501 lines
11 KiB
PHP
501 lines
11 KiB
PHP
|
<?php
|
||
|
/**
|
||
|
* SimplePie
|
||
|
*
|
||
|
* A PHP-Based RSS and Atom Feed Framework.
|
||
|
* Takes the hard work out of managing a complete RSS/Atom solution.
|
||
|
*
|
||
|
* Copyright (c) 2004-2012, Ryan Parman, Geoffrey Sneddon, Ryan McCue, and contributors
|
||
|
* All rights reserved.
|
||
|
*
|
||
|
* Redistribution and use in source and binary forms, with or without modification, are
|
||
|
* permitted provided that the following conditions are met:
|
||
|
*
|
||
|
* * Redistributions of source code must retain the above copyright notice, this list of
|
||
|
* conditions and the following disclaimer.
|
||
|
*
|
||
|
* * Redistributions in binary form must reproduce the above copyright notice, this list
|
||
|
* of conditions and the following disclaimer in the documentation and/or other materials
|
||
|
* provided with the distribution.
|
||
|
*
|
||
|
* * Neither the name of the SimplePie Team nor the names of its contributors may be used
|
||
|
* to endorse or promote products derived from this software without specific prior
|
||
|
* written permission.
|
||
|
*
|
||
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS
|
||
|
* OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
|
||
|
* AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDERS
|
||
|
* AND CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
||
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
||
|
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
||
|
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
|
||
|
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
||
|
* POSSIBILITY OF SUCH DAMAGE.
|
||
|
*
|
||
|
* @package SimplePie
|
||
|
* @version 1.3
|
||
|
* @copyright 2004-2012 Ryan Parman, Geoffrey Sneddon, Ryan McCue
|
||
|
* @author Ryan Parman
|
||
|
* @author Geoffrey Sneddon
|
||
|
* @author Ryan McCue
|
||
|
* @link http://simplepie.org/ SimplePie
|
||
|
* @license http://www.opensource.org/licenses/bsd-license.php BSD License
|
||
|
*/
|
||
|
|
||
|
|
||
|
/**
|
||
|
* HTTP Response Parser
|
||
|
*
|
||
|
* @package SimplePie
|
||
|
* @subpackage HTTP
|
||
|
*/
|
||
|
class SimplePie_HTTP_Parser
|
||
|
{
|
||
|
/**
|
||
|
* HTTP Version
|
||
|
*
|
||
|
* @var float
|
||
|
*/
|
||
|
public $http_version = 0.0;
|
||
|
|
||
|
/**
|
||
|
* Status code
|
||
|
*
|
||
|
* @var int
|
||
|
*/
|
||
|
public $status_code = 0;
|
||
|
|
||
|
/**
|
||
|
* Reason phrase
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
public $reason = '';
|
||
|
|
||
|
/**
|
||
|
* Key/value pairs of the headers
|
||
|
*
|
||
|
* @var array
|
||
|
*/
|
||
|
public $headers = array();
|
||
|
|
||
|
/**
|
||
|
* Body of the response
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
public $body = '';
|
||
|
|
||
|
/**
|
||
|
* Current state of the state machine
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
protected $state = 'http_version';
|
||
|
|
||
|
/**
|
||
|
* Input data
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
protected $data = '';
|
||
|
|
||
|
/**
|
||
|
* Input data length (to avoid calling strlen() everytime this is needed)
|
||
|
*
|
||
|
* @var int
|
||
|
*/
|
||
|
protected $data_length = 0;
|
||
|
|
||
|
/**
|
||
|
* Current position of the pointer
|
||
|
*
|
||
|
* @var int
|
||
|
*/
|
||
|
protected $position = 0;
|
||
|
|
||
|
/**
|
||
|
* Name of the hedaer currently being parsed
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
protected $name = '';
|
||
|
|
||
|
/**
|
||
|
* Value of the hedaer currently being parsed
|
||
|
*
|
||
|
* @var string
|
||
|
*/
|
||
|
protected $value = '';
|
||
|
|
||
|
/**
|
||
|
* Create an instance of the class with the input data
|
||
|
*
|
||
|
* @param string $data Input data
|
||
|
*/
|
||
|
public function __construct($data)
|
||
|
{
|
||
|
$this->data = $data;
|
||
|
$this->data_length = strlen($this->data);
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse the input data
|
||
|
*
|
||
|
* @return bool true on success, false on failure
|
||
|
*/
|
||
|
public function parse()
|
||
|
{
|
||
|
while ($this->state && $this->state !== 'emit' && $this->has_data())
|
||
|
{
|
||
|
$state = $this->state;
|
||
|
$this->$state();
|
||
|
}
|
||
|
$this->data = '';
|
||
|
if ($this->state === 'emit' || $this->state === 'body')
|
||
|
{
|
||
|
return true;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->http_version = '';
|
||
|
$this->status_code = '';
|
||
|
$this->reason = '';
|
||
|
$this->headers = array();
|
||
|
$this->body = '';
|
||
|
return false;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Check whether there is data beyond the pointer
|
||
|
*
|
||
|
* @return bool true if there is further data, false if not
|
||
|
*/
|
||
|
protected function has_data()
|
||
|
{
|
||
|
