From e5afca6c1e76bc5757846ee2c4698e47e66a8697 Mon Sep 17 00:00:00 2001
From: Evan You <yyx990803@gmail.com>
Date: Mon, 20 Nov 2023 22:49:16 +0800
Subject: [PATCH] wip: treat template with preprocessor as plain text

---
 .../compiler-core/src/parser/Tokenizer.ts     | 47 ++++++++++---------
 packages/compiler-core/src/parser/index.ts    | 19 ++++++--
 packages/compiler-sfc/__tests__/parse.spec.ts |  6 ++-
 packages/compiler-sfc/src/compileTemplate.ts  |  3 +-
 packages/compiler-sfc/src/parse.ts            |  6 ++-
 5 files changed, 48 insertions(+), 33 deletions(-)

diff --git a/packages/compiler-core/src/parser/Tokenizer.ts b/packages/compiler-core/src/parser/Tokenizer.ts
index 3fe2656a7..c0767c1ef 100644
--- a/packages/compiler-core/src/parser/Tokenizer.ts
+++ b/packages/compiler-core/src/parser/Tokenizer.ts
@@ -129,7 +129,7 @@ const enum State {
   BeforeSpecialS, // Decide if we deal with `<script` or `<style`
   BeforeSpecialT, // Decide if we deal with `<title` or `<textarea`
   SpecialStartSequence,
-  InSpecialTag,
+  InRCDATA,
 
   InEntity,
 
@@ -245,6 +245,11 @@ export default class Tokenizer {
 
   private readonly entityDecoder?: EntityDecoder
 
+  public mode = ParseMode.BASE
+  public get inSFCRoot() {
+    return this.mode === ParseMode.SFC && this.stack.length === 0
+  }
+
   constructor(
     private readonly stack: ElementNode[],
     private readonly cbs: Callbacks
@@ -256,8 +261,6 @@ export default class Tokenizer {
     }
   }
 
-  public mode = ParseMode.BASE
-
   public reset(): void {
     this.state = State.Text
     this.mode = ParseMode.BASE
@@ -328,8 +331,8 @@ export default class Tokenizer {
         this.delimiterIndex++
       }
     } else if (this.inRCDATA) {
-      this.state = State.InSpecialTag
-      this.stateInSpecialTag(c)
+      this.state = State.InRCDATA
+      this.stateInRCDATA(c)
     } else {
       this.state = State.Text
       this.stateText(c)
@@ -349,7 +352,7 @@ export default class Tokenizer {
       if (this.delimiterIndex === this.delimiterClose.length - 1) {
         this.cbs.oninterpolation(this.sectionStart, this.index + 1)
         if (this.inRCDATA) {
-          this.state = State.InSpecialTag
+          this.state = State.InRCDATA
         } else {
           this.state = State.Text
         }
@@ -386,7 +389,7 @@ export default class Tokenizer {
   }
 
   /** Look for an end tag. For <title> and <textarea>, also decode entities. */
-  private stateInSpecialTag(c: number): void {
+  private stateInRCDATA(c: number): void {
     if (this.sequenceIndex === this.currentSequence.length) {
       if (c === CharCodes.Gt || isWhitespace(c)) {
         const endOfText = this.index - this.currentSequence.length
@@ -413,8 +416,7 @@ export default class Tokenizer {
     } else if (this.sequenceIndex === 0) {
       if (
         this.currentSequence === Sequences.TitleEnd ||
-        (this.currentSequence === Sequences.TextareaEnd &&
-          !(this.mode === ParseMode.SFC && this.stack.length === 0))
+        (this.currentSequence === Sequences.TextareaEnd && !this.inSFCRoot)
       ) {
         // We have to parse entities in <title> and <textarea> tags.
         if (!__BROWSER__ && c === CharCodes.Amp) {
@@ -507,10 +509,14 @@ export default class Tokenizer {
   }
 
