|
1
|
|
|
<?php
|
|
2
|
|
|
|
|
3
|
|
|
/*
|
|
4
|
|
|
* Copyright (c) 2003 Jose Solorzano. All rights reserved.
|
|
5
|
|
|
* Redistribution of source must retain this copyright notice.
|
|
6
|
|
|
*
|
|
7
|
|
|
* Jose Solorzano (http://jexpert.us) is a software consultant.
|
|
8
|
|
|
*
|
|
9
|
|
|
* Contributions by:
|
|
10
|
|
|
* - Leo West (performance improvements)
|
|
11
|
|
|
*/
|
|
12
|
|
|
|
|
13
|
|
|
define ("NODE_TYPE_START",0);
|
|
14
|
|
|
define ("NODE_TYPE_ELEMENT",1);
|
|
15
|
|
|
define ("NODE_TYPE_ENDELEMENT",2);
|
|
16
|
|
|
define ("NODE_TYPE_TEXT",3);
|
|
17
|
|
|
define ("NODE_TYPE_COMMENT",4);
|
|
18
|
|
|
define ("NODE_TYPE_DONE",5);
|
|
19
|
|
|
|
|
20
|
|
|
/**
|
|
21
|
|
|
* Class HtmlParser.
|
|
22
|
|
|
* To use, create an instance of the class passing
|
|
23
|
|
|
* HTML text. Then invoke parse() until it's false.
|
|
24
|
|
|
* When parse() returns true, $iNodeType, $iNodeName
|
|
25
|
|
|
* $iNodeValue and $iNodeAttributes are updated.
|
|
26
|
|
|
*
|
|
27
|
|
|
* To create an HtmlParser instance you may also
|
|
28
|
|
|
* use convenience functions HtmlParser_ForFile
|
|
29
|
|
|
* and HtmlParser_ForURL.
|
|
30
|
|
|
*/
|
|
31
|
|
|
class HtmlParser {
|
|
32
|
|
|
|
|
33
|
|
|
/**
|
|
34
|
|
|
* Field iNodeType.
|
|
35
|
|
|
* May be one of the NODE_TYPE_* constants above.
|
|
36
|
|
|
*/
|
|
37
|
|
|
var $iNodeType;
|
|
38
|
|
|
|
|
39
|
|
|
/**
|
|
40
|
|
|
* Field iNodeName.
|
|
41
|
|
|
* For elements, it's the name of the element.
|
|
42
|
|
|
*/
|
|
43
|
|
|
var $iNodeName = "";
|
|
44
|
|
|
|
|
45
|
|
|
/**
|
|
46
|
|
|
* Field iNodeValue.
|
|
47
|
|
|
* For text nodes, it's the text.
|
|
48
|
|
|
*/
|
|
49
|
|
|
var $iNodeValue = "";
|
|
50
|
|
|
|
|
51
|
|
|
/**
|
|
52
|
|
|
* Field iNodeAttributes.
|
|
53
|
|
|
* A string-indexed array containing attribute values
|
|
54
|
|
|
* of the current node. Indexes are always lowercase.
|
|
55
|
|
|
*/
|
|
56
|
|
|
var $iNodeAttributes;
|
|
57
|
|
|
|
|
58
|
|
|
// The following fields should be
|
|
59
|
|
|
// considered private:
|
|
60
|
|
|
|
|
61
|
|
|
var $iHtmlText;
|
|
62
|
|
|
var $iHtmlTextLength;
|
|
63
|
|
|
var $iHtmlTextIndex = 0;
|
|
64
|
|
|
var $iHtmlCurrentChar;
|
|
65
|
|
|
var $BOE_ARRAY;
|
|
66
|
|
|
var $B_ARRAY;
|
|
67
|
|
|
var $BOS_ARRAY;
|
|
68
|
|
|
|
|
69
|
|
|
/**
|
|
70
|
|
|
* Constructor.
|
|
71
|
|
|
* Constructs an HtmlParser instance with
|
|
72
|
|
|
* the HTML text given.