return (bool) ($this->position < $this->data_length);
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* See if the next character is LWS
|
||
|
*
|
||
|
* @return bool true if the next character is LWS, false if not
|
||
|
*/
|
||
|
protected function is_linear_whitespace()
|
||
|
{
|
||
|
return (bool) ($this->data[$this->position] === "\x09"
|
||
|
|| $this->data[$this->position] === "\x20"
|
||
|
|| ($this->data[$this->position] === "\x0A"
|
||
|
&& isset($this->data[$this->position + 1])
|
||
|
&& ($this->data[$this->position + 1] === "\x09" || $this->data[$this->position + 1] === "\x20")));
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse the HTTP version
|
||
|
*/
|
||
|
protected function http_version()
|
||
|
{
|
||
|
if (strpos($this->data, "\x0A") !== false && strtoupper(substr($this->data, 0, 5)) === 'HTTP/')
|
||
|
{
|
||
|
$len = strspn($this->data, '0123456789.', 5);
|
||
|
$this->http_version = substr($this->data, 5, $len);
|
||
|
$this->position += 5 + $len;
|
||
|
if (substr_count($this->http_version, '.') <= 1)
|
||
|
{
|
||
|
$this->http_version = (float) $this->http_version;
|
||
|
$this->position += strspn($this->data, "\x09\x20", $this->position);
|
||
|
$this->state = 'status';
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = false;
|
||
|
}
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = false;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse the status code
|
||
|
*/
|
||
|
protected function status()
|
||
|
{
|
||
|
if ($len = strspn($this->data, '0123456789', $this->position))
|
||
|
{
|
||
|
$this->status_code = (int) substr($this->data, $this->position, $len);
|
||
|
$this->position += $len;
|
||
|
$this->state = 'reason';
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = false;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse the reason phrase
|
||
|
*/
|
||
|
protected function reason()
|
||
|
{
|
||
|
$len = strcspn($this->data, "\x0A", $this->position);
|
||
|
$this->reason = trim(substr($this->data, $this->position, $len), "\x09\x0D\x20");
|
||
|
$this->position += $len + 1;
|
||
|
$this->state = 'new_line';
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Deal with a new line, shifting data around as needed
|
||
|
*/
|
||
|
protected function new_line()
|
||
|
{
|
||
|
$this->value = trim($this->value, "\x0D\x20");
|
||
|
if ($this->name !== '' && $this->value !== '')
|
||
|
{
|
||
|
$this->name = strtolower($this->name);
|
||
|
// We should only use the last Content-Type header. c.f. issue #1
|
||
|
if (isset($this->headers[$this->name]) && $this->name !== 'content-type')
|
||
|
{
|
||
|
$this->headers[$this->name] .= ', ' . $this->value;
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->headers[$this->name] = $this->value;
|
||
|
}
|
||
|
}
|
||
|
$this->name = '';
|
||
|
$this->value = '';
|
||
|
if (substr($this->data[$this->position], 0, 2) === "\x0D\x0A")
|
||
|
{
|
||
|
$this->position += 2;
|
||
|
$this->state = 'body';
|
||
|
}
|
||
|
elseif ($this->data[$this->position] === "\x0A")
|
||
|
{
|
||
|
$this->position++;
|
||
|
$this->state = 'body';
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = 'name';
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse a header name
|
||
|
*/
|
||
|
protected function name()
|
||
|
{
|
||
|
$len = strcspn($this->data, "\x0A:", $this->position);
|
||
|
if (isset($this->data[$this->position + $len]))
|
||
|
{
|
||
|
if ($this->data[$this->position + $len] === "\x0A")
|
||
|
{
|
||
|
$this->position += $len;
|
||
|
$this->state = 'new_line';
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->name = substr($this->data, $this->position, $len);
|
||
|
$this->position += $len + 1;
|
||
|
$this->state = 'value';
|
||
|
}
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = false;
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse LWS, replacing consecutive LWS characters with a single space
|
||
|
*/
|
||
|
protected function linear_whitespace()
|
||
|
{
|
||
|
do
|
||
|
{
|
||
|
if (substr($this->data, $this->position, 2) === "\x0D\x0A")
|
||
|
{
|
||
|
$this->position += 2;
|
||
|
}
|
||
|
elseif ($this->data[$this->position] === "\x0A")
|
||
|
{
|
||
|
$this->position++;
|
||
|
}
|
||
|
$this->position += strspn($this->data, "\x09\x20", $this->position);
|
||
|
} while ($this->has_data() && $this->is_linear_whitespace());
|
||
|
$this->value .= "\x20";
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* See what state to move to while within non-quoted header values
|
||
|
*/
|
||
|
protected function value()
|
||
|
{
|
||
|
if ($this->is_linear_whitespace())
|
||
|
{
|
||
|
$this->linear_whitespace();
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
switch ($this->data[$this->position])
|
||
|
{
|
||
|
case '"':
|
||
|
// Workaround for ETags: we have to include the quotes as
|
||
|
// part of the tag.
