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 urllib.parse import splittype, splithost, splittag, urljoin
21 from urllib import splittype, splithost, splittag
22 from urlparse import urljoin
24 from m_lib.md5wrapper import md5wrapper
25 from m_lib.net.www.util import parse_time
27 from bkmk_objects import Robot
28 from parse_html import parse_html
41 if isinstance(e, str):
47 s.append("'%s'" % str(i).replace('\n', "\\n"))
48 return "(%s)" % ' '.join(s)
51 # Icon cache; maps URL to a tuple (content type, data)
52 # or None if there is no icon.
56 class robot_base(Robot):
59 def __init__(self, *args, **kw):
60 Robot.__init__(self, *args, **kw)
61 socket.setdefaulttimeout(int(self.timeout))
63 def check_url(self, bookmark):
65 self.start = int(time.time())
68 url_type, url_rest = splittype(bookmark.href)
69 url_host, url_path = splithost(url_rest)
70 url_path, url_tag = splittag(url_path) # noqa: E221
71 # multiple spaces before operator
73 url = "%s://%s%s" % (url_type, url_host, url_path)
74 error, redirect_code, redirect_to, headers, content = \
75 self.get(bookmark, url, True)
78 bookmark.error = error
82 self.set_redirect(bookmark, redirect_code, redirect_to)
90 size = headers["Content-Length"]
95 last_modified = headers["Last-Modified"]
100 last_modified = parse_time(last_modified)
105 last_modified = str(int(last_modified))
107 last_modified = bookmark.last_visit
110 bookmark.last_modified = last_modified
113 if url_type == "ftp": # Pass welcome message through MD5
114 ftp_welcome = self.get_ftp_welcome()
115 if not isinstance(ftp_welcome, bytes):
116 ftp_welcome = ftp_welcome.encode('utf-8')
117 md5.update(ftp_welcome)
119 if isinstance(content, bytes):
122 md5.update(content.encode('utf-8'))
123 bookmark.md5 = str(md5)
127 content_type = headers["Content-Type"]
128 self.log(" Content-Type: %s" % content_type)
129 if content_type is None:
130 if 'html' in content.lower():
131 content_type = 'text/html'
133 content_type = 'text/plain'
134 self.log(" Set Content-Type to: %s"
137 # extract charset from
138 # "text/html; foo; charset=UTF-8, bar; baz;"
139 content_type, charset = content_type.split(';', 1)
140 content_type = content_type.strip()
141 charset = charset.split('=')[1].strip().split(',')[0]
142 self.log(" HTTP charset : %s" % charset)
143 except (ValueError, IndexError):
145 self.log(" no charset in Content-Type header")
146 for ctype in ("text/html", "application/xhtml+xml"):
147 if content_type.startswith(ctype):
153 parser = parse_html(content, charset, self.log)
155 bookmark.real_title = parser.title
160 icon = "/favicon.ico"
162 "%s://%s%s" % (url_type, url_host, url_path), icon)
163 self.log(" looking for icon at: %s" % icon_url)
164 if icon_url in icons:
166 bookmark.icon_href = icon_url
167 content_type, bookmark.icon = icons[icon_url]
168 self.log(" cached icon: %s" % content_type)
170 self.log(" cached icon: no icon")
175 error, icon_redirect_code, \
176 icon_redirect_to, icon_headers, \
178 self.get(bookmark, _icon_url)
179 if icon_redirect_code:
180 _icon_url = icon_redirect_to
181 self.log(" redirect to : %s"
184 if icon_data is None:
185 raise IOError("No icon")
188 raise IOError("Too many redirects")
190 etype, emsg, _ = sys.exc_info()
191 self.log(" no icon : %s %s"
193 etype = emsg = _ = None
194 icons[icon_url] = None
196 content_type = icon_headers["Content-Type"]
197 if content_type and (
198 content_type.startswith("application/")
199 or content_type.startswith("image/")
200 or content_type.startswith("text/plain")
202 bookmark.icon_href = icon_url
203 self.log(" got icon : %s"
206 content_type.startswith("application/")
207 or content_type.startswith(
210 self.log(" non-image content type,"
212 content_type = 'image/x-icon'
213 if not isinstance(icon_data, bytes):
214 icon_data = icon_data.encode('utf-8')
215 bookmark.icon = "data:%s;base64,%s" \
216 % (content_type, b64encode(icon_data))
217 icons[icon_url] = (content_type,
221 self.log(" no icon :"
222 "bad content type '%s'"
225 icons[icon_url] = None
226 if parser and parser.refresh:
227 refresh = parser.refresh
229 url = refresh.split('=', 1)[1]
233 timeout = float(refresh.split(';')[0])
234 except (IndexError, ValueError):
235 self.set_redirect(bookmark, "html",
236 "Bad redirect to %s (%s)"
241 timeout = int(refresh.split(';')[0])
244 self.set_redirect(bookmark, "html",
249 except KeyError as key:
250 self.log(" no header: %s" % key)
253 bookmark.error = "Unexpected EOF (FTP server closed connection)"
254 self.log(' EOF: %s' % bookmark.error)
256 except KeyboardInterrupt:
257 self.log("Keyboard interrupt (^C)")
260 except socket.error as e:
261 bookmark.error = get_error(e)
262 self.log(bookmark.error)
266 traceback.print_exc()
267 bookmark.error = "Exception!"
268 self.log(' Exception: %s' % bookmark.error)
271 self.finish_check_url(bookmark)
276 def set_redirect(self, bookmark, errcode, newurl):
277 bookmark.moved = moved = "(%s) to %s" % (reloc_dict[errcode], newurl)
279 moved.encode('ascii')
280 except UnicodeEncodeError:
282 moved = moved.encode(bookmark.charset)
283 except (LookupError, TypeError, UnicodeEncodeError):
284 moved = moved.encode('utf-8')
285 self.log(' Moved: %s' % moved)
287 def finish_check_url(self, bookmark):
289 bookmark.last_tested = str(start)
290 now = int(time.time())
291 bookmark.test_time = str(now - start)