linguise / src / HTMLHelper.php

HTMLHelper.php in Linguise – AI Automatic Multilingual Translation trunk, at src/HTMLHelper.php

243 lines 8.4 KB

Raw Download Zip

1	<?php
2
3	namespace Linguise\WordPress;
4
5	defined('ABSPATH') \|\| die('');
6
7	/**
8	* A collection of HTML helper functions used by Linguise
9	*
10	* Mostly used to parse HTML content and protect HTML entities
11	* for processing by the Fragment handler system
12	*/
13	class HTMLHelper
14	{
15	/**
16	* Marker used to protect the HTML entities
17	*
18	* @var array
19	*/
20	protected static $marker_entity = [
21	'common' => 'linguise-internal-entity',
22	'named' => 'linguise-internal-entity1',
23	'numeric' => 'linguise-internal-entity2',
24	];
25
26	/**
27	* Protect the HTML entities in the source code.
28	*
29	* Adapted from: https://github.com/ivopetkov/html5-dom-document-php/blob/master/src/HTML5DOMDocument.php
30	*
31	* @param string $source The source code to be protected
32	*
33	* @return string The protected source code
34	*/
35	public static function protectEntity($source)
36	{
37	// Replace the entity with our own
38	$source = preg_replace('/&([a-zA-Z]*);/', self::$marker_entity['named'] . '-$1-end', $source);
39	$source = preg_replace('/&#([0-9]*);/', self::$marker_entity['numeric'] . '-$1-end', $source);
40
41	return $source;
42	}
43
44	/**
45	* Unprotect the HTML entities in the source code.
46	*
47	* @param string $html The HTML code to be unprotected
48	*
49	* @return string The unprotected HTML code
50	*/
51	public static function unprotectEntity($html)
52	{
53	if (strpos($html, self::$marker_entity['common']) !== false) {
54	$html = preg_replace('/' . self::$marker_entity['named'] . '-(.*?)-end/', '&$1;', $html);
55	$html = preg_replace('/' . self::$marker_entity['numeric'] . '-(.*?)-end/', '&#$1;', $html);
56	}
57
58	return $html;
59	}
60
61
62	/**
63	* Protect the HTML string before processing with DOMDocument.
64	*
65	* It does:
66	* - Add CDATA around script tags content
67	* - Preserve html entities
68	*
69	* Adapted from: https://github.com/ivopetkov/html5-dom-document-php/blob/master/src/HTML5DOMDocument.php
70	*
71	* @param string $source The HTML source code to be protected
72	*
73	* @return string The protected HTML source code
74	*/
75	private static function protectHTML($source)
76	{
77	// Add CDATA around script tags content
78	$matches = null;
79	preg_match_all('/<script(.*?)>/', $source, $matches);
80	if (isset($matches[0])) {
81	$matches[0] = array_unique($matches[0]);
82	foreach ($matches[0] as $match) {
83	if (substr($match, -2, 1) !== '/') { // check if ends with />
84	$source = str_replace($match, $match . '<![CDATA[-linguise-dom-internal-cdata', $source); // Add CDATA after the open tag
85	}
86	}
87	}
88
89	$source = str_replace('</script>', '-linguise-dom-internal-cdata]]></script>', $source); // Add CDATA before the end tag
90	$source = str_replace('<![CDATA[-linguise-dom-internal-cdata-linguise-dom-internal-cdata]]>', '', $source); // Clean empty script tags
91	$matches = null;
92	preg_match_all('/\<!\[CDATA\[-linguise-dom-internal-cdata.*?-linguise-dom-internal-cdata\]\]>/s', $source, $matches);
93	if (isset($matches[0])) {
94	$matches[0] = array_unique($matches[0]);
95	foreach ($matches[0] as $match) {
96	if (strpos($match, '</') !== false) { // check if contains </
97	$source = str_replace($match, str_replace('</', '<-linguise-dom-internal-cdata-endtagfix/', $match), $source);
98	}
99	}
100	}
101
102	// Preserve html entities
103	$source = self::protectEntity($source);
104
105	return $source;
106	}
107
108
109	/**
110	* Unclobber the CDATA internal
111	*
112	* NOTE: Only use this when processing a script content internal data not the whole HTML data.
113	*
114	* This is used to protect the CDATA internal data from being mangled by the DOMDocument.
115	*
116	* @param string $html_data The HTML data to be unclobbered
117	*
118	* @return string The unclobbered HTML data
119	*/
120	public static function unclobberCdataInternal($html_data)
121	{