|
||
|
if (strtolower($this->name) === 'etag')
|
||
|
{
|
||
|
$this->value .= '"';
|
||
|
$this->position++;
|
||
|
$this->state = 'value_char';
|
||
|
break;
|
||
|
}
|
||
|
$this->position++;
|
||
|
$this->state = 'quote';
|
||
|
break;
|
||
|
|
||
|
case "\x0A":
|
||
|
$this->position++;
|
||
|
$this->state = 'new_line';
|
||
|
break;
|
||
|
|
||
|
default:
|
||
|
$this->state = 'value_char';
|
||
|
break;
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse a header value while outside quotes
|
||
|
*/
|
||
|
protected function value_char()
|
||
|
{
|
||
|
$len = strcspn($this->data, "\x09\x20\x0A\"", $this->position);
|
||
|
$this->value .= substr($this->data, $this->position, $len);
|
||
|
$this->position += $len;
|
||
|
$this->state = 'value';
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* See what state to move to while within quoted header values
|
||
|
*/
|
||
|
protected function quote()
|
||
|
{
|
||
|
if ($this->is_linear_whitespace())
|
||
|
{
|
||
|
$this->linear_whitespace();
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
switch ($this->data[$this->position])
|
||
|
{
|
||
|
case '"':
|
||
|
$this->position++;
|
||
|
$this->state = 'value';
|
||
|
break;
|
||
|
|
||
|
case "\x0A":
|
||
|
$this->position++;
|
||
|
$this->state = 'new_line';
|
||
|
break;
|
||
|
|
||
|
case '\\':
|
||
|
$this->position++;
|
||
|
$this->state = 'quote_escaped';
|
||
|
break;
|
||
|
|
||
|
default:
|
||
|
$this->state = 'quote_char';
|
||
|
break;
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse a header value while within quotes
|
||
|
*/
|
||
|
protected function quote_char()
|
||
|
{
|
||
|
$len = strcspn($this->data, "\x09\x20\x0A\"\\", $this->position);
|
||
|
$this->value .= substr($this->data, $this->position, $len);
|
||
|
$this->position += $len;
|
||
|
$this->state = 'value';
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse an escaped character within quotes
|
||
|
*/
|
||
|
protected function quote_escaped()
|
||
|
{
|
||
|
$this->value .= $this->data[$this->position];
|
||
|
$this->position++;
|
||
|
$this->state = 'quote';
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parse the body
|
||
|
*/
|
||
|
protected function body()
|
||
|
{
|
||
|
$this->body = substr($this->data, $this->position);
|
||
|
if (!empty($this->headers['transfer-encoding']))
|
||
|
{
|
||
|
unset($this->headers['transfer-encoding']);
|
||
|
$this->state = 'chunked';
|
||
|
}
|
||
|
else
|
||
|
{
|
||
|
$this->state = 'emit';
|
||
|
}
|
||
|
}
|
||
|
|
||
|
/**
|
||
|
* Parsed a "Transfer-Encoding: chunked" body
|
||
|
*/
|
||
|
protected function chunked()
|
||
|
{
|
||
|
if (!preg_match('/^([0-9a-f]+)[^\r\n]*\r\n/i', trim($this->body)))
|
||
|
{
|
||
|
$this->state = 'emit';
|
||
|
return;
|
||
|
}
|
||
|
|
||
|
$decoded = '';
|
||
|
$encoded = $this->body;
|
||
|
|
||
|
while (true)
|
||
|
{
|
||
|
$is_chunked = (bool) preg_match( '/^([0-9a-f]+)[^\r\n]*\r\n/i', $encoded, $matches );
|
||
|
if (!$is_chunked)
|
||
|
{
|
||
|
// Looks like it's not chunked after all
|
||
|
$this->state = 'emit';
|
||
|
return;
|
||
|
}
|
||
|
|
||
|
$length = hexdec(trim($matches[1]));
|
||
|
if ($length === 0)
|
||
|
{
|
||
|
// Ignore trailer headers
|
||
|
$this->state = 'emit';
|
||
|
$this->body = $decoded;
|
||
|
return;
|
||
|
}
|
||
|
|
||
|
$chunk_length = strlen($matches[0]);
|
||
|
$decoded .= $part = substr($encoded, $chunk_length, $length);
|
||
|
$encoded = substr($encoded, $chunk_length + $length + 2);
|
||
|
|
||
|
if (trim($encoded) === '0' || empty($encoded))
|
||
|
{
|
||
|
$this->state = 'emit';
|
||
|
$this->body = $decoded;
|
||
|
return;
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
}
|