   private startSpecial(sequence: Uint8Array, offset: number) {
+    this.enterRCDATA(sequence, offset)
+    this.state = State.SpecialStartSequence
+  }
+
+  public enterRCDATA(sequence: Uint8Array, offset: number) {
     this.inRCDATA = true
     this.currentSequence = sequence
     this.sequenceIndex = offset
-    this.state = State.SpecialStartSequence
   }
 
   private stateBeforeTagName(c: number): void {
@@ -525,7 +531,7 @@ export default class Tokenizer {
       if (this.mode === ParseMode.BASE) {
         // no special tags in base mode
         this.state = State.InTagName
-      } else if (this.mode === ParseMode.SFC && this.stack.length === 0) {
+      } else if (this.inSFCRoot) {
         // SFC mode + root level
         // - everything except <template> is RAWTEXT
         // - <template> with lang other than html is also RAWTEXT
@@ -560,8 +566,7 @@ export default class Tokenizer {
     if (isEndOfTagSection(c)) {
       const tag = this.buffer.slice(this.sectionStart, this.index)
       if (tag !== 'template') {
-        this.inRCDATA = true
-        this.currentSequence = toCharCodes(`</` + tag)
+        this.enterRCDATA(toCharCodes(`</` + tag), 0)
       }
       this.handleTagName(c)
     }
@@ -603,8 +608,7 @@ export default class Tokenizer {
     if (c === CharCodes.Gt) {
       this.cbs.onopentagend(this.index)
       if (this.inRCDATA) {
-        this.state = State.InSpecialTag
-        this.sequenceIndex = 0
+        this.state = State.InRCDATA
       } else {
         this.state = State.Text
       }
@@ -827,7 +831,7 @@ export default class Tokenizer {
       this.state = State.InEntity
       this.entityStart = this.index
       this.entityDecoder!.startEntity(
-        this.baseState === State.Text || this.baseState === State.InSpecialTag
+        this.baseState === State.Text || this.baseState === State.InRCDATA
           ? DecodingMode.Legacy
           : DecodingMode.Attribute
       )
@@ -885,8 +889,8 @@ export default class Tokenizer {
           this.stateSpecialStartSequence(c)
           break
         }
-        case State.InSpecialTag: {
-          this.stateInSpecialTag(c)
+        case State.InRCDATA: {
+          this.stateInRCDATA(c)
           break
         }
         case State.CDATASequence: {
@@ -1016,7 +1020,7 @@ export default class Tokenizer {
     if (this.sectionStart !== this.index) {
       if (
         this.state === State.Text ||
-        (this.state === State.InSpecialTag && this.sequenceIndex === 0)
+        (this.state === State.InRCDATA && this.sequenceIndex === 0)
       ) {
         this.cbs.ontext(this.sectionStart, this.index)
         this.sectionStart = this.index
@@ -1083,10 +1087,7 @@ export default class Tokenizer {
 