122	// Unclobber the CDATA internal
123	$html_data = str_replace('<![CDATA[-linguise-dom-internal-cdata', '', $html_data);
124	$html_data = str_replace('-linguise-dom-internal-cdata]]>', '', $html_data);
125	$html_data = str_replace('<-linguise-dom-internal-cdata-endtagfix/', '</', $html_data);
126
127	return $html_data;
128	}
129
130	/**
131	* Clobber back the CDATA internal
132	*
133	* NOTE: Only use this when processing a script content internal data not the whole HTML data.
134	*
135	* This is used to protect the CDATA internal data from being mangled by the DOMDocument.
136	*
137	* @param string $html_data The HTML data to be clobbered
138	*
139	* @return string The clobbered HTML data
140	*/
141	public static function clobberCdataInternal($html_data)
142	{
143	// Append prefix and suffix
144	return '<![CDATA[-linguise-dom-internal-cdata' . $html_data . '-linguise-dom-internal-cdata]]>';
145	}
146
147	/**
148	* Load the HTML data into a DOMDocument object.
149	*
150	* @param string $html_data The HTML data to be loaded
151	*
152	* @return \DOMDocument\|null The loaded HTML DOM object
153	*/
154	public static function loadHTML($html_data)
155	{
156	// Check if DOMDocument is available or xml extension is loaded
157	if (!class_exists('\\DOMDocument') \|\| !extension_loaded('xml')) {
158	// @codeCoverageIgnoreStart
159	return null;
160	// @codeCoverageIgnoreEnd
161	}
162	if (empty($html_data)) {
163	return null;
164	}
165
166	// Load HTML
167	$html_dom = new \DOMDocument();
168	@$html_dom->loadHTML(self::protectHTML($html_data), LIBXML_HTML_NOIMPLIED \| LIBXML_HTML_NODEFDTD); // phpcs:ignore Generic.PHP.NoSilencedErrors.Discouraged
169
170	/**
171	* Avoid mangling the CSS and JS code with encoded HTML entities
172	*
173	* See: https://www.php.net/manual/en/domdocument.savehtml.php#119813
174	*
175	* While testing, we found this issue with css inner text got weirdly mangled
176	* following the issue above, I manage to correct this by adding proper content-type equiv
177	* which is really weird honestly but it manages to fix the issue.
178	*/
179	$has_utf8 = false;
180	$meta_attrs = $html_dom->getElementsByTagName('meta');
181	foreach ($meta_attrs as $meta) {
182	if ($meta->hasAttribute('http-equiv') && strtolower($meta->getAttribute('http-equiv')) === 'content-type') {
183	// force UTF-8s
184	$meta->setAttribute('content', 'text/html; charset=UTF-8');
185	$has_utf8 = true;
186	break;
187	}
188	}
189
190	if (!$has_utf8) {
191	// We didn't found any meta tag with content-type equiv, add our own
192	$meta = $html_dom->createElement('meta');
193	$meta->setAttribute('http-equiv', 'Content-Type');
194	$meta->setAttribute('content', 'text/html; charset=UTF-8');
195	$head_doc = $html_dom->getElementsByTagName('head');
196	if ($head_doc->length > 0) {
197	// Add to head tag on the child as the first node
198	$head = $head_doc->item(0);
199	@$head->insertBefore($meta, $head->firstChild); // phpcs:ignore Generic.PHP.NoSilencedErrors.Discouraged -- ignore any errors for now
200	}
201	}
202
203	return $html_dom;
204	}
205
206	/**
207	* Save the HTML data into a string.
208	*
209	* @param \DOMDocument $dom The DOMDocument object to be saved
210	*
211	* @return string The saved HTML data
212	*/
213	public static function saveHTML($dom)
214	{
215	// Save HTML
216	$html_data = $dom->saveHTML();
217	if ($html_data === false) {
218	// @codeCoverageIgnoreStart
219	return '';
220	// @codeCoverageIgnoreEnd
221	}
222
223	// Unprotect HTML entities
224	$html_data = self::unprotectEntity($html_data);
225
226	// Unprotect HTML
227	$code_to_be_removed = [
228	'linguise-dom-internal-content',
229	'<![CDATA[-linguise-dom-internal-cdata',
230	'-linguise-dom-internal-cdata]]>',
231	'-linguise-dom-internal-cdata-endtagfix'
232	];
233	foreach ($code_to_be_removed as $code) {
234	$html_data = str_replace($code, '', $html_data);
235	}
236
237	// Unmangle stuff like &#xE5;
238	$html_data = preg_replace('/&#x([0-9A-Fa-f]+);/', '&#x$1;', $html_data);
239
240	return $html_data;
241	}
242	}
243