Save a webpage (and its linked assets) into a single HTML file.

pagesnap

#! /usr/bin/env php
<?php

/***
 * PageSnap
 *
 * Script for archiving a HTML page (and associated media assets) into a single file.  
 *
 * Requires: 
 *
 * - PhpQuery: http://code.google.com/p/phpquery
 * - Parallel Curl: http://github.com/petewarden/parallelcurl
 */

require 'parallelcurl.php';
require 'phpQuery.php';

class PageSnap {
  
  public $url=null;
  private $doc=null;

  //associative array mapping assets URLs to DOM Element
  private $asset_urls=array(); 
  private $multi_curl=null;

  
  public $selectors=array(
    'link[href]',
     'script[src]',
     'img[src]',
     'object[src]'
     //TODO: iframe frame
  );

  function __construct($url,$filename=null){
       $this->url=$url;
       $this->filename= $filename ? $filename :  preg_replace('@^http://(www.)?@','',$url) . '.html';   
       $this->doc=phpQuery::newDocument(file_get_contents($url));

       $this->multi_curl = new ParallelCurl(10,array(
                   CURLOPT_FOLLOWLOCATION => true,
                   CURLOPT_USERAGENT => "Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1", 
       ));
       return $this;
  }
  
  /**
   * Parallel Curl callback.  
   *
   */

  function set_data_url($response_body,$url,$ch,$element){
    if (!in_array($url,$this->asset_urls)){
       file_put_contents('php://stderr',"fetching {$url}\n");
       $data=array();
       $data[]= 'data:' . preg_replace('/;.*$/','',curl_getinfo($ch,CURLINFO_CONTENT_TYPE));
       $data[]= 'base64,'.base64_encode($response_body);
       foreach(array('href','src') as $attr) if ($element->hasAttribute($attr))  pq($element)->attr($attr,implode(';',$data)); 
       $this->asset_urls[]=$url;
   } 
  }

  /**
   * Resolve relative URLs.
   *
   * @author Stefano Faenza
   * http://www.stefanoforenza.com/how-to-build-an-absolute-url-in-php/
   */

  private function absolute_url($u,$p){
        $url = parse_url( $u );
        $page = parse_url( $p );

        if ( strpos( $u , '/' ) === 0 )
        {
                //already absolute              
        } else {
                $basePath = '';
                if (
                        isset( $page[ 'path' ] )
                        && strpos( ltrim( $page[ 'path' ], '/' ), '/' )
                )
                {
                        $baseTokens = explode( '/', $page[ 'path' ] );
                        array_pop( $baseTokens ); // strip basename                     
                        $baseTokens[] = $u;
                        $u = join( '/', $baseTokens );
                }
        }
        if ( ! isset( $url[ 'host' ]))
        {
                $u = 'http://'.$page[ 'host' ].'/'.ltrim( $u, '/' );
        }
        return $u;
  }


 /**
  * Maps assets URL to the DOM elements that reference them. 
  *
  */
  private function get_urls($url){   
    foreach($this->selectors as $selector){
      $i=0;
      foreach(pq($selector) as $element){       

         //move on to the next iteration if link is not shortcut icon or stylesheet
         if ( pq($element)->attr('rel') && !in_array( strtolower(pq($element)->attr('rel')), array('shortcut icon','stylesheet'))){
            continue;
         }   
         foreach(array('src','href') as $attr){
             if ($element->hasAttribute($attr)) {
                    $absolute_url = $this->absolute_url(pq($element)->attr($attr), $url);
                    $this->asset_urls[$absolute_url]=$element; 
             }
          }
     }      
    }
  }

  /**
   * Updates the DOM with the data URL its linked assets and saves the document in the current working directory.
   */
  function fetch($url=null){
    $url or $url = $this->url;
    $this->get_urls($url);
    foreach($this->asset_urls as $u => $element){
           $this->multi_curl->startRequest($u,array($this,'set_data_url'),$element);
    }
    $this->multi_curl->finishAllRequests();

    //TODO: gzip file 
    file_put_contents( basename($this->filename), $this->doc->html());
  }



}
$p = new PageSnap($argv[1],$argv[2]);
$p->fetch(); 
?>

Although this script is itself smart, a better name would make it even more!

I would love a webservice that ran this, great idea

can you work on a solution for css files that @import other css files? Is there any work around for that?