4get/lib/fuckhtml.php

611 lines
11 KiB
PHP
Raw Normal View History

2023-07-22 18:41:14 +00:00
<?php
class fuckhtml{
public function __construct($html = null, $isfile = false){
if($html !== null){
$this->load($html, $isfile);
}
}
public function load($html, $isfile = false){
if(is_array($html)){
if(!isset($html["innerHTML"])){
throw new Exception("(load) Supplied array doesn't contain an innerHTML index");
2023-07-22 18:41:14 +00:00
}
$html = $html["innerHTML"];
}
if($isfile){
$handle = fopen($html, "r");
$fetch = fread($handle, filesize($html));
fclose($handle);
$this->html = $fetch;
}else{
$this->html = $html;
}
$this->strlen = strlen($this->html);
}
public function getloadedhtml(){
return $this->html;
}
2023-07-22 18:41:14 +00:00
public function getElementsByTagName(string $tagname){
$out = [];
/*
Scrape start of the tag. Example
<div class="mydiv"> ...
*/
if($tagname == "*"){
$tagname = '[A-Za-z0-9._-]+';
2023-07-22 18:41:14 +00:00
}else{
$tagname = preg_quote(strtolower($tagname));
}
preg_match_all(
'/<\s*(' . $tagname . ')(\s(?:[^>\'"]*|"[^"]*"|\'[^\']*\')+)?\s*>/i',
/* '/<\s*(' . $tagname . ')(\s[\S\s]*?)?>/i', */
$this->html,
$starting_tags,
PREG_OFFSET_CAPTURE
);
for($i=0; $i<count($starting_tags[0]); $i++){
/*
Parse attributes
*/
$attributes = [];
preg_match_all(
2024-03-25 02:31:19 +00:00
'/([^\/\s\\=]+)(?:\s*=\s*("[^"]*"|\'[^\']*\'|[^\s]*))?/i',
2023-07-22 18:41:14 +00:00
$starting_tags[2][$i][0],
$regex_attributes
);
for($k=0; $k<count($regex_attributes[0]); $k++){
if(trim($regex_attributes[2][$k]) == ""){
$attributes[$regex_attributes[1][$k]] =
"true";
continue;
}
2024-03-25 02:31:19 +00:00
$attributes[strtolower($regex_attributes[1][$k])] =
2023-07-22 18:41:14 +00:00
trim($regex_attributes[2][$k], "'\" \n\r\t\v\x00");
}
$out[] = [
"tagName" => strtolower($starting_tags[1][$i][0]),
"startPos" => $starting_tags[0][$i][1],
"endPos" => 0,
"startTag" => $starting_tags[0][$i][0],
"attributes" => $attributes,
"innerHTML" => null
];
}
/*
Get innerHTML
*/
// get closing tag positions
preg_match_all(
'/<\s*\/\s*(' . $tagname . ')\s*>/i',
$this->html,
$regex_closing_tags,
PREG_OFFSET_CAPTURE
);
// merge opening and closing tags together
for($i=0; $i<count($regex_closing_tags[1]); $i++){
$out[] = [
"tagName" => strtolower($regex_closing_tags[1][$i][0]),
"endTag" => $regex_closing_tags[0][$i][0],
"startPos" => $regex_closing_tags[0][$i][1]
];
}
usort(
$out,
function($a, $b){
return $a["startPos"] > $b["startPos"];
}
);
// compute the indent level for each element
2023-07-22 18:41:14 +00:00
$level = [];
$count = count($out);
for($i=0; $i<$count; $i++){
if(!isset($level[$out[$i]["tagName"]])){
$level[$out[$i]["tagName"]] = 0;
}
if(isset($out[$i]["startTag"])){
// encountered starting tag
$level[$out[$i]["tagName"]]++;
$out[$i]["level"] = $level[$out[$i]["tagName"]];
}else{
// encountered closing tag
$out[$i]["level"] = $level[$out[$i]["tagName"]];
$level[$out[$i]["tagName"]]--;
}
}
// if the indent level is the same for a div,
// we encountered _THE_ closing tag
for($i=0; $i<$count; $i++){
if(!isset($out[$i]["startTag"])){
continue;
}
for($k=$i; $k<$count; $k++){
if(
