Escape unindented indented pseudo setext headers (#19350)

Co-authored-by: fisker <lionkay@gmail.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>

changelog_unreleased/markdown/19350.md

@@ -0,0 +1,31 @@
+#### Escape unindented indented pseudo setext headers (#19350 by @tats-u)
+
+Indented lines composed of only `=` or `-` were previously unindented without being escaped, which caused them to be mistakenly parsed as setext headers. With this change, Prettier now escapes such lines to prevent the misinterpretation:
+
+[Check parse results on CommonMark playground](https://spec.commonmark.org/dingus/?text=%3C!--%20Input%20--%3E%0ANot%20a%20header%0A%20%20%20%20%3D%3D%3D%0ANot%20a%20header%0A%20%20%20%20---%0ANot%20a%20list%20item%20or%20a%20header%0A%20%20%20%20-%0A%0A------%0A%0A%3C!--%20Prettier%20stable%20--%3E%0ANot%20a%20header%0A%3D%3D%3D%0ANot%20a%20header%0A---%0ANot%20a%20list%20item%20or%20a%20header%20-%0A%0A------%0A%0A%3C!--%20Prettier%20main%20--%3E%0ANot%20a%20header%0A%5C%3D%3D%3D%0ANot%20a%20header%0A%5C---%0ANot%20a%20list%20item%20or%20a%20header%0A%5C-)
+
+<!-- prettier-ignore -->
+```markdown
+<!-- Input -->
+Not a header
+    ===
+Not a header
+    ---
+Not a list item or a header
+    -
+
+<!-- Prettier stable -->
+Not a header
+===
+Not a header
+---
+Not a list item or a header -
+
+<!-- Prettier main -->
+Not a header
+\===
+Not a header
+\---
+Not a list item or a header
+\-
+```

src/language-markdown/print/mdast.js

@@ -22,6 +22,7 @@ import {
   getFencedCodeBlockValue,
   getNthListSiblingIndex,
   isAutolink,
+  isNewLine,
   isPrettierIgnore,
   splitText,
 } from "../utilities.js";
@@ -65,6 +66,19 @@ function hasFakeWhitespaceAfterNextToken(path) {
   return afterNext?.type === "whitespace" && afterNext.value === "";
 }
 
+/**
+ * @param {AstPath} path
+ * @returns {boolean}
+ */
+function isNextTokenFakeSetextH2Line(path) {
+  if (!isNewLine(path.node) || path.next?.value !== "-") {
+    return false;
+  }
+
+  const afterNext = path.siblings[path.index + 2];
+  return !afterNext || isNewLine(afterNext);
+}
+
 function printMdast(path, options, print) {
   const { node } = path;
 
@@ -116,7 +130,9 @@ function printMdast(path, options, print) {
     case "sentence":
       return printSentence(path, print);
     case "word":
-      return options.parser !== "mdx" ? printWord(path) : printWordLegacy(path);
+      return options.parser !== "mdx"
+        ? printWord(path, options)
+        : printWordLegacy(path);
     case "whitespace": {
       const { next } = path;
 
@@ -125,7 +141,9 @@ function printMdast(path, options, print) {
         next &&
         /^>|^(?:[*+-]|#{1,6}|\d+[).])$/.test(next.value) &&
         // Avoid https://github.com/prettier/prettier/issues/18861
-        !hasFakeWhitespaceAfterNextToken(path)
+        !hasFakeWhitespaceAfterNextToken(path) &&
+        // Next fake setext h2 `-` is going to be escaped, so no need to join adjacent words
+        !(options.proseWrap === "preserve" && isNextTokenFakeSetextH2Line(path))
           ? "never"
           : options.proseWrap;
 

src/language-markdown/print/word.js

@@ -1,5 +1,7 @@
 import { PUNCTUATION_REGEXP } from "../constants.evaluate.js";
-import { isAutolink } from "../utilities.js";
+import { isAutolink, isNewLine } from "../utilities.js";
+
+const fakeSetextHeaderRegex = /^(?:=+|-+)$/;
 
