use custom pandoc writer to preserve raw html blocks

apps/vscode/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## 1.116.0 (Not yet released)
+
+- Fix issue with raw html blocks being removed from document by Visual Editor (<https://github.com/quarto-dev/quarto/issues/552>)
+
 ## 1.115.0 (Release on 20 September 2024)
 
 - Suppress background highlight on first line (<https://github.com/quarto-dev/quarto/pull/537>).

apps/vscode/package.json

@@ -11,7 +11,7 @@
     "pandoc",
     "quarto"
   ],
-  "version": "1.115.0",
+  "version": "1.116.0",
   "repository": {
     "type": "git",
     "url": "https://github.com/quarto-dev/quarto/tree/main/apps/vscode"

packages/editor-server/src/resources/md-writer.lua

@@ -0,0 +1,23 @@
+---@diagnostic disable: undefined-global
+---
+--- Writer for markdown that preserves raw html attributes.
+--- 
+--- This changed in pandoc 3.2, see https://github.com/rstudio/rstudio/issues/15189.
+
+local html_formats = pandoc.List{'html', 'html4', 'html5'}
+function Writer (doc, opts)
+  if opts.extensions:includes 'raw_attribute' then
+    doc = doc:walk{
+      RawBlock = function (raw)
+        if html_formats:includes(raw.format) then
+          local md = pandoc.write(pandoc.Pandoc(raw), 'markdown-raw_html')
+          return pandoc.RawBlock('markdown', md)
+        end
+      end
+    }
+  end
+  return pandoc.write(doc, {format = 'markdown', extensions = opts.extensions}, opts)
+end
+
+Extensions = pandoc.format.extensions 'markdown'
+Template = pandoc.template.default 'markdown'

packages/editor-server/src/server/pandoc.ts

@@ -55,6 +55,13 @@ import { EditorServerOptions } from './server';
 
 export function pandocServer(options: EditorServerOptions) : PandocServer {
 
+  // work around change in pandoc 3.2+: https://github.com/jgm/pandoc/issues/9677 by
+  // using a custom pandoc writer
+  function substituteCustomMarkdownWriter(format: string): string {
+    const customWriterScript = path.join(options.pandoc.resourcesDir, 'md-writer.lua')
+    return format.replace(/^markdown(?=[-+]|$)/, customWriterScript);
+  }
+
   async function runPandoc(args: readonly string[] | null, stdin?: string) : Promise<string> {
     return pandoc(options.pandoc, args, stdin);
   }
@@ -100,9 +107,9 @@ export function pandocServer(options: EditorServerOptions) : PandocServer {
       return ast;
     },
     async astToMarkdown(ast: PandocAst, format: string, options: string[]): Promise<string> {
-      const markdown = await runPandoc(
+     const markdown = await runPandoc(
         ["--from", "json",
-         "--to", format, ...options],
+         "--to", substituteCustomMarkdownWriter(format), ...options],
          JSON.stringify(ast)
       );
       return markdown;