isset($out[$k]["endTag"]) &&
$out[$i]["tagName"] == $out[$k]["tagName"] &&
$out[$i]["level"]
=== $out[$k]["level"]
){
$startlen = strlen($out[$i]["startTag"]);
$endlen = strlen($out[$k]["endTag"]);
$out[$i]["endPos"] = $out[$k]["startPos"] + $endlen;
$out[$i]["innerHTML"] =
substr(
$this->html,
$out[$i]["startPos"] + $startlen,
$out[$k]["startPos"] - ($out[$i]["startPos"] + $startlen)
);
$out[$i]["outerHTML"] =
substr(
$this->html,
$out[$i]["startPos"],
$out[$k]["startPos"] - $out[$i]["startPos"] + $endlen
);
break;
}
}
}
// filter out ending divs
for($i=0; $i<$count; $i++){
if(isset($out[$i]["endTag"])){
unset($out[$i]);
}
unset($out[$i]["startTag"]);
}
return array_values($out);
}
public function getElementsByAttributeName(string $name, $collection = null){
if($collection === null){
$collection = $this->getElementsByTagName("*");
}elseif(is_string($collection)){
$collection = $this->getElementsByTagName($collection);
}
$return = [];
foreach($collection as $elem){
foreach($elem["attributes"] as $attrib_name => $attrib_value){
if($attrib_name == $name){
$return[] = $elem;
continue 2;
}
}
}
return $return;
}
public function getElementsByFuzzyAttributeValue(string $name, string $value, $collection = null){
$elems = $this->getElementsByAttributeName($name, $collection);
2024-06-13 02:41:02 +00:00
$value =
explode(
" ",
trim(
preg_replace(
'/ +/',
" ",
$value
)
)
);
2023-07-22 18:41:14 +00:00
$return = [];
foreach($elems as $elem){
foreach($elem["attributes"] as $attrib_name => $attrib_value){
$attrib_value = explode(" ", $attrib_value);
$ac = count($attrib_value);
$nc = count($value);
$cr = 0;
for($i=0; $i<$nc; $i++){
for($k=0; $k<$ac; $k++){
if($value[$i] == $attrib_value[$k]){
$cr++;
}
}
}
if($cr === $nc){
$return[] = $elem;
continue 2;
}
}
}
return $return;
}
public function getElementsByAttributeValue(string $name, string $value, $collection = null){
$elems = $this->getElementsByAttributeName($name, $collection);
$return = [];
foreach($elems as $elem){
foreach($elem["attributes"] as $attrib_name => $attrib_value){
if($attrib_value == $value){
$return[] = $elem;
continue 2;
}
}
}
return $return;
}
public function getElementById(string $idname, $collection = null){
$id = $this->getElementsByAttributeValue("id", $idname, $collection);
if(count($id) !== 0){
return $id[0];
}
return false;
}
public function getElementsByClassName(string $classname, $collection = null){
return $this->getElementsByFuzzyAttributeValue("class", $classname, $collection);
}
public function getTextContent($html, $whitespace = false, $trim = true){
if(is_array($html)){
if(!isset($html["innerHTML"])){
throw new Exception("(getTextContent) Supplied array doesn't contain an innerHTML index");
2023-07-22 18:41:14 +00:00
}
2023-07-22 18:41:14 +00:00
$html = $html["innerHTML"];
}
$html = preg_split('/\n|<\/?br>/i', $html);
2023-07-22 18:41:14 +00:00
$out = "";
for($i=0; $i<count($html); $i++){
$tmp =
html_entity_decode(
strip_tags(
$html[$i]
),
ENT_QUOTES | ENT_XML1, "UTF-8"
);
if($trim){
$tmp = trim($tmp);
}
$out .= $tmp;
if($whitespace === true){
$out .= "\n";
}else{
$out .= " ";
}
}
if($trim){
return trim($out);
}
return $out;
