Robots/bkmk_robot_base.py

   1 """Base class for robots
   2
   3 This file is a part of Bookmarks database and Internet robot.
   4
   5 """
   6
   7 __author__ = "Oleg Broytman <phd@phdru.name>"
   8 __copyright__ = "Copyright (C) 2000-2023 PhiloSoft Design"
   9 __license__ = "GNU GPL"
  10
  11 __all__ = ['robot_base', 'get_error']
  12
  13
  14 from base64 import b64encode
  15 import sys
  16 import socket
  17 import time
  18 import urllib
  19 from urlparse import urljoin
  20
  21 from m_lib.md5wrapper import md5wrapper
  22 from m_lib.net.www.util import parse_time
  23
  24 from bkmk_objects import Robot
  25 from parse_html import parse_html
  26
  27
  28 reloc_dict = {
  29   301: "perm.",
  30   302: "temp2.",
  31   303: "temp3.",
  32   307: "temp7.",
  33   "html": "html"
  34 }
  35
  36
  37 def get_error(e):
  38     if isinstance(e, str):
  39         return e
  40
  41     else:
  42         s = []
  43         for i in e:
  44             s.append("'%s'" % str(i).replace('\n', "\\n"))
  45         return "(%s)" % ' '.join(s)
  46
  47
  48 # Icon cache; maps URL to a tuple (content type, data)
  49 # or None if there is no icon.
  50 icons = {}
  51
  52
  53 class robot_base(Robot):
  54     timeout = 60
  55
  56     def __init__(self, *args, **kw):
  57         Robot.__init__(self, *args, **kw)
  58         socket.setdefaulttimeout(int(self.timeout))
  59
  60     def check_url(self, bookmark):
  61         try:
  62             self.start = int(time.time())
  63             bookmark.icon = None
  64
  65             url_type, url_rest = urllib.splittype(bookmark.href)
  66             url_host, url_path = urllib.splithost(url_rest)
  67             url_path, url_tag  = urllib.splittag(url_path)  # noqa: E221
  68             #                            multiple spaces before operator
  69
  70             url = "%s://%s%s" % (url_type, url_host, url_path)
  71             error, redirect_code, redirect_to, headers, content = \
  72                 self.get(bookmark, url, True)
  73
  74             if error:
  75                 bookmark.error = error
  76                 return 1
  77
  78             if redirect_code:
  79                 self.set_redirect(bookmark, redirect_code, redirect_to)
  80                 return 1
  81
  82             size = 0
  83             last_modified = None
  84
  85             if headers:
  86                 try:
  87                     size = headers["Content-Length"]
  88                 except KeyError:
  89                     size = len(content)
  90
  91                 try:
  92                     last_modified = headers["Last-Modified"]
  93                 except KeyError:
  94                     pass
  95
  96                 if last_modified:
  97                     last_modified = parse_time(last_modified)
  98             else:
  99                 size = len(content)
 100
 101             if last_modified:
 102                 last_modified = str(int(last_modified))
 103             else:
 104                 last_modified = bookmark.last_visit
 105
 106             bookmark.size = size
 107             bookmark.last_modified = last_modified
 108
 109             md5 = md5wrapper()
 110             if url_type == "ftp":  # Pass welcome message through MD5
 111                 md5.update(self.get_ftp_welcome())
 112
 113             md5.update(content)
 114             bookmark.md5 = str(md5)
 115
 116             if headers:
 117                 try:
 118                     content_type = headers["Content-Type"]
 119                     self.log("   Content-Type: %s" % content_type)
 120                     try:
 121                         # extract charset from
 122                         # "text/html; foo; charset=UTF-8, bar; baz;"
 123                         content_type, charset = content_type.split(';', 1)
 124                         content_type = content_type.strip()
 125                         charset = charset.split('=')[1].strip().split(',')[0]
 126                         self.log("   HTTP charset   : %s" % charset)
 127                     except (ValueError, IndexError):
 128                         charset = None
 129                         self.log("   no charset in Content-Type header")
 130                     for ctype in ("text/html", "application/xhtml+xml"):
 131                         if content_type.startswith(ctype):
 132                             html = True
 133                             break
 134                     else:
 135                         html = False
 136                     if html:
 137                         parser = parse_html(content, charset, self.log)
 138                         if parser:
 139                             bookmark.real_title = parser.title
 140                             icon = parser.icon
 141                         else:
 142                             icon = None
 143                         if not icon:
 144                             icon = "/favicon.ico"
 145                         icon_url = urljoin(
 146                             "%s://%s%s" % (url_type, url_host, url_path), icon)
 147                         self.log("   looking for icon at: %s" % icon_url)
 148                         if icon_url in icons:
 149                             if icons[icon_url]:
 150                                 bookmark.icon_href = icon_url
 151                                 content_type, bookmark.icon = icons[icon_url]
 152                                 self.log("   cached icon: %s" % content_type)
 153                             else:
 154                                 self.log("   cached icon: no icon")
 155                         else:
 156                             try:
 157                                 _icon_url = icon_url