|
|
73
|
|
|
*/
|
|
74
|
|
|
function HtmlParser ($aHtmlText) {
|
|
75
|
|
|
$this->iHtmlText = $aHtmlText;
|
|
76
|
|
|
$this->iHtmlTextLength = strlen($aHtmlText);
|
|
77
|
|
|
$this->iNodeAttributes = array();
|
|
78
|
|
|
$this->setTextIndex (0);
|
|
79
|
|
|
|
|
80
|
|
|
$this->BOE_ARRAY = array (" ", "\t", "\r", "\n", "=" );
|
|
81
|
|
|
$this->B_ARRAY = array (" ", "\t", "\r", "\n" );
|
|
82
|
|
|
$this->BOS_ARRAY = array (" ", "\t", "\r", "\n", "/" );
|
|
83
|
|
|
}
|
|
84
|
|
|
|
|
85
|
|
|
/**
|
|
86
|
|
|
* Method parse.
|
|
87
|
|
|
* Parses the next node. Returns false only if
|
|
88
|
|
|
* the end of the HTML text has been reached.
|
|
89
|
|
|
* Updates values of iNode* fields.
|
|
90
|
|
|
*/
|
|
91
|
|
|
function parse() {
|
|
92
|
|
|
$text = $this->skipToElement();
|
|
93
|
|
View Code Duplication |
if ($text != "") {
|
|
94
|
|
|
$this->iNodeType = NODE_TYPE_TEXT;
|
|
95
|
|
|
$this->iNodeName = "Text";
|
|
96
|
|
|
$this->iNodeValue = $text;
|
|
97
|
|
|
return true;
|
|
98
|
|
|
}
|
|
99
|
|
|
return $this->readTag();
|
|
100
|
|
|
}
|
|
101
|
|
|
|
|
102
|
|
|
function clearAttributes() {
|
|
103
|
|
|
$this->iNodeAttributes = array();
|
|
104
|
|
|
}
|
|
105
|
|
|
|
|
106
|
|
|
function readTag() {
|
|
107
|
|
|
if ($this->iCurrentChar != "<") {
|
|
|
|
|
|
|
108
|
|
|
$this->iNodeType = NODE_TYPE_DONE;
|
|
109
|
|
|
return false;
|
|
110
|
|
|
}
|
|
111
|
|
|
$this->clearAttributes();
|
|
112
|
|
|
$this->skipMaxInTag ("<", 1);
|
|
113
|
|
|
if ($this->iCurrentChar == '/') {
|
|
114
|
|
|
$this->moveNext();
|
|
115
|
|
|
$name = $this->skipToBlanksInTag();
|
|
116
|
|
|
$this->iNodeType = NODE_TYPE_ENDELEMENT;
|
|
117
|
|
|
$this->iNodeName = $name;
|
|
118
|
|
|
$this->iNodeValue = "";
|
|
119
|
|
|
$this->skipEndOfTag();
|
|
120
|
|
|
return true;
|
|
121
|
|
|
}
|
|
122
|
|
|
$name = $this->skipToBlanksOrSlashInTag();
|
|
123
|
|
|
if (!$this->isValidTagIdentifier ($name)) {
|
|
124
|
|
|
$comment = false;
|
|
125
|
|
|
if (strpos($name, "!--") === 0) {
|
|
126
|
|
|
$ppos = strpos($name, "--", 3);
|
|
|
|
|
|
|
127
|
|
|
if (strpos($name, "--", 3) === (strlen($name) - 2)) {
|
|
128
|
|
|
$this->iNodeType = NODE_TYPE_COMMENT;
|
|
129
|
|
|
$this->iNodeName = "Comment";
|
|
130
|
|
|
$this->iNodeValue = "<" . $name . ">";
|
|
131
|
|
|
$comment = true;
|
|
132
|
|
|
}
|
|
133
|
|
|
else {
|
|
134
|
|
|
$rest = $this->skipToStringInTag ("-->");
|
|
135
|
|
|
if ($rest != "") {
|
|
136
|
|
|
$this->iNodeType = NODE_TYPE_COMMENT;
|
|
137
|
|
|
$this->iNodeName = "Comment";
|
|
138
|
|
|