}
2023-08-08 07:09:47 +00:00
public function parseJsObject(string $json){
$bracket = false;
$is_close_bracket = false;
$escape = false;
2023-12-11 15:25:05 +00:00
$lastchar = false;
2023-08-08 07:09:47 +00:00
$json_out = null;
$last_char = null;
2024-10-30 01:29:17 +00:00
$keyword_check = null;
2023-08-08 07:09:47 +00:00
for($i=0; $i<strlen($json); $i++){
switch($json[$i]){
case "\"":
case "'":
if($escape === true){
break;
}
if($json[$i] == $bracket){
$bracket = false;
$is_close_bracket = true;
2024-10-30 01:29:17 +00:00
2023-08-08 07:09:47 +00:00
}else{
if($bracket === false){
$bracket = $json[$i];
}
}
break;
default:
$is_close_bracket = false;
break;
}
2023-12-11 15:25:05 +00:00
if(
$json[$i] == "\\" &&
!(
$lastchar !== false &&
$lastchar . $json[$i] == "\\\\"
)
){
$escape = true;
}else{
$escape = false;
}
2023-08-08 07:09:47 +00:00
if(
$bracket === false &&
$is_close_bracket === false
){
2024-10-30 01:29:17 +00:00
// do keyword check
$keyword_check .= $json[$i];
if(in_array($json[$i], [":", "{"])){
$keyword_check = substr($keyword_check, 0, -1);
if(
preg_match(
'/function|array|return/i',
$keyword_check
)
){
$json_out =
preg_replace(
'/[{"]*' . preg_quote($keyword_check, "/") . '$/',
"",
$json_out
);
}
$keyword_check = null;
}
2023-08-08 07:09:47 +00:00
// here we know we're not iterating over a quoted string
switch($json[$i]){
case "[":
case "{":
$json_out .= $json[$i];
break;
case "]":
case "}":
case ",":
case ":":
if(!in_array($last_char, ["[", "{", "}", "]", "\""])){
$json_out .= "\"";
}
$json_out .= $json[$i];
break;
default:
if(in_array($last_char, ["{", "[", ",", ":"])){
$json_out .= "\"";
}
$json_out .= $json[$i];
break;
}
}else{
$json_out .= $json[$i];
}
$last_char = $json[$i];
}
return json_decode($json_out, true);
}
public function parseJsString($string){
return
preg_replace_callback(
2024-02-18 04:22:19 +00:00
'/\\\u[A-Fa-f0-9]{4}|\\\x[A-Fa-f0-9]{2}|\\\n|\\\r/',
function($match){
2024-02-18 04:22:19 +00:00
switch($match[0][1]){
2024-02-18 04:22:19 +00:00
case "u":
return json_decode('"' . $match[0] . '"');
break;
2024-02-18 04:22:19 +00:00
case "x":
return mb_convert_encoding(
stripcslashes($match[0]),
"utf-8",
"windows-1252"
);
break;
default:
return " ";
break;
}
},
$string
);
}
2024-12-17 05:31:15 +00:00
public function extract_json($json){
$len = strlen($json);
$array_level = 0;
$object_level = 0;
$in_quote = null;
$start = null;
for($i=0; $i<$len; $i++){
switch($json[$i]){
case "[":
if($in_quote === null){
$array_level++;
if($start === null){
$start = $i;
}
}
break;
case "]":
if($in_quote === null){
$array_level--;
}
break;
case "{":
if($in_quote === null){
$object_level++;
if($start === null){
$start = $i;
}
}
break;
case "}":
if($in_quote === null){
$object_level--;
}
break;
case "\"":
case "'":
if(
$i !== 0 &&
$json[$i - 1] !== "\\"
){
// found a non-escaped quote
if($in_quote === null){
// open quote
$in_quote = $json[$i];
}elseif($in_quote === $json[$i]){
// close quote
$in_quote = null;
}
}
break;
}
if(
$start !== null &&
$array_level === 0 &&
$object_level === 0
){
return substr($json, $start, $i - $start + 1);
break;
}
}
}
2023-07-22 18:41:14 +00:00
}