[DomCrawler] Add argument $normalizeWhitespace to Crawler::innerText() and make it return the first non-empty text

otsch · nicolas-grekas · commit bb0c214059ce · 2023-01-11T14:00:31.000+01:00
diff --git a/src/Symfony/Component/DomCrawler/CHANGELOG.md b/src/Symfony/Component/DomCrawler/CHANGELOG.md
@@ -5,6 +5,8 @@ CHANGELOG
 ---
 
  * Add `CrawlerSelectorCount` test constraint
+ * Add argument `$normalizeWhitespace` to `Crawler::innerText()`
+ * Make `Crawler::innerText()` return the first non-empty text
 
 6.0
 ---
diff --git a/src/Symfony/Component/DomCrawler/Crawler.php b/src/Symfony/Component/DomCrawler/Crawler.php
@@ -553,18 +553,34 @@ public function text(string $default = null, bool $normalizeWhitespace = true):
         $text = $this->getNode(0)->nodeValue;
 
         if ($normalizeWhitespace) {
-            return trim(preg_replace("/(?:[ \n\r\t\x0C]{2,}+|[\n\r\t\x0C])/", ' ', $text), " \n\r\t\x0C");
+            return $this->normalizeWhitespace($text);
         }
 
         return $text;
     }
 
     /**
      * Returns only the inner text that is the direct descendent of the current node, excluding any child nodes.
+     *
+     * @param bool $normalizeWhitespace Whether whitespaces should be trimmed and normalized to single spaces
      */
-    public function innerText(): string
+    public function innerText(/* bool $normalizeWhitespace = true */): string
     {
-        return $this->filterXPath('.//text()')->text();
+        $normalizeWhitespace = 1 <= \func_num_args() ? func_get_arg(0) : true;
+
+        foreach ($this->getNode(0)->childNodes as $childNode) {
+            if (\XML_TEXT_NODE !== $childNode->nodeType) {
+                continue;
+            }
+            if (!$normalizeWhitespace) {
+                return $childNode->nodeValue;
+            }
+            if ('' !== trim($childNode->nodeValue)) {
+                return $this->normalizeWhitespace($childNode->nodeValue);
+            }
+        }
+
+        return '';
     }
 
     /**
@@ -1189,4 +1205,9 @@ private function isValidHtml5Heading(string $heading): bool
     {
         return 1 === preg_match('/^\x{FEFF}?\s*(<!--[^>]*?-->\s*)*$/u', $heading);
     }
+
+    private function normalizeWhitespace(string $string): string
+    {
+        return trim(preg_replace("/(?:[ \n\r\t\x0C]{2,}+|[\n\r\t\x0C])/", ' ', $string), " \n\r\t\x0C");
+    }
 }
diff --git a/src/Symfony/Component/DomCrawler/Tests/AbstractCrawlerTest.php b/src/Symfony/Component/DomCrawler/Tests/AbstractCrawlerTest.php
@@ -348,12 +348,56 @@ public function testText()
         $this->assertSame('my value', $this->createTestCrawler(null)->filterXPath('//ol')->text('my value'));
     }
 
-    public function testInnerText()
+    public function provideInnerTextExamples()
     {
-        self::assertCount(1, $crawler = $this->createTestCrawler()->filterXPath('//*[@id="complex-element"]'));
+        return [
+            [
+                '//*[@id="complex-elements"]/*[@class="one"]',     // XPath query
+                'Parent text Child text',                       // Result of Crawler::text()
+                'Parent text',                                  // Result of Crawler::innerText()
+                ' Parent text ',                                // Result of Crawler::innerText(false)
+            ],
+            [
+                '//*[@id="complex-elements"]/*[@class="two"]',
+                'Child text Parent text',
+                'Parent text',
+                ' ',
+            ],
+            [
+                '//*[@id="complex-elements"]/*[@class="three"]',
+                'Parent text Child text Parent text',
+                'Parent text',
+                ' Parent text ',
+            ],
+            [
+                '//*[@id="complex-elements"]/*[@class="four"]',
+                'Child text',
+                '',
+                '  ',
+            ],
+            [
+                '//*[@id="complex-elements"]/*[@class="five"]',
+                'Child text Another child',
+                '',
+                '  ',
+            ],
+        ];
+    }
+
+    /**
+     * @dataProvider provideInnerTextExamples
+     */
+    public function testInnerText(
+        string $xPathQuery,
+        string $expectedText,
+        string $expectedInnerText,
+        string $expectedInnerTextNormalizeWhitespaceFalse,
+    ) {
+        self::assertCount(1, $crawler = $this->createTestCrawler()->filterXPath($xPathQuery));
 
-        self::assertSame('Parent text Child text', $crawler->text());
-        self::assertSame('Parent text', $crawler->innerText());
+        self::assertSame($expectedText, $crawler->text());
+        self::assertSame($expectedInnerText, $crawler->innerText());
+        self::assertSame($expectedInnerTextNormalizeWhitespaceFalse, $crawler->innerText(false));
     }
 
     public function testHtml()
@@ -1265,9 +1309,12 @@ public function createTestCrawler($uri = null)
                         <div id="child2" xmlns:foo="http://example.com"></div>
                     </div>
                     <div id="sibling"><img /></div>
-                    <div id="complex-element">
-                        Parent text
-                        <span>Child text</span>
+                    <div id="complex-elements">
+                        <div class="one"> Parent text <span>Child text</span> </div>
+                        <div class="two"> <span>Child text</span> Parent text </div>
+                        <div class="three"> Parent text <span>Child text</span> Parent text </div>
+                        <div class="four">  <span>Child text</span>  </div>
+                        <div class="five"><span>Child text</span>  <span>Another child</span></div>
                     </div>
                 </body>
             </html>
