Add Infection + PHPStan CI check on PHP 8.3 and fix PHP 7.x template serialization

.github/workflows/ci.yml

@@ -28,7 +28,7 @@ jobs:
             8.4
         ]
         composer: [basic]
-    timeout-minutes: 10
+    timeout-minutes: 30
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
@@ -55,7 +55,8 @@ jobs:
       - name: Install dependencies
         run: |
           if [[ "${{ matrix.php }}" == "8.3" ]]; then
-            composer require phpstan/phpstan --no-update
+            composer config allow-plugins.infection/extension-installer true
+            composer require phpstan/phpstan:^2.1 infection/infection:^0.32.7 --dev --no-update --no-interaction
           fi;
 
           if [[ "${{ matrix.composer }}" == "lowest" ]]; then
@@ -78,6 +79,12 @@ jobs:
         run: |
           php vendor/bin/phpstan analyse -c phpstan.neon --no-progress
 
+      - name: Run infection with phpstan integration
+        if: ${{ matrix.php == '8.3' }}
+        run: |
+          mkdir -p build/logs
+          XDEBUG_MODE=coverage php vendor/bin/infection run --configuration=infection.json5
+
       - name: Upload coverage results to Coveralls
         continue-on-error: true
         env:

infection.json5

@@ -0,0 +1,21 @@
+{
+  "$schema": "vendor/infection/infection/resources/schema.json",
+  "source": {
+    "directories": [
+      "src"
+    ]
+  },
+  "phpUnit": {
+    "configDir": "."
+  },
+  "phpStan": {
+    "configDir": "."
+  },
+  "staticAnalysisTool": "phpstan",
+  "threads": "max",
+  "minMsi": 0,
+  "minCoveredMsi": 0,
+  "logs": {
+    "text": "build/logs/infection.log"
+  }
+}

src/voku/helper/HtmlDomParser.php

@@ -1185,48 +1185,81 @@
      *
      * @param \DOMNode $node
      */
     private function serializeNode(\DOMNode $node): string
     {
-        // For script/style on PHP < 8.0 use ownerDocument to avoid fresh-doc
-        // libxml injecting "\n" inside raw-text content.
-        $useOwnerDoc = \PHP_VERSION_ID < 80000
-            && $node instanceof \DOMElement
-            && \in_array(\strtolower($node->tagName), ['script', 'style'], true);
-
-        if (!$useOwnerDoc) {
-            $document = new \DOMDocument('1.0', $this->getEncoding());
-            $document->preserveWhiteSpace = true;
-            $document->formatOutput = false;
-
-            $importedNode = $document->importNode($node, true);
-            // @phpstan-ignore instanceof.alwaysTrue (importNode() returns DOMNode here)
-            if (!$importedNode instanceof \DOMNode) {
-                return '';
-            }
+        if (\PHP_VERSION_ID < 80000 && $node instanceof \DOMElement) {
+            return $this->serializeElementNodeForPhpLt8($node);
+        }
 
-            $document->appendChild($importedNode);
+        $document = new \DOMDocument('1.0', $this->getEncoding());
+        $document->preserveWhiteSpace = true;
+        $document->formatOutput = false;
 
-            $content = $document->saveHTML($importedNode);
-        } else {
-            // PHP < 8.0 script/style: serialize from original document and
-            // strip only the trailing "\n" that older libxml appends after
-            // raw-text elements.
-            $ownerDoc = $node->ownerDocument;
-            $content = $ownerDoc !== null ? $ownerDoc->saveHTML($node) : false;
-            // Older libxml appends exactly one synthetic trailing "\n" here;
-            // preserve any real user-provided trailing newlines in the content.
-            if ($content !== false && \substr($content, -1) === "\n") {
-                $content = \substr($content, 0, -1);
-            }
+        $importedNode = $document->importNode($node, true);
+        // @phpstan-ignore instanceof.alwaysTrue (importNode() returns DOMNode here)
+        if (!$importedNode instanceof \DOMNode) {
+            return '';
         }
 