$this->iNodeValue = "<" . $name . $rest;
|
|
139
|
|
|
$comment = true;
|
|
|
|
|
|
|
140
|
|
|
// Already skipped end of tag
|
|
141
|
|
|
return true;
|
|
142
|
|
|
}
|
|
143
|
|
|
}
|
|
144
|
|
|
}
|
|
145
|
|
View Code Duplication |
if (!$comment) {
|
|
146
|
|
|
$this->iNodeType = NODE_TYPE_TEXT;
|
|
147
|
|
|
$this->iNodeName = "Text";
|
|
148
|
|
|
$this->iNodeValue = "<" . $name;
|
|
149
|
|
|
return true;
|
|
150
|
|
|
}
|
|
151
|
|
|
}
|
|
152
|
|
|
else {
|
|
153
|
|
|
$this->iNodeType = NODE_TYPE_ELEMENT;
|
|
154
|
|
|
$this->iNodeValue = "";
|
|
155
|
|
|
$this->iNodeName = $name;
|
|
156
|
|
|
while ($this->skipBlanksInTag()) {
|
|
157
|
|
|
$attrName = $this->skipToBlanksOrEqualsInTag();
|
|
158
|
|
|
if ($attrName != "" && $attrName != "/") {
|
|
159
|
|
|
$this->skipBlanksInTag();
|
|
160
|
|
|
if ($this->iCurrentChar == "=") {
|
|
161
|
|
|
$this->skipEqualsInTag();
|
|
162
|
|
|
$this->skipBlanksInTag();
|
|
163
|
|
|
$value = $this->readValueInTag();
|
|
164
|
|
|
$this->iNodeAttributes[strtolower($attrName)] = $value;
|
|
165
|
|
|
}
|
|
166
|
|
|
else {
|
|
167
|
|
|
$this->iNodeAttributes[strtolower($attrName)] = "";
|
|
168
|
|
|
}
|
|
169
|
|
|
}
|
|
170
|
|
|
}
|
|
171
|
|
|
}
|
|
172
|
|
|
$this->skipEndOfTag();
|
|
173
|
|
|
return true;
|
|
174
|
|
|
}
|
|
175
|
|
|
|
|
176
|
|
|
function isValidTagIdentifier ($name) {
|
|
177
|
|
|
return ereg ("^[A-Za-z0-9_\\-]+$", $name);
|
|
178
|
|
|
}
|
|
179
|
|
|
|
|
180
|
|
|
function skipBlanksInTag() {
|
|
181
|
|
|
return "" != ($this->skipInTag ($this->B_ARRAY));
|
|
182
|
|
|
}
|
|
183
|
|
|
|
|
184
|
|
|
function skipToBlanksOrEqualsInTag() {
|
|
185
|
|
|
return $this->skipToInTag ($this->BOE_ARRAY);
|
|
186
|
|
|
}
|
|
187
|
|
|
|
|
188
|
|
|
function skipToBlanksInTag() {
|
|
189
|
|
|
return $this->skipToInTag ($this->B_ARRAY);
|
|
190
|
|
|
}
|
|
191
|
|
|
|
|
192
|
|
|
function skipToBlanksOrSlashInTag() {
|
|
193
|
|
|
return $this->skipToInTag ($this->BOS_ARRAY);
|
|
194
|
|
|
}
|
|
195
|
|
|
|
|
196
|
|
|
function skipEqualsInTag() {
|
|
197
|
|
|
return $this->skipMaxInTag ("=", 1);
|
|
198
|
|
|
}
|
|
199
|
|
|
|
|
200
|
|
|
function readValueInTag() {
|
|
201
|
|
|
$ch = $this->iCurrentChar;
|
|
202
|
|
|
$value = "";
|
|
|
|
|
|
|
203
|
|
|
if ($ch == "\"") {
|
|
204
|
|
|
$this->skipMaxInTag ("\"", 1);
|
|
205
|
|
|
$value = $this->skipToInTag ("\"");
|
|
206
|
|
|
$this->skipMaxInTag ("\"", 1);
|
|
207
|
|
|
}
|
|
208
|
|
|
else if ($ch == "'") {
|
|
209
|
|
|
$this->skipMaxInTag ("'", 1);
|
|
210
|
|
|
$value = $this->skipToInTag ("'");