 /**
  * @import AstPath from "../../common/ast-path.js"
@@ -8,25 +10,36 @@ import { isAutolink } from "../utilities.js";
 
 /**
  * @param {AstPath} path
+ * @param {*} options
  * @return {Doc}
  */
-function printWord(path) {
+function printWord(path, options) {
   const { node } = path;
   const emphasisOrStrong = path.findAncestor(
     (p) => p.type === "emphasis" || p.type === "strong",
   );
-  if (!emphasisOrStrong) {
-    return node.value;
-  }
-  const { previous, next, grandparent } = path;
   let text = node.value;
+  if (!emphasisOrStrong) {
+    if (
+      options.proseWrap === "preserve" &&
+      path.parent.type === "sentence" &&
+      fakeSetextHeaderRegex.test(text) &&
+      isNewLine(path.previous) &&
+      (path.isLast || isNewLine(path.next))
+    ) {
+      // escape indented pseudo setext header, e.g. `Previous line↵␣␣␣␣===`
+      return `\\${text}`;
+    }
+
+    return text;
+  }
 
   // escape leading `*` or `_` if it's the first character in an emphasis/strong
   if (
     path.isFirst &&
     (text.startsWith("*") || text.startsWith("_")) &&
     path.callParent(() => path.isFirst) &&
-    grandparent === emphasisOrStrong
+    path.grandparent === emphasisOrStrong
   ) {
     text = `\\${text}`;
   }
@@ -44,9 +57,9 @@ function printWord(path) {
       }
       if (
         canOpenOrCloseStrongOrEmphasis(
-          preceding.at(-1) || previous?.value.at(-1),
+          preceding.at(-1) || path.previous?.value.at(-1),
           delimiterRun,
-          following[0] || next?.value[0],
+          following[0] || path.next?.value[0],
         )
       ) {
         return `${preceding}\\${delimiterRun}${following}`;

src/language-markdown/utilities.js

@@ -321,6 +321,8 @@ function isSetextHeading(node) {
   return start.line !== end.line;
 }
 
+const isNewLine = (node) => node?.type === "whitespace" && node.value === "\n";
+
 export {
   getFencedCodeBlockValue,
   getNthListSiblingIndex,
@@ -330,6 +332,7 @@ export {
   INLINE_NODE_TYPES,
   INLINE_NODE_WRAPPER_TYPES,
   isAutolink,
+  isNewLine,
   isPrettierIgnore,
   isSetextHeading,
   KIND_CJ_LETTER,

tests/format/markdown/heading/setext/__snapshots__/format.test.js.snap

@@ -124,6 +124,55 @@ Bar
 ================================================================================
 `;
 
+exports[`not-header.md format 1`] = `
+====================================options=====================================
+parsers: ["markdown"]
+                                                      printWidth: 80 (default) |
+=====================================input======================================
+\\===
+    ===
+\\===
+    ---
+\\---
+    -
+\\-
+    =
+\\=
+\\\\---
+\\\\===
+    \\-
+    \\=
+    \\\\-
+    \\\\=
+-foo ---**bar**===あ===
+
+\\-
+    -
+
+=====================================output=====================================
+\\===
+\\===
+\\===
+\\---
+\\---
+\\-
+\\-
+\\=
+\\=
+\\\\---
+\\\\===
+\\-
+\\=
+\\\\-
+\\\\=
+-foo ---**bar**===あ===
+
+\\-
+\\-
+
+================================================================================
+`;
+
 exports[`setext.md format 1`] = `
 ====================================options=====================================
 parsers: ["markdown"]

tests/format/markdown/heading/setext/not-header.md

@@ -0,0 +1,19 @@
+\===
+    ===
+\===
+    ---
+\---
+    -
+\-
+    =
+\=
+\\---
+\\===
+    \-
+    \=
+    \\-
+    \\=
+-foo ---**bar**===あ===
+
+\-
+    -