yt_dlp/extractor/openload.py

   1 import json
   2 import os
   3 import subprocess
   4 import tempfile
   5
   6 from ..compat import compat_urlparse
   7 from ..utils import (
   8     ExtractorError,
   9     Popen,
  10     check_executable,
  11     encodeArgument,
  12     get_exe_version,
  13     is_outdated_version,
  14 )
  15
  16
  17 def cookie_to_dict(cookie):
  18     cookie_dict = {
  19         'name': cookie.name,
  20         'value': cookie.value,
  21     }
  22     if cookie.port_specified:
  23         cookie_dict['port'] = cookie.port
  24     if cookie.domain_specified:
  25         cookie_dict['domain'] = cookie.domain
  26     if cookie.path_specified:
  27         cookie_dict['path'] = cookie.path
  28     if cookie.expires is not None:
  29         cookie_dict['expires'] = cookie.expires
  30     if cookie.secure is not None:
  31         cookie_dict['secure'] = cookie.secure
  32     if cookie.discard is not None:
  33         cookie_dict['discard'] = cookie.discard
  34     try:
  35         if (cookie.has_nonstandard_attr('httpOnly')
  36                 or cookie.has_nonstandard_attr('httponly')
  37                 or cookie.has_nonstandard_attr('HttpOnly')):
  38             cookie_dict['httponly'] = True
  39     except TypeError:
  40         pass
  41     return cookie_dict
  42
  43
  44 def cookie_jar_to_list(cookie_jar):
  45     return [cookie_to_dict(cookie) for cookie in cookie_jar]
  46
  47
  48 class PhantomJSwrapper:
  49     """PhantomJS wrapper class
  50
  51     This class is experimental.
  52     """
  53
  54     _TEMPLATE = r'''
  55         phantom.onError = function(msg, trace) {{
  56           var msgStack = ['PHANTOM ERROR: ' + msg];
  57           if(trace && trace.length) {{
  58             msgStack.push('TRACE:');
  59             trace.forEach(function(t) {{
  60               msgStack.push(' -> ' + (t.file || t.sourceURL) + ': ' + t.line
  61                 + (t.function ? ' (in function ' + t.function +')' : ''));
  62             }});
  63           }}
  64           console.error(msgStack.join('\n'));
  65           phantom.exit(1);
  66         }};
  67         var page = require('webpage').create();
  68         var fs = require('fs');
  69         var read = {{ mode: 'r', charset: 'utf-8' }};
  70         var write = {{ mode: 'w', charset: 'utf-8' }};
  71         JSON.parse(fs.read("{cookies}", read)).forEach(function(x) {{
  72           phantom.addCookie(x);
  73         }});
  74         page.settings.resourceTimeout = {timeout};
  75         page.settings.userAgent = "{ua}";
  76         page.onLoadStarted = function() {{
  77           page.evaluate(function() {{
  78             delete window._phantom;
  79             delete window.callPhantom;
  80           }});
  81         }};
  82         var saveAndExit = function() {{
  83           fs.write("{html}", page.content, write);
  84           fs.write("{cookies}", JSON.stringify(phantom.cookies), write);
  85           phantom.exit();
  86         }};
  87         page.onLoadFinished = function(status) {{
  88           if(page.url === "") {{
  89             page.setContent(fs.read("{html}", read), "{url}");
  90           }}
  91           else {{
  92             {jscode}
  93           }}
  94         }};
  95         page.open("");
  96     '''
  97
  98     _TMP_FILE_NAMES = ['script', 'html', 'cookies']
  99
 100     @staticmethod
 101     def _version():
 102         return get_exe_version('phantomjs', version_re=r'([0-9.]+)')
 103
 104     def __init__(self, extractor, required_version=None, timeout=10000):
 105         self._TMP_FILES = {}
 106
 107         self.exe = check_executable('phantomjs', ['-v'])
 108         if not self.exe:
 109             raise ExtractorError('PhantomJS executable not found in PATH, '
 110                                  'download it from http://phantomjs.org',
 111                                  expected=True)
 112
 113         self.extractor = extractor
 114
 115         if required_version:
 116             version = self._version()
 117             if is_outdated_version(version, required_version):
 118                 self.extractor._downloader.report_warning(
 119                     'Your copy of PhantomJS is outdated, update it to version '
 120                     '%s or newer if you encounter any errors.' % required_version)
 121
 122         self.options = {
 123             'timeout': timeout,
 124         }
 125         for name in self._TMP_FILE_NAMES:
 126             tmp = tempfile.NamedTemporaryFile(delete=False)
 127             tmp.close()