|
|
211
|
|
|
$this->skipMaxInTag ("'", 1);
|
|
212
|
|
|
}
|
|
213
|
|
|
else {
|
|
214
|
|
|
$value = $this->skipToBlanksInTag();
|
|
215
|
|
|
}
|
|
216
|
|
|
return $value;
|
|
217
|
|
|
}
|
|
218
|
|
|
|
|
219
|
|
|
function setTextIndex ($index) {
|
|
220
|
|
|
$this->iHtmlTextIndex = $index;
|
|
221
|
|
|
if ($index >= $this->iHtmlTextLength) {
|
|
222
|
|
|
$this->iCurrentChar = -1;
|
|
223
|
|
|
}
|
|
224
|
|
|
else {
|
|
225
|
|
|
$this->iCurrentChar = $this->iHtmlText{$index};
|
|
226
|
|
|
}
|
|
227
|
|
|
}
|
|
228
|
|
|
|
|
229
|
|
|
function moveNext() {
|
|
230
|
|
|
if ($this->iHtmlTextIndex < $this->iHtmlTextLength) {
|
|
231
|
|
|
$this->setTextIndex ($this->iHtmlTextIndex + 1);
|
|
232
|
|
|
return true;
|
|
233
|
|
|
}
|
|
234
|
|
|
else {
|
|
235
|
|
|
return false;
|
|
236
|
|
|
}
|
|
237
|
|
|
}
|
|
238
|
|
|
|
|
239
|
|
|
function skipEndOfTag() {
|
|
240
|
|
|
while (($ch = $this->iCurrentChar) !== -1) {
|
|
241
|
|
|
if ($ch == ">") {
|
|
242
|
|
|
$this->moveNext();
|
|
243
|
|
|
return;
|
|
244
|
|
|
}
|
|
245
|
|
|
$this->moveNext();
|
|
246
|
|
|
}
|
|
247
|
|
|
}
|
|
248
|
|
|
|
|
249
|
|
View Code Duplication |
function skipInTag ($chars) {
|
|
|
|
|
|
|
250
|
|
|
$sb = "";
|
|
251
|
|
|
while (($ch = $this->iCurrentChar) !== -1) {
|
|
252
|
|
|
if ($ch == ">") {
|
|
253
|
|
|
return $sb;
|
|
254
|
|
|
} else {
|
|
255
|
|
|
$match = false;
|
|
256
|
|
|
for ($idx = 0; $idx < count($chars); $idx++) {
|
|
|
|
|
|
|
257
|
|
|
if ($ch == $chars[$idx]) {
|
|
258
|
|
|
$match = true;
|
|
259
|
|
|
break;
|
|
260
|
|
|
}
|
|
261
|
|
|
}
|
|
262
|
|
|
if (!$match) {
|
|
263
|
|
|
return $sb;
|
|
264
|
|
|
}
|
|
265
|
|
|
$sb .= $ch;
|
|
266
|
|
|
$this->moveNext();
|
|
267
|
|
|
}
|
|
268
|
|
|
}
|
|
269
|
|
|
return $sb;
|
|
270
|
|
|
}
|
|
271
|
|
|
|
|
272
|
|
|
function skipMaxInTag ($chars, $maxChars) {
|
|
273
|
|
|
$sb = "";
|
|
274
|
|
|
$count = 0;
|
|
275
|
|
|
while (($ch = $this->iCurrentChar) !== -1 && $count++ < $maxChars) {
|
|
276
|
|
|
if ($ch == ">") {
|
|
277
|
|
|
return $sb;
|
|
278
|
|
|
} else {
|
|
279
|
|
|
$match = false;
|
|
280
|
|
|
for ($idx = 0; $idx < count($chars); $idx++) {
|
|
|
|
|
|
|
281
|
|
|
if ($ch == $chars[$idx]) {
|
|
282
|
|
|
$match = true;
|
|
283
|
|
|
break;
|
|
284
|
|
|
}
|
|
285
|
|
|
}
|
|
286
|
|
|
if (!$match) {
|
|
287
|
|
|
return $sb;
|
|
288
|
|
|
}
|
|
289
|
|
|
$sb .= $ch;
|
|
290
|
|
|
$this->moveNext();
|
|
291
|
|
|
}
|
|
292
|
|
|
}
|
|
293
|
|
|
return $sb;
|
|
294
|
|
|
}
|
|
295
|
|
|
|
|
296
|
|
View Code Duplication |
