yt_dlp/extractor/openload.py

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