 128             self._TMP_FILES[name] = tmp
 129
 130     def __del__(self):
 131         for name in self._TMP_FILE_NAMES:
 132             try:
 133                 os.remove(self._TMP_FILES[name].name)
 134             except (OSError, KeyError):
 135                 pass
 136
 137     def _save_cookies(self, url):
 138         cookies = cookie_jar_to_list(self.extractor._downloader.cookiejar)
 139         for cookie in cookies:
 140             if 'path' not in cookie:
 141                 cookie['path'] = '/'
 142             if 'domain' not in cookie:
 143                 cookie['domain'] = compat_urlparse.urlparse(url).netloc
 144         with open(self._TMP_FILES['cookies'].name, 'wb') as f:
 145             f.write(json.dumps(cookies).encode('utf-8'))
 146
 147     def _load_cookies(self):
 148         with open(self._TMP_FILES['cookies'].name, 'rb') as f:
 149             cookies = json.loads(f.read().decode('utf-8'))
 150         for cookie in cookies:
 151             if cookie['httponly'] is True:
 152                 cookie['rest'] = {'httpOnly': None}
 153             if 'expiry' in cookie:
 154                 cookie['expire_time'] = cookie['expiry']
 155             self.extractor._set_cookie(**cookie)
 156
 157     def get(self, url, html=None, video_id=None, note=None, note2='Executing JS on webpage', headers={}, jscode='saveAndExit();'):
 158         """
 159         Downloads webpage (if needed) and executes JS
 160
 161         Params:
 162             url: website url
 163             html: optional, html code of website
 164             video_id: video id
 165             note: optional, displayed when downloading webpage
 166             note2: optional, displayed when executing JS
 167             headers: custom http headers
 168             jscode: code to be executed when page is loaded
 169
 170         Returns tuple with:
 171             * downloaded website (after JS execution)
 172             * anything you print with `console.log` (but not inside `page.execute`!)
 173
 174         In most cases you don't need to add any `jscode`.
 175         It is executed in `page.onLoadFinished`.
 176         `saveAndExit();` is mandatory, use it instead of `phantom.exit()`
 177         It is possible to wait for some element on the webpage, for example:
 178             var check = function() {
 179               var elementFound = page.evaluate(function() {
 180                 return document.querySelector('#b.done') !== null;
 181               });
 182               if(elementFound)
 183                 saveAndExit();
 184               else
 185                 window.setTimeout(check, 500);
 186             }
 187
 188             page.evaluate(function(){
 189               document.querySelector('#a').click();
 190             });
 191             check();
 192         """
 193         if 'saveAndExit();' not in jscode:
 194             raise ExtractorError('`saveAndExit();` not found in `jscode`')
 195         if not html:
 196             html = self.extractor._download_webpage(url, video_id, note=note, headers=headers)
 197         with open(self._TMP_FILES['html'].name, 'wb') as f:
 198             f.write(html.encode('utf-8'))
 199
 200         self._save_cookies(url)
 201
 202         replaces = self.options
 203         replaces['url'] = url
 204         user_agent = headers.get('User-Agent') or self.extractor.get_param('http_headers')['User-Agent']
 205         replaces['ua'] = user_agent.replace('"', '\\"')
 206         replaces['jscode'] = jscode
 207
 208         for x in self._TMP_FILE_NAMES:
 209             replaces[x] = self._TMP_FILES[x].name.replace('\\', '\\\\').replace('"', '\\"')
 210
 211         with open(self._TMP_FILES['script'].name, 'wb') as f:
 212             f.write(self._TEMPLATE.format(**replaces).encode('utf-8'))
 213
 214         if video_id is None:
 215             self.extractor.to_screen(f'{note2}')
 216         else:
 217             self.extractor.to_screen(f'{video_id}: {note2}')
 218
 219         p = Popen(
 220             [self.exe, '--ssl-protocol=any', self._TMP_FILES['script'].name],
 221             stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 222         out, err = p.communicate_or_kill()
 223         if p.returncode != 0:
 224             raise ExtractorError(
 225                 'Executing JS failed\n:' + encodeArgument(err))
 226         with open(self._TMP_FILES['html'].name, 'rb') as f:
 227             html = f.read().decode('utf-8')
 228
 229         self._load_cookies()
 230
 231         return (html, encodeArgument(out))