Merge pull request #8936 from quarto-dev/improve/auto-max-backticks

markdown - change markdown gen to automatically emit sufficiently-many : and `s

src/core/handlers/base.ts

@@ -27,7 +27,12 @@ import {
   optionCommentPatternFromLanguage,
 } from "../lib/partition-cell-options.ts";
 import { ConcreteSchema } from "../lib/yaml-schema/types.ts";
-import { pandocBlock, pandocList, pandocRawStr } from "../pandoc/codegen.ts";
+import {
+  pandocCode,
+  pandocDiv,
+  pandocList,
+  pandocRawStr,
+} from "../pandoc/codegen.ts";
 
 import {
   kCapLoc,
@@ -61,7 +66,13 @@ import {
   kTblCapLoc,
 } from "../../config/constants.ts";
 import { DirectiveCell } from "../lib/break-quarto-md-types.ts";
-import { basename, dirname, join, relative, resolve } from "../../deno_ral/path.ts";
+import {
+  basename,
+  dirname,
+  join,
+  relative,
+  resolve,
+} from "../../deno_ral/path.ts";
 import { figuresDir, inputFilesDir } from "../render.ts";
 import { ensureDirSync } from "fs/mod.ts";
 import { mappedStringFromFile } from "../mapped-text.ts";
@@ -629,12 +640,9 @@ export const baseHandler: LanguageHandler = {
 
     const unrolledOutput = isPowerpointOutput && !hasLayoutAttributes;
 
-    const t3 = pandocBlock("```");
-    const t4 = pandocBlock("````");
-
     const cellBlock = unrolledOutput
       ? pandocList({ skipFirstLineBreak: true })
-      : pandocBlock(":::")({
+      : pandocDiv({
         classes: ["cell", ...classes],
         attrs,
       });
@@ -667,7 +675,7 @@ export const baseHandler: LanguageHandler = {
 
     switch (options.echo) {
       case true: {
-        const cellInput = t3({
+        const cellInput = pandocCode({
           classes: cellInputClasses,
           attrs: cellInputAttrs,
         });
@@ -676,11 +684,11 @@ export const baseHandler: LanguageHandler = {
         break;
       }
       case "fenced": {
-        const cellInput = t4({
+        const cellInput = pandocCode({
           classes: ["markdown", ...cellInputClasses.slice(1)], // replace the language class with markdown
           attrs: cellInputAttrs,
         });
-        const cellFence = t3({
+        const cellFence = pandocCode({
           language: this.languageName,
           skipFirstLineBreak: true,
         });
@@ -695,7 +703,7 @@ export const baseHandler: LanguageHandler = {
       }
     }
 
-    const divBlock = pandocBlock(":::");
+    const divBlock = pandocDiv;
 
     // PandocNodes ignore self-pushes (n.push(n))
     // this makes it much easier to write the logic around "unrolled blocks"

src/core/pandoc/codegen.ts

@@ -142,7 +142,7 @@ export function pandocList(opts: {
   };
 }
 
-export function pandocBlock(delimiter: string) {
+export function pandocBlock(delimiterCharacter: ":" | "`") {
   return function (
     opts?: {
       language?: string;
@@ -197,21 +197,42 @@ export function pandocBlock(delimiter: string) {
         }
       },
       emit: function (ls: EitherString[]) {
+        const innerLs: EitherString[] = [];
         const lb = skipFirstLineBreak ? "" : "\n";
-        ls.push(`${lb}${delimiter}${attrString()}\n`);
+
         for (const entry of contents!) {
-          entry.emit(ls);
+          entry.emit(innerLs);
         }
-        if (!asMappedString(ls[ls.length - 1] || "\n").value.endsWith("\n")) {
-          ls.push(`\n`);
+        if (
+          !asMappedString(innerLs[innerLs.length - 1] || "\n").value.endsWith(
+            "\n",
+          )
+        ) {
+          innerLs.push(`\n`);
         }
+
+        // now find the longest streak of delimiter characters in the innerLs
+        const longestStreak = Math.max(...innerLs.map((eitherS) => {
+          const s = asMappedString(eitherS).value;
+          return s.match(new RegExp(`${delimiterCharacter}+`, "g"))?.[0]
+            ?.length || 0;
+        }));
+        const delimiter = delimiterCharacter.repeat(
+          Math.max(3, longestStreak + 1),
+        );
+        ls.push(`${lb}${delimiter}${attrString()}\n`);
+
+        // FIXME this will incur a runtime of eventually O(n * m) where n is the number of lines and m is the depth of
+        // the PandocNode tree.
+        ls.push(...innerLs);
+
         ls.push(`${delimiter}\n`);
       },
     };
   };
 }
 
-export const pandocDiv = pandocBlock(":::");
-export const pandocCode = pandocBlock("```");
+export const pandocDiv = pandocBlock(":");
+export const pandocCode = pandocBlock("`");
 export const pandocFigure = pandocHtmlBlock("figure");
 export const pandocFigCaption = pandocHtmlBlock("figcaption");

src/execute/ojs/compile.ts

@@ -299,7 +299,7 @@ export async function ojsCompile(
         );
         ojsParseError(err, cellSrc);
 
-        const preDiv = pandocBlock("````")({
+        const preDiv = pandocCode({
           classes: ["numberLines", "java"],
           attrs: [
             `startFrom="${cellStartingLoc - 1}"`,
@@ -543,7 +543,7 @@ export async function ojsCompile(
           `startFrom="${cellStartingLoc - 1}"`,
           `source-offset="${cell.sourceOffset}"`,
         );
-        const srcDiv = pandocBlock("````")({
+        const srcDiv = pandocCode({
           classes: ourClasses,
           attrs: ourAttrs,
         });

tests/docs/smoke-all/mermaid/backticks.qmd

@@ -0,0 +1,21 @@
+---
+title: "mermaid backtick test"
+keep-md: true
+_quarto:
+  tests:
+    html:
+      ensureFileRegexMatches:
+        - []
+        - ["<p>This would have been"] # it can't be in a paragraph, it should be inside a code cell.
+---
+
+It's not important right now whether this syntax is correct for mermaid;
+the relevant aspect is that our pandoc codegen wasn't taking the
+inner number of backticks into account. This test checks for that.
+
+````{mermaid}
+%%| echo: true
+```
+This would have been a problem.
+```
+````