work around syntax highlighter stripping leading blank lines

The syntaxhighlighter library strips blank lines at the start of code
block it is trying to highlight. In addition to just being wrong, this
means that line numbers on the source page will not line up with the
real files.

Work around this by providing the library with fake objects that ignore
the attempt to strip these lines. This uses a fake string and a fake
element to hold it. This is an ugly workaround, but allows us to
continue using the syntax highlighter via npm.

root/static/js/syntaxhighlighter.mjs

@@ -132,19 +132,16 @@ const processPackages = function(code) {
     return code.replace(/(<code class="p(?:er)?l keyword">(use|package|require)<\/code> <code class="p(?:er)?l plain">)([A-Z_a-z][0-9A-Z_a-z]*(?:::[0-9A-Z_a-z]+)*)(.*?<\/code>)/g, replace_pattern);
 };
 
-const getCodeLinesHtml = Renderer.prototype.getCodeLinesHtml;
-Renderer.prototype.getCodeLinesHtml = function() {
-    // the syntax highlighter has a bug that strips spaces from the first line.
-    // replace any leading whitespace with an entity, preventing that.
-    // html = html.replace(/^ /, "&#32;");
-    // html = html.replace(/^\t/, "&#9;");
-    let html = getCodeLinesHtml.apply(this, arguments);
-    return processPackages.call(this, html);
+const processUrls = Renderer.prototype.processUrls;
+Renderer.prototype.processUrls = function(html, ...args) {
+    html = processPackages.apply(this, [html]);
+    html = processUrls.apply(this, [html, ...args]);
+    return html;
 };
 
 const getHtml = Renderer.prototype.getHtml;
-Renderer.prototype.getHtml = function() {
-    let html = getHtml.call(this);
+Renderer.prototype.getHtml = function(...args) {
+    let html = getHtml.call(this, ...args);
     html = html.replace(/\s+(<(tbody|table|div)\b)/g, '$1');
     html = html.replace(/(<\/(tbody|table|div)>)\s+/g, '$1');
     return html;
@@ -231,14 +228,44 @@ for (const code of document.querySelectorAll(".content pre > code")) {
 
     config.package_target_type = source ? 'source' : 'pod';
 
-    // highlighter strips leading blank lines, throwing off line numbers.
-    // add a blank line for the highlighter to strip
-    // const html = code.innerHTML;
-    // if (html.match(/^ *\n/)) {
-    //     code.innerHTML = "\n " + html;
-    // }
+    let highlightObject = code;
+
+    const html = code.innerHTML;
+    if (html.match(/^ *\n+/)) {
+        // highlighter strips leading blank lines, throwing off line numbers.
+        // use this awful hack to bypass it. depends on specific details inside
+        // the syntaxhighlighter module
+
+        const fakeCode = {
+            className: code.className,
+            id: code.id,
+            title: code.title,
+            innerHTML: {
+                toString: function() {
+                    return html
+                },
+                replace: function(search, replace) {
+                    if (search.toString() == /^[ ]*[\n]+|[\n]*[ ]*$/g.toString()) {
+                        return html.replace(/\n$/g, '');
+                    }
+                    return html.replace(search, replace);
+                },
+            },
+        };
+        const parentNode = code.parentNode;
+        fakeCode.parentNode = {
+            replaceChild: function(newEl, oldEl) {
+                if (oldEl === fakeCode) {
+                    oldEl = code
+                }
+                parentNode.replaceChild(newEl, oldEl);
+            },
+        };
+
+        highlightObject = fakeCode;
+    }
 
-    SyntaxHighlighter.highlight(config, code);
+    SyntaxHighlighter.highlight(config, highlightObject);
 
     const pod_lines = pre.dataset.podLines;
     if (pod_lines) {