+        $document->appendChild($importedNode);
+
+        $content = $document->saveHTML($importedNode);
+
         if ($content === false) {
             return '';
         }
 
         return $content;
     }
 
+    /**
+     * On PHP < 8.0, saveHTML($node) injects formatting newlines for detached
+     * block-level elements, so serialize a temporary whole document instead.
+     *
+     * @param \DOMElement $node
+     *
+     * @return string
+     */
+    private function serializeElementNodeForPhpLt8(\DOMElement $node): string
+    {
+        $document = new \DOMDocument('1.0', $this->getEncoding());
+        $document->preserveWhiteSpace = true;
+        $document->formatOutput = false;
+
+        $importedNode = $document->importNode($node, true);
+        // @phpstan-ignore instanceof.alwaysTrue (importNode() returns DOMNode here)
+        if (!$importedNode instanceof \DOMElement) {
+            return '';
+        }
+
+        $document->appendChild($importedNode);
+
+        $content = $document->saveHTML();
+        if ($content === false) {
+            return '';
+        }
+
+        $content = (string) \preg_replace('/^<!DOCTYPE[^>]+>\s*/i', '', $content);
+
+        $tagName = \strtolower($importedNode->tagName);
+        if ($tagName !== 'html') {
+            $content = (string) \preg_replace('/^<html[^>]*>/i', '', $content);
+            $content = (string) \preg_replace('/<\/html>\s*$/i', '', $content);
+
+            if ($tagName !== 'body') {
+                $content = (string) \preg_replace('/^<body[^>]*>/i', '', $content);
+                $content = (string) \preg_replace('/<\/body>\s*$/i', '', $content);
+                $content = \str_replace('<body></body>', '', $content);
+            }
+        }
+
+        if (\substr($content, -1) === "\n") {
+            $content = \substr($content, 0, -1);
+        }
+
+        return $content;
+    }
+
     /**
      * Serialize the single element that was imported via the node-backed
      * constructor, for PHP < 8.0.

tests/HtmlSerializationRegressionTest.php

@@ -72,6 +72,34 @@
         static::assertSame('before<span>middle</span><strong>after</strong>', $parser->innerHtml());
     }
 
+    public function testSerializeElementNodeDoesNotAppendTrailingNewline()
+    {
+        $document = HtmlDomParser::str_get_html(
+            '<div><span>one</span><br><p>two</p><template id="card"><section><h2>Title</h2><p>Body</p></section></template></div>'
+        );
+
+        $serializeElementNodeForPhpLt8 = new \ReflectionMethod(HtmlDomParser::class, 'serializeElementNodeForPhpLt8');
+        if (\PHP_VERSION_ID < 80100) {
+            // This version check is only for Reflection behavior: private method
+            // access still needs setAccessible() when PHP_VERSION_ID < 80100
+            // (PHP 8.0 and earlier).
+            $serializeElementNodeForPhpLt8->setAccessible(true);
+        }
+
+        $spanHtml = $serializeElementNodeForPhpLt8->invoke($document, $document->getElementByTagName('span')->getNode());
+        $brHtml = $serializeElementNodeForPhpLt8->invoke($document, $document->getElementByTagName('br')->getNode());
+        $pHtml = $serializeElementNodeForPhpLt8->invoke($document, $document->getElementByTagName('p')->getNode());
+
+        static::assertSame(
+            '<span>one</span><br><p>two</p>',
+            $spanHtml . $brHtml . $pHtml
+        );
+        static::assertSame(
+            '<template id="card"><section><h2>Title</h2><p>Body</p></section></template>',
+            $serializeElementNodeForPhpLt8->invoke($document, $document->findOne('template')->getNode())
+        );
+    }
+
     public function testNodeBackedTextNodeHtmlPreservesTextVerbatim()
     {
         $document = HtmlDomParser::str_get_html('<div>before<span>middle</span>after</div>');