From 14871f80669d3f299407ce4d0e35e67fa90bc1a5 Mon Sep 17 00:00:00 2001
From: Andres Rey <andreskrey@gmail.com>
Date: Wed, 5 Sep 2018 21:21:11 +0100
Subject: Update initial parsing and add isWhitespace trait function.

---
 src/Nodes/NodeTrait.php | 10 ++++++++--
 src/Readability.php     | 25 ++++++++++++++++++++++++-
 2 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/Nodes/NodeTrait.php b/src/Nodes/NodeTrait.php
index b4ca746..7661e15 100644
--- a/src/Nodes/NodeTrait.php
+++ b/src/Nodes/NodeTrait.php
@@ -473,9 +473,9 @@ trait NodeTrait
      */
     public function isPhrasingContent()
     {
-        return $this->nodeType === XML_TEXT_NODE || !in_array($this->tagName, $this->phrasing_elems) ||
+        return $this->nodeType === XML_TEXT_NODE || !in_array($this->nodeName , $this->phrasing_elems) ||
             (!is_null($this->childNodes) &&
-                ($this->tagName === 'a' || $this->tagName === 'del' || $this->tagName === 'ins') &&
+                ($this->nodeName === 'a' || $this->nodeName === 'del' || $this->nodeName === 'ins') &&
                 array_reduce(iterator_to_array($this->childNodes), function ($carry, $node) {
                     return $carry || $node->isPhrasingContent();
                 })
@@ -493,4 +493,10 @@ trait NodeTrait
 
         return !$this->hasAttribute('hidden');
     }
+
+    public function isWhitespace()
+    {
+        return ($this->nodeType === XML_TEXT_NODE && mb_strlen(trim($this->textContent)) === 0) ||
+            ($this->nodeType === XML_ELEMENT_NODE && $this->nodeName === 'br');
+    }
 }
diff --git a/src/Readability.php b/src/Readability.php
index f913d80..3e168fe 100644
--- a/src/Readability.php
+++ b/src/Readability.php
@@ -671,13 +671,36 @@ class Readability
 
             // Turn all divs that don't have children block level elements into p's
             if ($node->nodeName === 'div') {
+                // Put phrasing content into paragraphs.
+                $p = null;
+                $childNode = $node->firstChild;
+                while ($childNode) {
+                    $nextSibling = $childNode->nextSibling;
+                    if ($childNode->isPhrasingContent()) {
+                        if ($p !== null) {
+                            $p->appendChild($childNode);
+                        } else if (!$childNode->isWhitespace()) {
+                            $p = $this->dom->createElement('p');
+                            $node->replaceChild($p, $childNode);
+                            $p->appendChild($childNode);
+                        }
+                    } else if ($p !== null) {
+                        while ($p->lastChild && $p->lastChild->isWhitespace()) {
+                            $p->removeChild($p->lastChild);
+                        }
+                        $p = null;
+                    }
+                    $childNode = $nextSibling;
+                }
+
+
                 /*
                  * Sites like http://mobile.slate.com encloses each paragraph with a DIV
                  * element. DIVs with only a P element inside and no text content can be
                  * safely converted into plain P elements to avoid confusing the scoring
                  * algorithm with DIVs with are, in practice, paragraphs.
                  */
-                if ($node->hasSingleTagInsideElement('p')) {
+                if ($node->hasSingleTagInsideElement('p') && $node->getLinkDensity() < 0.25) {
                     $this->logger->debug(sprintf('[Get Nodes] Found DIV with a single P node, removing DIV. Node content is: \'%s\'', substr($node->nodeValue, 0, 128)));
                     $pNode = $node->getChildren(true)[0];
                     $node->parentNode->replaceChild($pNode, $node);
-- 
cgit v1.2.3