## The Design System Testing Gap Your design system has a Storybook with 200+ component stories. Every component has unit tests for its props and behavior. But does anyone actually verify that the `Button` component with `variant="destructive"` and `size="sm"` renders with the correct red background and 32px height? This is the gap visual testing fills. It validates the visual output of every component, in every variant, at every breakpoint. ## The Testing Matrix For a typical design system button component, the testing matrix looks like this: | Dimension | Values | |-----------|--------| | Variant | default, secondary, destructive, outline, ghost, link | | Size | sm, md, lg | | State | default, hover, focus, disabled, loading | | Content | text, icon, text+icon | | Theme | light, dark | That is 6 x 3 x 5 x 3 x 2 = **540 visual states** for a single component. Multiply by the dozens of components in your system, and you quickly understand why manual visual QA is impossible. ## Automating With Storybook Storybook is the natural integration point for design system visual testing. Each story represents a visual state: ```tsx // Button.stories.tsx export const AllVariants: Story = { render: () => (

{(['default', 'secondary', 'destructive', 'outline'] as const).map( (variant) => (

) )}

), } ``` Then use a visual testing tool to capture and compare screenshots of every story on every PR. ## Handling Theme Variants Design systems typically support light and dark themes. Your visual tests need to cover both: ```typescript const themes = ['light', 'dark'] as const for (const theme of themes) { test(`button renders in ${theme} mode`, async ({ page }) => { await page.goto(`/iframe.html?id=button--all-variants`) await page.evaluate((t) => { document.documentElement.classList.toggle('dark', t === 'dark') }, theme) await expect(page).toHaveScreenshot(`button-${theme}.png`) }) } ``` ## Token Validation Beyond visual comparison, you can validate that components use the correct design tokens: ```typescript test('button uses correct design tokens', async ({ page }) => { await page.goto('/iframe.html?id=button--default') const button = page.locator('button') const styles = await button.evaluate((el) => { const computed = getComputedStyle(el) return { backgroundColor: computed.backgroundColor, borderRadius: computed.borderRadius, fontWeight: computed.fontWeight, padding: computed.padding, } }) expect(styles.borderRadius).toBe('8px') expect(styles.fontWeight).toBe('500') }) ``` This catches CSS drift at the token level, not just the visual level. ## Responsive Component Testing Components need to work at every viewport size. For a card component that switches from horizontal to vertical layout at mobile: ```typescript test('card switches layout at mobile', async ({ page }) => { // Desktop - horizontal layout await page.setViewportSize({ width: 1024, height: 768 }) await page.goto('/iframe.html?id=card--responsive') await expect(page.locator('.card')).toHaveScreenshot('card-desktop.png') // Mobile - vertical layout await page.setViewportSize({ width: 375, height: 812 }) await expect(page.locator('.card')).toHaveScreenshot('card-mobile.png') }) ``` ## The Approval Workflow When visual tests detect changes, someone needs to decide whether the change is intentional. For design systems, this workflow should involve: 1. **Engineering review** for code correctness 2. **Design review** for visual correctness 3. **Automated checks** for accessibility (contrast ratios, text sizes) Tools like Chromatic integrate this workflow directly into your PR process, showing side-by-side comparisons with one-click approval. ## Measuring Coverage Track which components have visual tests and which do not: ```bash # Compare stories with visual test coverage total_stories=$(find src -name "*.stories.tsx" | wc -l) tested_stories=$(grep -rl "toHaveScreenshot" tests/ | wc -l) echo "Visual test coverage: $tested_stories/$total_stories" ``` Aim for 100% coverage of your core components (buttons, inputs, cards, navigation) and at least 80% coverage of composite components. ## Continue with ScanU If you want to apply these techniques in production, start with a focused set of pages and run baseline screenshot comparison after every meaningful UI change. You can review plans on [Pricing](/pricing), implementation details in the [FAQ](/faq), and product capabilities on [Features](/features).

洞察 与工程

洞察与工程