Formatter: Preserve Frontmatter when formatting documents

javascript/packages/formatter/src/format-helpers.ts

@@ -521,3 +521,14 @@ export function isHerbDisableComment(node: Node): boolean {
 
   return trimmed.startsWith("herb:disable")
 }
+
+/**
+ * Check if a text node is YAML frontmatter (starts and ends with ---)
+ */
+export function isFrontmatter(node: Node): node is HTMLTextNode {
+  if (!isNode(node, HTMLTextNode)) return false
+
+  const content = node.content.trim()
+
+  return content.startsWith("---") && /---\s*$/.test(content)
+}

javascript/packages/formatter/src/format-printer.ts

@@ -33,6 +33,7 @@ import {
   isBlockLevelNode,
   isClosingPunctuation,
   isContentPreserving,
+  isFrontmatter,
   isHerbDisableComment,
   isInlineElement,
   isNonWhitespaceNode,
@@ -641,15 +642,16 @@ export class FormatPrinter extends Printer {
   // --- Visitor methods ---
 
   visitDocumentNode(node: DocumentNode) {
-    const hasTextFlow = this.isInTextFlowContext(null, node.children)
+    const children = this.formatFrontmatter(node)
+    const hasTextFlow = this.isInTextFlowContext(null, children)
 
     if (hasTextFlow) {
-      const children = filterSignificantChildren(node.children)
+      const filteredChildren = filterSignificantChildren(children)
 
       const wasInlineMode = this.inlineMode
       this.inlineMode = true
 
-      this.visitTextFlowChildren(children)
+      this.visitTextFlowChildren(filteredChildren)
 
       this.inlineMode = wasInlineMode
 
@@ -659,10 +661,10 @@ export class FormatPrinter extends Printer {
     let lastWasMeaningful = false
     let hasHandledSpacing = false
 
-    for (let i = 0; i < node.children.length; i++) {
-      const child = node.children[i]
+    for (let i = 0; i < children.length; i++) {
+      const child = children[i]
 
-      if (shouldPreserveUserSpacing(child, node.children, i)) {
+      if (shouldPreserveUserSpacing(child, children, i)) {
         this.push("")
         hasHandledSpacing = true
         continue
@@ -672,8 +674,8 @@ export class FormatPrinter extends Printer {
         continue
       }
 
-      if (shouldAppendToLastLine(child, node.children, i)) {
-        this.appendChildToLastLine(child, node.children, i)
+      if (shouldAppendToLastLine(child, children, i)) {
+        this.appendChildToLastLine(child, children, i)
         lastWasMeaningful = true
         hasHandledSpacing = false
         continue
@@ -1473,6 +1475,21 @@ export class FormatPrinter extends Printer {
 
   // --- Utility methods ---
 
+  private formatFrontmatter(node: DocumentNode): Node[] {
+    const firstChild = node.children[0]
+    const hasFrontmatter = firstChild && isFrontmatter(firstChild)
+
+    if (!hasFrontmatter) return node.children
+
+    this.push(firstChild.content.trimEnd())
+
+    const remaining = node.children.slice(1)
+
+    if (remaining.length > 0) this.push("")
+
+    return remaining
+  }
+
   /**
    * Append a child node to the last output line
    */

javascript/packages/formatter/test/erb-formatter/erb-formatter-fixtures.test.ts

@@ -590,7 +590,10 @@ describe("ERB Formatter Fixture Tests", () => {
       const result = formatter.format(source)
 
       expect(result).toBe(dedent`
-        --- title: "My Page" layout: "application" ---
+        ---
+        title: "My Page"
+        layout: "application"
+        ---
 
         <div class="page">
           <h1><%= @title || "Default Title" %></h1>

javascript/packages/formatter/test/frontmatter.test.ts

@@ -0,0 +1,229 @@
+import { describe, test, expect, beforeAll } from "vitest"
+import { Herb } from "@herb-tools/node-wasm"
+import { Formatter } from "../src"
+
+import dedent from "dedent"
+
+let formatter: Formatter
+
+describe("@herb-tools/formatter", () => {
+  beforeAll(async () => {
+    await Herb.load()
+
+    formatter = new Formatter(Herb, {
+      indentWidth: 2,
+      maxLineLength: 80
+    })
+  })
+
+  test("preserves YAML frontmatter with no formatting", () => {
+    const source = dedent`
+      ---
+      title: My Page
+      layout: application
+      published: true
+      ---
+
+      <div class="container">
+        <h1><%= @title %></h1>
+      </div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("preserves frontmatter with ERB content after it", () => {
+    const source = dedent`
+      ---
+      title: Test
+      ---
+
+      <% if Rails.env.development? %>
+        <p>Debug mode</p>
+      <% end %>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("preserves frontmatter indentation as-is", () => {
+    const source = dedent`
+      ---
+      nested:
+        key: value
+        another:
+          deep: true
+      ---
+
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("normalizes whitespace after frontmatter", () => {
+    const source = dedent`
+      ---
+      title: Test
+      ---
+
+
+
+
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(dedent`
+      ---
+      title: Test
+      ---
+
+      <div>Content</div>
+    `)
+  })
+
+  test("handles frontmatter with arrays and objects", () => {
+    const source = dedent`
+      ---
+      tags:
+        - ruby
+        - rails
+        - erb
+      metadata:
+        author: John Doe
+        date: 2024-01-01
+      ---
+
+      <article>
+        <h1>Title</h1>
+      </article>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("formats HTML but preserves frontmatter when HTML is messy", () => {
+    const source = dedent`
+      ---
+      title: Test
+      ---
+
+      <div    class="container"   >
+          <h1>Title</h1>
+            <p>Text</p>
+      </div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(dedent`
+      ---
+      title: Test
+      ---
+
+      <div class="container">
+        <h1>Title</h1>
+        <p>Text</p>
+      </div>
+    `)
+  })
+
+  test("does not treat --- in the middle of document as frontmatter", () => {
+    const source = dedent`
+      <div>
+        <p>Some content</p>
+        ---
+        <p>More content</p>
+      </div>
+    `
+    const result = formatter.format(source)
+
+    expect(result).toEqual(source)
+  })
+
+  test("frontmatter must end with --- on its own line", () => {
+    const source = dedent`
+      ---
+      title: Test --- not closing
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+
+    expect(result).toEqual(dedent`
+      --- title: Test --- not closing
+
+      <div>Content</div>
+    `)
+  })
+
+  test("empty frontmatter block", () => {
+    const source = dedent`
+      ---
+      ---
+
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("frontmatter with comments", () => {
+    const source = dedent`
+      ---
+      # This is a YAML comment
+      title: My Page
+      # Another comment
+      layout: application
+      ---
+
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(source)
+  })
+
+  test("frontmatter adds newline", () => {
+    const source = dedent`
+      ---
+      title: My Page
+      ---
+      <div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(dedent`
+      ---
+      title: My Page
+      ---
+
+      <div>Content</div>
+    `)
+  })
+
+  test("frontmatter with no newline after ---", () => {
+    const source = dedent`
+      ---
+      title: My Page
+      ---<div>Content</div>
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(dedent`
+      ---
+      title: My Page
+      ---
+
+      <div>Content</div>
+    `)
+  })
+
+  // TODO: maybe we can improve this in the future
+  test("frontmatter with text after ---", () => {
+    const source = dedent`
+      ---
+      title: My Page
+      ---
+      Content
+    `
+    const result = formatter.format(source)
+    expect(result).toEqual(dedent`
+      --- title: My Page --- Content
+    `)
+  })
+})