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 = self.get(bookmark, url, True)
  72
  73             if error:
  74                 bookmark.error = error
  75                 return 1
  76
  77             if redirect_code:
  78                 self.set_redirect(bookmark, redirect_code, redirect_to)
  79                 return 1
  80
  81             size = 0
  82             last_modified = None
  83
  84             if headers:
  85                 try:
  86                     size = headers["Content-Length"]
  87                 except KeyError:
  88                     size = len(content)
  89
  90                 try:
  91                     last_modified = headers["Last-Modified"]
  92                 except KeyError:
  93                     pass
  94
  95                 if last_modified:
  96                     last_modified = parse_time(last_modified)
  97             else:
  98                 size = len(content)
  99
 100             if last_modified:
 101                 last_modified = str(int(last_modified))
 102             else:
 103                 last_modified = bookmark.last_visit
 104
 105             bookmark.size = size
 106             bookmark.last_modified = last_modified
 107
 108             md5 = md5wrapper()
 109             if url_type == "ftp":  # Pass welcome message through MD5
 110                 md5.update(self.get_ftp_welcome())
 111
 112             md5.update(content)
 113             bookmark.md5 = str(md5)
 114
 115             if headers:
 116                 try:
 117                     content_type = headers["Content-Type"]
 118                     self.log("   Content-Type: %s" % content_type)
 119                     try:
 120                         # extract charset from "text/html; foo; charset=UTF-8, bar; baz;"
 121                         content_type, charset = content_type.split(';', 1)
 122                         content_type = content_type.strip()
 123                         charset = charset.split('=')[1].strip().split(',')[0]
 124                         self.log("   HTTP charset   : %s" % charset)
 125                     except (ValueError, IndexError):
 126                         charset = None
 127                         self.log("   no charset in Content-Type header")
 128                     for ctype in ("text/html", "application/xhtml+xml"):
 129                         if content_type.startswith(ctype):
 130                             html = True
 131                             break
 132                     else:
 133                         html = False
 134                     if html:
 135                         parser = parse_html(content, charset, self.log)
 136                         if parser:
 137                             bookmark.real_title = parser.title
 138                             icon = parser.icon
 139                         else:
 140                             icon = None
 141                         if not icon:
 142                             icon = "/favicon.ico"
 143                         icon_url = urljoin("%s://%s%s" % (url_type, url_host, url_path), icon)
 144                         self.log("   looking for icon at: %s" % icon_url)
 145                         if icon_url in icons:
 146                             if icons[icon_url]:
 147                                 bookmark.icon_href = icon_url
 148                                 content_type, bookmark.icon = icons[icon_url]
 149                                 self.log("   cached icon: %s" % content_type)
 150                             else:
 151                                 self.log("   cached icon: no icon")
 152                         else:
 153                             try:
 154                                 _icon_url = icon_url
 155                                 for i in range(8):
 156                                     error, icon_redirect_code, icon_redirect_to, \
 157                                         icon_headers, icon_data = \
 158                                         self.get(bookmark, _icon_url)
 159                                     if icon_redirect_code:
 160                                         _icon_url = icon_redirect_to
 161                                         self.log("   redirect to : %s" % _icon_url)
 162                                     else:
 163                                         if icon_data is None:
 164                                             raise IOError("No icon")
 165                                         break
 166                                 else:
 167                                     raise IOError("Too many redirects")
 168                             except:
 169                                 etype, emsg, tb = sys.exc_info()
 170                                 self.log("   no icon        : %s %s" % (etype, emsg))
 171                                 etype = emsg = tb = None
 172                                 icons[icon_url] = None
 173                             else:
 174                                 content_type = icon_headers["Content-Type"]
 175                                 if content_type.startswith("application/") \
 176                                    or content_type.startswith("image/") \
 177                                    or content_type.startswith("text/plain"):
 178                                     bookmark.icon_href = icon_url
 179                                     self.log("   got icon       : %s" % content_type)
 180                                     if content_type.startswith("application/") \
 181                                        or content_type.startswith("text/plain"):
 182                                         self.log("   non-image content type, assume x-icon")
 183                                         content_type = 'image/x-icon'
 184                                     bookmark.icon = "data:%s;base64,%s" % (content_type, b64encode(icon_data))
 185                                     icons[icon_url] = (content_type, bookmark.icon)
 186                                 else:
 187                                     self.log("   no icon        : bad content type '%s'" % content_type)
 188                                     icons[icon_url] = None
 189                         if parser and parser.refresh:
 190                             refresh = parser.refresh
 191                             try:
 192                                 url = refresh.split('=', 1)[1]
 193                             except IndexError:
 194                                 url = "self"
 195                             try:
 196                                 timeout = float(refresh.split(';')[0])
 197                             except (IndexError, ValueError):
 198                                 self.set_redirect(bookmark, "html", "Bad redirect to %s (%s)" % (url, refresh))
 199                             else:
 200                                 try:
 201                                     timeout = int(refresh.split(';')[0])
 202                                 except ValueError:
 203                                     pass  # float timeout
 204                                 self.set_redirect(bookmark, "html", "%s (%s sec)" % (url, timeout))
 205
 206                 except KeyError as key:
 207                     self.log("   no header: %s" % key)
 208
 209         except EOFError:
 210             bookmark.error = "Unexpected EOF (FTP server closed connection)"
 211             self.log('   EOF: %s' % bookmark.error)
 212
 213         except KeyboardInterrupt:
 214             self.log("Keyboard interrupt (^C)")
 215             return 0
 216
 217         except socket.error as e:
 218             bookmark.error = get_error(e)
 219             self.log(bookmark.error)
 220
 221         except:
 222             import traceback
 223             traceback.print_exc()
 224             bookmark.error = "Exception!"
 225             self.log('   Exception: %s' % bookmark.error)
 226
 227         finally:
 228             self.finish_check_url(bookmark)
 229
 230         # Tested
 231         return 1
 232
 233     def set_redirect(self, bookmark, errcode, newurl):
 234         bookmark.moved = "(%s) to %s" % (reloc_dict[errcode], newurl)
 235         self.log('   Moved: %s' % bookmark.moved)
 236
 237     def finish_check_url(self, bookmark):
 238         start = self.start
 239         bookmark.last_tested = str(start)
 240         now = int(time.time())
 241         bookmark.test_time = str(now - start)