1 """Base class for robots
3 This file is a part of Bookmarks database and Internet robot.
7 __author__ = "Oleg Broytman <phd@phdru.name>"
8 __copyright__ = "Copyright (C) 2000-2023 PhiloSoft Design"
9 __license__ = "GNU GPL"
11 __all__ = ['robot_base', 'get_error']
14 from base64 import b64encode
19 from urlparse import urljoin
21 from m_lib.md5wrapper import md5wrapper
22 from m_lib.net.www.util import parse_time
24 from bkmk_objects import Robot
25 from parse_html import parse_html
38 if isinstance(e, str):
44 s.append("'%s'" % str(i).replace('\n', "\\n"))
45 return "(%s)" % ' '.join(s)
48 # Icon cache; maps URL to a tuple (content type, data)
49 # or None if there is no icon.
53 class robot_base(Robot):
56 def __init__(self, *args, **kw):
57 Robot.__init__(self, *args, **kw)
58 socket.setdefaulttimeout(int(self.timeout))
60 def check_url(self, bookmark):
62 self.start = int(time.time())
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
70 url = "%s://%s%s" % (url_type, url_host, url_path)
71 error, redirect_code, redirect_to, headers, content = self.get(bookmark, url, True)
74 bookmark.error = error
78 self.set_redirect(bookmark, redirect_code, redirect_to)
86 size = headers["Content-Length"]
91 last_modified = headers["Last-Modified"]
96 last_modified = parse_time(last_modified)
101 last_modified = str(int(last_modified))
103 last_modified = bookmark.last_visit
106 bookmark.last_modified = last_modified
109 if url_type == "ftp": # Pass welcome message through MD5
110 md5.update(self.get_ftp_welcome())
113 bookmark.md5 = str(md5)
117 content_type = headers["Content-Type"]
118 self.log(" Content-Type: %s" % content_type)
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):
127 self.log(" no charset in Content-Type header")
128 for ctype in ("text/html", "application/xhtml+xml"):
129 if content_type.startswith(ctype):
135 parser = parse_html(content, charset, self.log)
137 bookmark.real_title = parser.title
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:
147 bookmark.icon_href = icon_url
148 content_type, bookmark.icon = icons[icon_url]
149 self.log(" cached icon: %s" % content_type)
151 self.log(" cached icon: no icon")
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)
163 if icon_data is None:
164 raise IOError("No icon")
167 raise IOError("Too many redirects")
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
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)
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
192 url = refresh.split('=', 1)[1]
196 timeout = float(refresh.split(';')[0])
197 except (IndexError, ValueError):
198 self.set_redirect(bookmark, "html", "Bad redirect to %s (%s)" % (url, refresh))
201 timeout = int(refresh.split(';')[0])
204 self.set_redirect(bookmark, "html", "%s (%s sec)" % (url, timeout))
206 except KeyError as key:
207 self.log(" no header: %s" % key)
210 bookmark.error = "Unexpected EOF (FTP server closed connection)"
211 self.log(' EOF: %s' % bookmark.error)
213 except KeyboardInterrupt:
214 self.log("Keyboard interrupt (^C)")
217 except socket.error as e:
218 bookmark.error = get_error(e)
219 self.log(bookmark.error)
223 traceback.print_exc()
224 bookmark.error = "Exception!"
225 self.log(' Exception: %s' % bookmark.error)
228 self.finish_check_url(bookmark)
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)
237 def finish_check_url(self, bookmark):
239 bookmark.last_tested = str(start)
240 now = int(time.time())
241 bookmark.test_time = str(now - start)