Merge pull request #12 from elecena/fix/entities-handling

Improve XML entities handling

Author: macbre

src/XMLParser.php

@@ -17,6 +17,8 @@ class XMLParser implements \Iterator
     private ?string $currentTagName = null;
     private array $currentTagAttributes = [];
 
+    private string $currentTagContent = '';
+
     /**
      * The stack of the XML node names as go deeper into the tree.
      *
@@ -84,6 +86,8 @@ public function startXML(\XMLParser $parser, string $tagName, array $attributes)
         $this->currentTagName = $tagName;
         $this->currentTagAttributes = $attributes;
 
+        $this->currentTagContent = '';
+
         // append to the queue of items to iterate over
         $this->nodesQueue[] = new Nodes\XMLNodeOpen(
             name: $this->currentTagName,
@@ -94,19 +98,31 @@ public function startXML(\XMLParser $parser, string $tagName, array $attributes)
         $this->nodeNamesStack[] = $tagName;
     }
 
+    /**
+     * The XML parser "emits" separate characters when the node has the content with XML entities.
+     *
+     * For instance: <loc>https://example.com/index.html?ACTION=1004&amp;SITE=3</loc>
+     *
+     * Would emit: 'https://example.com/index.html?ACTION=1004', '&' and 'SITE=3' separately.
+     *
+     * So, just accumulate the characters as we're getting them and "emit" the XMLNodeContent instance
+     * when the node is closed.
+     */
     public function charXML(\XMLParser $parser, string $tagContent): void
+    {
+        $this->currentTagContent .= $tagContent;
+    }
+
+    public function endXML(\XMLParser $parser, string $tagName): void
     {
         // append to the queue of items to iterate over
         $this->nodesQueue[] = new Nodes\XMLNodeContent(
             name: $this->currentTagName,
             attributes: $this->currentTagAttributes,
-            content: $tagContent,
+            content: $this->currentTagContent,
             parentName: array_slice($this->nodeNamesStack, -2, 1)[0] ?: null
         );
-    }
 
-    public function endXML(\XMLParser $parser, string $tagName): void
-    {
         // Pop the node name off the end of stack
         array_pop($this->nodeNamesStack);
 

tests/XMLParserEntitiesTest.php

@@ -0,0 +1,26 @@
+<?php
+
+use Elecena\XmlIterator\Nodes\XMLNodeContent;
+use Elecena\XmlIterator\Nodes\XMLNodeOpen;
+use Elecena\XmlIterator\Nodes\XMLNodeClose;
+
+class XMLParserEntitiesTest extends XMLParserTestCase
+{
+    protected function getParserStream()
+    {
+        return fopen(__DIR__ . '/fixtures/sitemap-entities.xml', mode: 'rt');
+    }
+
+    public function testParsesTheLocNodesWithAmpersands(): void
+    {
+        $locations = [];
+
+        foreach($this->getParser()->iterateByNodeContent('loc') as $item) {
+            $locations[] = $item->content;
+        }
+
+        $this->assertCount(8, $locations);
+        $this->assertEquals('https://www.reichelt.com/index.html?ACTION=1004&SITE=1', $locations[0]);
+        $this->assertEquals('https://www.reichelt.com/magazin/en/sitemap.xml', $locations[7]);
+    }
+}

tests/XMLParserTest.php

@@ -47,11 +47,9 @@ public function testParsesTheLocNodes(): void
     {
         $locations = [];
 
-        foreach($this->getParser() as $item) {
-            if ($item instanceof XMLNodeContent && $item->name === 'loc') {
-                $locations[] = $item->content;
-                $this->assertEquals('sitemap', $item->parentName);
-            }
+        foreach($this->getParser()->iterateByNodeContent('loc') as $item) {
+            $locations[] = $item->content;
+            $this->assertEquals('sitemap', $item->parentName);
         }
 
         $this->assertCount(8, $locations);

tests/fixtures/sitemap-entities.xml

@@ -0,0 +1,28 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<sitemapindex xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+    <!-- https://www.reichelt.com/sitemap.xml -->
+    <sitemap>
+        <loc>https://www.reichelt.com/index.html?ACTION=1004&amp;SITE=1</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/index.html?ACTION=1004&amp;SITE=2</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/index.html?ACTION=1004&amp;SITE=3</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/index.html?ACTION=1004&amp;SITE=4</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.de/magazin/sitemap.xml</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/magazin/fr/sitemap.xml</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/magazin/nl/sitemap.xml</loc>
+    </sitemap>
+    <sitemap>
+        <loc>https://www.reichelt.com/magazin/en/sitemap.xml</loc>
+    </sitemap>
+</sitemapindex>