Merge pull request #87 from samipuro/fix/qpxml-pi-prefix

Fix `QPXML::pi($prefix)` always returning NULL

Author: samipuro

CHANGELOG.md

@@ -20,6 +20,7 @@ QueryPath Changelog
 - Fix processing instructions gaining an extra `?` each time an HTML-parsed document was serialized, so `<?php echo $title; ?>` came back out of `html()`, `innerHTML()`, `innerXML()`, `innerXHTML()`, `xml()`, `html5()`, `innerHTML5()`, and `writeXML()` as `<?php echo $title; ??>`. libxml's HTML parser keeps the closing `?` as part of the node's data, unlike its XML parser and the Masterminds HTML5 parser, so QueryPath now strips it on load. `DOMProcessingInstruction::$data` consequently no longer has a stray `?` on the end for documents read with `htmlqp()` or `qp()` on an `.html` file. Note that this normalisation applies only to documents QueryPath parses itself, and that taking the underlying `DOMDocument` out of QueryPath and calling libxml's own `saveHTML()` on it will emit `<?php ... >` without the terminator, since libxml's HTML serializer never adds one
 - With libxml2 2.14+, `htmlqp()` parses `<?php ... ?>` as a comment (`<!--?php ... ?-->`) and no longer throws `QueryPath\ParseException` on malformed HTML. Use `html5qp()` to keep processing instructions
 - Add `QueryPath\Document`, a `DOMDocument` subclass QueryPath parses into. The type is how QueryPath records that a document's processing instruction data does not carry the closing `?`, which cannot be determined by inspecting the document afterwards. It travels with the document, so every route to a second `DOMQuery` over one document -- iteration, `add()`, `remove()`, `replaceAll()`, `branch()`, `QueryPath::with()`, and the bundled extensions -- serializes it correctly. A `DOMDocument` supplied by the caller is a plain `DOMDocument`, makes no such promise, and is still serialized exactly as it was handed over
+- Fix `QPXML::pi($prefix)` always returning `NULL` with an "Undefined property" warning instead of the matching processing instruction's text (issue #79)
 - Fix fatal error when running a CSS selector against a match set that contains non-element nodes (text, comment, CDATA or processing instruction), as `contents()` returns. Those nodes now simply do not match, instead of calling element-only DOM methods on them
 - Fix `find('a > b')` throwing a `TypeError` when the left-hand side matched the root element, whose parent is the document rather than an element
 - Fix the `:text` pseudo-class so it matches jQuery: it selects `input` elements whose `type` attribute is absent or is `text` (case-insensitively). It never indicated, and still does not indicate, whether a node is a text node

src/Extension/QPXML.php

@@ -142,15 +142,8 @@ public function pi($prefix = null, $text = null)
 		}
 		foreach ($this->qp->get() as $ele) {
 			foreach ($ele->childNodes as $node) {
-				if ($node->nodeType == XML_PI_NODE) {
-					if (isset($prefix)) {
-						if ($node->tagName == $prefix) {
-							return $node->textContent;
-						}
-					} else {
-						// Return first match.
-						return $node->textContent;
-					}
+				if ($node->nodeType == XML_PI_NODE && ($prefix === null || $node->target === $prefix)) {
+					return $node->textContent;
 				}
 			} // foreach
 		} // foreach

tests/QueryPath/Extension/QPXMLTest.php

@@ -39,6 +39,16 @@ public function testProcessingInstruction()
 	{
 		$this->assertEquals('This is a processing instruction.', trim(qp($this->file, 'third')->pi()));
 		$msg = "Message";
-		$this->assertEquals($msg, qp($this->file, 'second')->pi('qp', $msg)->top()->find('second')->pi());
+		$this->assertEquals($msg, qp($this->file, 'second')->pi('qp', $msg)->top()->find('second')->pi('qp'));
+	}
+
+	public function testProcessingInstructionByPrefix()
+	{
+		$xml = '<?xml version="1.0"?><root><?first one?><?second two?></root>';
+
+		$qp = qp($xml, 'root');
+
+		$this->assertEquals('two', $qp->pi('second'));
+		$this->assertNull($qp->pi('missing'));
 	}
 }