   private emitCodePoint(cp: number, consumed: number): void {
     if (!__BROWSER__) {
-      if (
-        this.baseState !== State.Text &&
-        this.baseState !== State.InSpecialTag
-      ) {
+      if (this.baseState !== State.Text && this.baseState !== State.InRCDATA) {
         if (this.sectionStart < this.entityStart) {
           this.cbs.onattribdata(this.sectionStart, this.entityStart)
         }
diff --git a/packages/compiler-core/src/parser/index.ts b/packages/compiler-core/src/parser/index.ts
index 8bdcfb38e..99c7e0879 100644
--- a/packages/compiler-core/src/parser/index.ts
+++ b/packages/compiler-core/src/parser/index.ts
@@ -111,10 +111,9 @@ const tokenizer = new Tokenizer(stack, {
   onopentagname(start, end) {
     const name = getSlice(start, end)
     // in SFC mode, root-level tags locations are for its inner content.
-    const startIndex =
-      tokenizer.mode === ParseMode.SFC && stack.length === 0
-        ? end + fastForward(end, CharCodes.Gt) + 1
-        : start - 1
+    const startIndex = tokenizer.inSFCRoot
+      ? end + fastForward(end, CharCodes.Gt) + 1
+      : start - 1
     currentElement = {
       type: NodeTypes.ELEMENT,
       tag: name,
@@ -296,6 +295,16 @@ const tokenizer = new Tokenizer(stack, {
                 ? getLoc(currentAttrStartIndex, currentAttrEndIndex)
                 : getLoc(currentAttrStartIndex - 1, currentAttrEndIndex + 1)
           }
+          if (
+            currentAttrValue &&
+            tokenizer.inSFCRoot &&
+            currentElement.tag === 'template' &&
+            currentProp.name === 'lang'
+          ) {
+            // SFC root template with preprocessor lang, force tokenizer to
+            // RCDATA mode
+            tokenizer.enterRCDATA(toCharCodes(`</template`), 0)
+          }
         } else {
           // directive
           currentProp.rawExp = currentAttrValue
@@ -464,7 +473,7 @@ function onText(content: string, start: number, end: number) {
 
 function onCloseTag(el: ElementNode, end: number) {
   // attach end position
-  if (tokenizer.mode === ParseMode.SFC && stack.length === 0) {
+  if (tokenizer.inSFCRoot) {
     // SFC root tag, end position should be inner end
     if (el.children.length) {
       el.loc.end = extend({}, el.children[el.children.length - 1].loc.end)
diff --git a/packages/compiler-sfc/__tests__/parse.spec.ts b/packages/compiler-sfc/__tests__/parse.spec.ts
index 65a318c0b..6ae3427c7 100644
--- a/packages/compiler-sfc/__tests__/parse.spec.ts
+++ b/packages/compiler-sfc/__tests__/parse.spec.ts
@@ -203,13 +203,15 @@ h1 { color: red }
   })
 
   // #1120
-  test('alternative template lang should be treated as plain text', () => {
-    const content = `p(v-if="1 < 2") test`
+  test('template with preprocessor lang should be treated as plain text', () => {
+    const content = `p(v-if="1 < 2") test <div/>`
     const { descriptor, errors } = parse(
       `<template lang="pug">` + content + `</template>`
     )
     expect(errors.length).toBe(0)
     expect(descriptor.template!.content).toBe(content)
+    // should not attempt to parse the content
+    expect(descriptor.template!.ast.children.length).toBe(1)
   })
 
   //#2566
diff --git a/packages/compiler-sfc/src/compileTemplate.ts b/packages/compiler-sfc/src/compileTemplate.ts
index 8c5e492fa..d8c105908 100644
--- a/packages/compiler-sfc/src/compileTemplate.ts
+++ b/packages/compiler-sfc/src/compileTemplate.ts
@@ -132,7 +132,8 @@ export function compileTemplate(
     try {
       return doCompileTemplate({
         ...options,
-        source: preprocess(options, preprocessor)
+        source: preprocess(options, preprocessor),
+        ast: undefined // invalidate AST if template goes through preprocessor
       })
     } catch (e: any) {
       return {
diff --git a/packages/compiler-sfc/src/parse.ts b/packages/compiler-sfc/src/parse.ts
index 22fde2155..3330254d2 100644
--- a/packages/compiler-sfc/src/parse.ts
+++ b/packages/compiler-sfc/src/parse.ts
@@ -245,8 +245,10 @@ export function parse(
         )
       }
     }
-    // no need to genMap for template as its AST already accounts for the
-    // position in the SFC
+    // only genMap for template when it needs preprocessor
+    if (descriptor.template && descriptor.template.lang) {
+      genMap(descriptor.template)
+    }
     genMap(descriptor.script)
     descriptor.styles.forEach(genMap)
     descriptor.customBlocks.forEach(genMap)