function skipToInTag ($chars) {
|
|
|
|
|
|
|
297
|
|
|
$sb = "";
|
|
298
|
|
|
while (($ch = $this->iCurrentChar) !== -1) {
|
|
299
|
|
|
$match = $ch == ">";
|
|
300
|
|
|
if (!$match) {
|
|
301
|
|
|
for ($idx = 0; $idx < count($chars); $idx++) {
|
|
|
|
|
|
|
302
|
|
|
if ($ch == $chars[$idx]) {
|
|
303
|
|
|
$match = true;
|
|
304
|
|
|
break;
|
|
305
|
|
|
}
|
|
306
|
|
|
}
|
|
307
|
|
|
}
|
|
308
|
|
|
if ($match) {
|
|
309
|
|
|
return $sb;
|
|
310
|
|
|
}
|
|
311
|
|
|
$sb .= $ch;
|
|
312
|
|
|
$this->moveNext();
|
|
313
|
|
|
}
|
|
314
|
|
|
return $sb;
|
|
315
|
|
|
}
|
|
316
|
|
|
|
|
317
|
|
|
function skipToElement() {
|
|
318
|
|
|
$sb = "";
|
|
319
|
|
|
while (($ch = $this->iCurrentChar) !== -1) {
|
|
320
|
|
|
if ($ch == "<") {
|
|
321
|
|
|
return $sb;
|
|
322
|
|
|
}
|
|
323
|
|
|
$sb .= $ch;
|
|
324
|
|
|
$this->moveNext();
|
|
325
|
|
|
}
|
|
326
|
|
|
return $sb;
|
|
327
|
|
|
}
|
|
328
|
|
|
|
|
329
|
|
|
/**
|
|
330
|
|
|
* Returns text between current position and $needle,
|
|
331
|
|
|
* inclusive, or "" if not found. The current index is moved to a point
|
|
332
|
|
|
* after the location of $needle, or not moved at all
|
|
333
|
|
|
* if nothing is found.
|
|
334
|
|
|
*/
|
|
335
|
|
|
function skipToStringInTag ($needle) {
|
|
336
|
|
|
$pos = strpos ($this->iHtmlText, $needle, $this->iHtmlTextIndex);
|
|
337
|
|
|
if ($pos === false) {
|
|
338
|
|
|
return "";
|
|
339
|
|
|
}
|
|
340
|
|
|
$top = $pos + strlen($needle);
|
|
341
|
|
|
$retvalue = substr ($this->iHtmlText, $this->iHtmlTextIndex, $top - $this->iHtmlTextIndex);
|
|
342
|
|
|
$this->setTextIndex ($top);
|
|
343
|
|
|
return $retvalue;
|
|
344
|
|
|
}
|
|
345
|
|
|
}
|
|
346
|
|
|
|
|
347
|
|
|
function HtmlParser_ForFile ($fileName) {
|
|
348
|
|
|
return HtmlParser_ForURL($fileName);
|
|
349
|
|
|
}
|
|
350
|
|
|
|
|
351
|
|
|
function HtmlParser_ForURL ($url) {
|
|
352
|
|
|
$fp = fopen ($url, "r");
|
|
353
|
|
|
$content = "";
|
|
354
|
|
|
while (true) {
|
|
355
|
|
|
$data = fread ($fp, 8192);
|
|
356
|
|
|
if (strlen($data) == 0) {
|
|
357
|
|
|
break;
|
|
358
|
|
|
}
|
|
359
|
|
|
$content .= $data;
|
|
360
|
|
|
}
|
|
361
|
|
|
fclose ($fp);
|
|
362
|
|
|
return new HtmlParser ($content);
|
|
363
|
|
|
}
|
|
364
|
|
|
|
|
365
|
|
|
php?>
|
|
|
|
|
|
|
366
|
|
|
|
In PHP it is possible to write to properties without declaring them. For example, the following is perfectly valid PHP code:
Generally, it is a good practice to explictly declare properties to avoid accidental typos and provide IDE auto-completion: