Fix renderer content loss in non-HTML renderers (markdown/plain/ansi)

src/Renderer/AnsiRenderer.php

@@ -30,6 +30,7 @@
 use Djot\Node\Inline\Code;
 use Djot\Node\Inline\Delete;
 use Djot\Node\Inline\Emphasis;
+use Djot\Node\Inline\EscapedText;
 use Djot\Node\Inline\FootnoteRef;
 use Djot\Node\Inline\HardBreak;
 use Djot\Node\Inline\Highlight;
@@ -403,6 +404,7 @@ protected function renderNode(Node $node): string
             $node instanceof LineBlock => $this->renderLineBlock($node),
             $node instanceof Footnote => $this->renderFootnote($node),
             $node instanceof Text => $node->getContent(),
+            $node instanceof EscapedText => $node->getContent(),
             $node instanceof Abbreviation => $this->renderAbbreviation($node),
             $node instanceof Emphasis => $this->renderEmphasis($node),
             $node instanceof Strong => $this->renderStrong($node),
@@ -608,7 +610,16 @@ protected function renderThematicBreak(): string
 
     protected function renderDiv(Div $node): string
     {
-        return $this->renderChildren($node);
+        $body = $this->renderChildren($node);
+        // A Div's quoted title (e.g. an admonition title carried as the `title`
+        // attribute) is preserved as a leading bold line instead of being
+        // dropped.
+        $title = $node->getAttribute('title');
+        if (is_string($title) && $title !== '') {
+            return $this->style($title, self::BOLD) . "\n\n" . $body;
+        }
+
+        return $body;
     }
 
     protected function renderTable(Table $node): string

src/Renderer/MarkdownRenderer.php

@@ -6,12 +6,14 @@
 
 use Djot\Event\RenderEvent;
 use Djot\Node\Block\BlockQuote;
+use Djot\Node\Block\Caption;
 use Djot\Node\Block\CodeBlock;
 use Djot\Node\Block\Comment;
 use Djot\Node\Block\DefinitionDescription;
 use Djot\Node\Block\DefinitionList;
 use Djot\Node\Block\DefinitionTerm;
 use Djot\Node\Block\Div;
+use Djot\Node\Block\Figure;
 use Djot\Node\Block\Footnote;
 use Djot\Node\Block\Heading;
 use Djot\Node\Block\LineBlock;
@@ -24,9 +26,11 @@
 use Djot\Node\Block\TableRow;
 use Djot\Node\Block\ThematicBreak;
 use Djot\Node\Document;
+use Djot\Node\Inline\Abbreviation;
 use Djot\Node\Inline\Code;
 use Djot\Node\Inline\Delete;
 use Djot\Node\Inline\Emphasis;
+use Djot\Node\Inline\EscapedText;
 use Djot\Node\Inline\FootnoteRef;
 use Djot\Node\Inline\HardBreak;
 use Djot\Node\Inline\Highlight;
@@ -140,6 +144,14 @@ protected function renderNode(Node $node): string
             $node instanceof LineBlock => $this->renderLineBlock($node),
             $node instanceof Footnote => $this->renderFootnote($node),
             $node instanceof Text => $this->escapeText($node->getContent()),
+            // Keep the backslash so the literal stays literal when re-parsed as
+            // Markdown: a bare `.` from `\.` would turn `1\. x` back into an
+            // ordered list. EscapedText only ever holds escaped ASCII
+            // punctuation, all of which CommonMark allows a `\` before.
+            $node instanceof EscapedText => '\\' . $node->getContent(),
+            $node instanceof Figure => $this->renderFigure($node),
+            $node instanceof Caption => $this->renderCaption($node),
+            $node instanceof Abbreviation => $this->renderAbbreviation($node),
             $node instanceof Emphasis => $this->renderEmphasis($node),
             $node instanceof Strong => $this->renderStrong($node),
             $node instanceof Code => $this->renderCode($node),
@@ -287,8 +299,17 @@ protected function renderDefinitionDescription(DefinitionDescription $node): str
 
     protected function renderDiv(Div $node): string
     {
-        // Divs don't exist in Markdown, just render content
-        return $this->renderChildren($node);
+        // Divs/admonitions don't exist in Markdown; render the content. A Div's
+        // quoted title (e.g. an admonition title carried as the `title`
+        // attribute) would otherwise be lost - preserve it as a leading bold
+        // line.
+        $body = $this->renderChildren($node);
+        $title = $node->getAttribute('title');
+        if (is_string($title) && $title !== '') {
+            return '**' . $this->escapeText($title) . "**\n\n" . $body;
+        }
+
+        return $body;
     }
 
     protected function renderTable(Table $node): string
@@ -475,6 +496,48 @@ protected function renderRawInline(RawInline $node): string
         return '';
     }
 
+    /**
+     * A figure renders its target then its caption as a separate block
+     * (Markdown has no figure element). A BLANK line before the caption is
+     * required, not just a newline: against a block-quote target a single
+     * newline would make the caption a lazy continuation of the quote and
+     * swallow it.
+     */
+    protected function renderFigure(Figure $node): string
+    {
+        $output = '';
+        foreach ($node->getChildren() as $child) {
+            if ($child instanceof Caption) {
+                $output = rtrim($output) . "\n\n" . $this->renderCaption($child);
+            } else {
+                $output .= $this->renderNode($child);
+            }
+        }
+
+        return $output;
+    }
+
+    protected function renderCaption(Caption $node): string
+    {
+        return trim($this->renderChildren($node)) . "\n\n";
+    }
+
+    /**
+     * Markdown has no abbreviation syntax; emit inline abbr HTML so the title
+     * is preserved (mirrors how subscript/superscript fall back to inline HTML).
+     */
+    protected function renderAbbreviation(Abbreviation $node): string
+    {
+        // The whole element is raw inline HTML, so both the title (attribute)
+        // and the text (element content) need HTML escaping, NOT Markdown text
+        // escaping: a `"` in the title or a `<` in the text would otherwise
+        // break the tag / be misparsed as markup downstream.
+        $title = htmlspecialchars($node->getTitle(), ENT_QUOTES, 'UTF-8');
+        $text = htmlspecialchars($this->renderChildren($node), ENT_QUOTES, 'UTF-8');
+
+        return '<abbr title="' . $title . '">' . $text . '</abbr>';
+    }
+
     protected function escapeText(string $text): string
     {
         // Escape special Markdown characters in text

src/Renderer/PlainTextRenderer.php

@@ -11,6 +11,7 @@
 use Djot\Node\Block\DefinitionDescription;
 use Djot\Node\Block\DefinitionList;
 use Djot\Node\Block\DefinitionTerm;
+use Djot\Node\Block\Div;
 use Djot\Node\Block\Footnote;
 use Djot\Node\Block\Heading;
 use Djot\Node\Block\LineBlock;
@@ -25,6 +26,7 @@
 use Djot\Node\Document;
 use Djot\Node\Inline\Code;
 use Djot\Node\Inline\Delete;
+use Djot\Node\Inline\EscapedText;
 use Djot\Node\Inline\FootnoteRef;
 use Djot\Node\Inline\HardBreak;
 use Djot\Node\Inline\Image;
@@ -114,6 +116,7 @@ protected function renderNode(Node $node): string
 
         return match (true) {
             $node instanceof Document => $this->renderChildren($node),
+            $node instanceof Div => $this->renderDiv($node),
             $node instanceof Paragraph => $this->renderParagraph($node),
             $node instanceof Heading => $this->renderHeading($node),
             $node instanceof CodeBlock => $this->renderCodeBlock($node),
@@ -132,6 +135,7 @@ protected function renderNode(Node $node): string
             $node instanceof LineBlock => $this->renderLineBlock($node),
             $node instanceof Footnote => $this->renderFootnote($node),
             $node instanceof Text => $node->getContent(),
+            $node instanceof EscapedText => $node->getContent(),
             $node instanceof Code => $node->getContent(),
             $node instanceof Math => $node->getContent(),
             $node instanceof Image => $node->getAlt(),
@@ -161,6 +165,19 @@ protected function renderParagraph(Paragraph $node): string
         return $this->renderChildren($node) . "\n\n";
     }
 
+    protected function renderDiv(Div $node): string
+    {
+        $body = $this->renderChildren($node);
+        // A Div's quoted title (e.g. an admonition title carried as the `title`
+        // attribute) is preserved as a leading line instead of being dropped.
+        $title = $node->getAttribute('title');
+        if (is_string($title) && $title !== '') {
+            return $title . "\n\n" . $body;
+        }
+
+        return $body;
+    }
+
     protected function renderHeading(Heading $node): string
     {
         return $this->renderChildren($node) . "\n\n";

tests/TestCase/Renderer/RendererContentLossTest.php

@@ -0,0 +1,54 @@
+<?php
+
+declare(strict_types=1);
+
+namespace Djot\Test\TestCase\Renderer;
+
+use Djot\DjotConverter;
+use PHPUnit\Framework\TestCase;
+
+/**
+ * Renderer content loss: the non-HTML renderers (markdown/plain/ansi) must not
+ * silently drop content the HTML renderer keeps.
+ *
+ * Ported from carve-php commit 217b72f.
+ */
+class RendererContentLossTest extends TestCase
+{
+    public function testEscapedTextIsNotDroppedInNonHtmlRenderers(): void
+    {
+        $md = DjotConverter::markdown()->convert('a \*lit\* b');
+        $this->assertStringContainsString('lit', $md);
+        $this->assertStringContainsString('*', $md);
+
+        $plain = DjotConverter::plainText()->convert('a \*lit\* b');
+        $this->assertStringContainsString('*lit*', $plain);
+
+        $ansi = DjotConverter::ansi()->convert('a \*lit\* b');
+        $this->assertStringContainsString('*lit*', $ansi);
+    }
+
+    public function testAbbreviationTitlePreservedInMarkdown(): void
+    {
+        $md = DjotConverter::markdown()->convert("The HTML spec.\n\n*[HTML]: HyperText Markup Language");
+        $this->assertStringContainsString('HyperText Markup Language', $md);
+    }
+
+    public function testFigureCaptionNotGluedInMarkdown(): void
+    {
+        $md = DjotConverter::markdown()->convert("![a](i.png)\n^ Cap text");
+        // caption sits on its own line, not glued to the image
+        $this->assertStringNotContainsString('i.png)Cap', $md);
+        $this->assertStringContainsString('Cap text', $md);
+    }
+
+    public function testDivTitlePreservedInNonHtmlRenderers(): void
+    {
+        // A Div carries a `title` attribute via Djot's attribute syntax; the
+        // non-HTML renderers must surface it as a leading line, not drop it.
+        $src = "{title=\"Heads up\"}\n:::\nbody\n:::";
+        $this->assertStringContainsString('Heads up', DjotConverter::markdown()->convert($src));
+        $this->assertStringContainsString('Heads up', DjotConverter::plainText()->convert($src));
+        $this->assertStringContainsString('Heads up', DjotConverter::ansi()->convert($src));
+    }
+}