for isolated logic and smoke tests with Playwright for verifying the app loads correctly in a browser. The AppKit template ships with both configured, but you need to update the default selectors after customizing your app.

In Action

“Using TypeScript, write a Vitest unit test for a data transformation function.”

import { describe, it, expect } from "vitest";

function formatRevenue(amount: number): string {
  return new Intl.NumberFormat("en-US", {
    style: "currency",
    currency: "USD",
  }).format(amount);
}

describe("formatRevenue", () => {
  it("formats a positive number as USD currency", () => {
    expect(formatRevenue(1234.5)).toBe("$1,234.50");
  });

  it("handles zero", () => {
    expect(formatRevenue(0)).toBe("$0.00");
  });
});

Key decisions:

• Place test files next to the code they test (e.g., src/utils.test.ts alongside src/utils.ts)
• Do not write tests for SQL files or auto-generated types — those are validated by the type generator and the SQL warehouse at runtime
• Focus unit tests on data transformations, formatters, and business logic functions
• Keep smoke tests simple: verify the app loads and renders initial content

More Patterns

Updating the default Playwright smoke test

“Using TypeScript, fix the template’s smoke test to match your app’s actual heading.”

import { test, expect } from "@playwright/test";

test("app loads and displays dashboard", async ({ page }) => {
  await page.goto("/");

  // Template default (will fail after you customize):
  // await expect(page.getByRole("heading", { name: "Minimal Databricks App" })).toBeVisible();

  // Update to match YOUR app's heading:
  await expect(
    page.getByRole("heading", { name: "Revenue Dashboard" })
  ).toBeVisible();
});

The template ships with a smoke test that checks for the default heading text. After you rename your app, the test fails. Update the selector to match your actual heading.

Handling Playwright strict mode

“Using TypeScript, write Playwright selectors that avoid strict mode violations.”

// Strict mode fails when a selector matches multiple elements.
// Use these in priority order:

// 1. Role + name (most reliable)
await expect(page.getByRole("heading", { name: "Dashboard" })).toBeVisible();

// 2. Role for interactive elements
await expect(page.getByRole("button", { name: "Submit" })).toBeVisible();

// 3. Exact text match
await expect(page.getByText("Total Revenue", { exact: true })).toBeVisible();

// 4. First match (last resort)
await expect(page.getByText("Loading").first()).toBeVisible();

Playwright’s strict mode throws when a selector matches more than one element. Start with getByRole — it is the most specific and resilient to layout changes.

Keeping smoke tests fast

“Using TypeScript, write a smoke test that validates initial data rendering without waiting for slow queries.”

test("data table renders within 5 seconds", async ({ page }) => {
  await page.goto("/");

  // Wait for the table to appear (SQL warehouse queries may take a moment)
  await expect(page.getByRole("table")).toBeVisible({ timeout: 5000 });

  // Verify at least one data row rendered
  const rows = page.getByRole("row");
  await expect(rows).not.toHaveCount(0);
});

Smoke tests should verify the app loads and displays initial data — not exercise every feature. Keep them under 5 seconds. If a query takes longer, the issue is in your SQL or warehouse configuration, not the test.

Watch Out For

• Leaving the default heading selector unchanged — the template’s smoke test checks for “Minimal Databricks App”. After you customize your app title, the test fails silently in CI. Update the selector immediately after scaffolding.
• Testing SQL query files or generated types — SQL files are validated by the SQL warehouse at runtime, and types are generated automatically. Writing unit tests for either adds maintenance cost without meaningful coverage.
• Broad Playwright selectors that match multiple elements — page.getByText("Submit") fails in strict mode if there are two submit buttons. Use getByRole with a name parameter for uniqueness.
• Setting long timeouts to mask slow queries — if your smoke test needs a 30-second timeout, the problem is the query, not the test. Investigate SQL performance or warehouse sizing before increasing timeouts.