 158                                 for i in range(8):
 159                                     error, icon_redirect_code, \
 160                                         icon_redirect_to, icon_headers, \
 161                                         icon_data = \
 162                                         self.get(bookmark, _icon_url)
 163                                     if icon_redirect_code:
 164                                         _icon_url = icon_redirect_to
 165                                         self.log("   redirect to : %s"
 166                                                  % _icon_url)
 167                                     else:
 168                                         if icon_data is None:
 169                                             raise IOError("No icon")
 170                                         break
 171                                 else:
 172                                     raise IOError("Too many redirects")
 173                             except:
 174                                 etype, emsg, tb = sys.exc_info()
 175                                 self.log("   no icon        : %s %s"
 176                                          % (etype, emsg))
 177                                 etype = emsg = tb = None
 178                                 icons[icon_url] = None
 179                             else:
 180                                 content_type = icon_headers["Content-Type"]
 181                                 if content_type.startswith("application/") \
 182                                    or content_type.startswith("image/") \
 183                                    or content_type.startswith("text/plain"):
 184                                     bookmark.icon_href = icon_url
 185                                     self.log("   got icon       : %s"
 186                                              % content_type)
 187                                     if (
 188                                         content_type.startswith("application/")
 189                                         or content_type.startswith(
 190                                             "text/plain")
 191                                     ):
 192                                         self.log("   non-image content type,"
 193                                                  " assume x-icon")
 194                                         content_type = 'image/x-icon'
 195                                     bookmark.icon = "data:%s;base64,%s" \
 196                                         % (content_type, b64encode(icon_data))
 197                                     icons[icon_url] = (content_type,
 198                                                        bookmark.icon
 199                                                        )
 200                                 else:
 201                                     self.log("   no icon        :"
 202                                              "bad content type '%s'"
 203                                              % content_type
 204                                              )
 205                                     icons[icon_url] = None
 206                         if parser and parser.refresh:
 207                             refresh = parser.refresh
 208                             try:
 209                                 url = refresh.split('=', 1)[1]
 210                             except IndexError:
 211                                 url = "self"
 212                             try:
 213                                 timeout = float(refresh.split(';')[0])
 214                             except (IndexError, ValueError):
 215                                 self.set_redirect(bookmark, "html",
 216                                                   "Bad redirect to %s (%s)"
 217                                                   % (url, refresh)
 218                                                   )
 219                             else:
 220                                 try:
 221                                     timeout = int(refresh.split(';')[0])
 222                                 except ValueError:
 223                                     pass  # float timeout
 224                                 self.set_redirect(bookmark, "html",
 225                                                   "%s (%s sec)"
 226                                                   % (url, timeout)
 227                                                   )
 228
 229                 except KeyError as key:
 230                     self.log("   no header: %s" % key)
 231
 232         except EOFError:
 233             bookmark.error = "Unexpected EOF (FTP server closed connection)"
 234             self.log('   EOF: %s' % bookmark.error)
 235
 236         except KeyboardInterrupt:
 237             self.log("Keyboard interrupt (^C)")
 238             return 0
 239
 240         except socket.error as e:
 241             bookmark.error = get_error(e)
 242             self.log(bookmark.error)
 243
 244         except:
 245             import traceback
 246             traceback.print_exc()
 247             bookmark.error = "Exception!"
 248             self.log('   Exception: %s' % bookmark.error)
 249
 250         finally:
 251             self.finish_check_url(bookmark)
 252
 253         # Tested
 254         return 1
 255
 256     def set_redirect(self, bookmark, errcode, newurl):
 257         bookmark.moved = "(%s) to %s" % (reloc_dict[errcode], newurl)
 258         self.log('   Moved: %s' % bookmark.moved)
 259
 260     def finish_check_url(self, bookmark):
 261         start = self.start
 262         bookmark.last_tested = str(start)
 263         now = int(time.time())
 264         bookmark.test_time = str(now - start)