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) 2000-2024 PhiloSoft Design"
9 __license__ = "GNU GPL"
11 __all__ = ['robot_base', 'get_error']
14 from base64 import b64encode
15 from urllib.parse import urljoin
20 from m_lib.md5wrapper import md5wrapper
21 from m_lib.net.www.util import parse_time
23 from bkmk_objects import Robot
24 from parse_html import parse_html
27 # Fake headers to pretend this is a real browser
28 _user_agent = "Mozilla/5.0 (X11; U; Linux 2.6 i686; en)"
29 " Gecko/20001221 Firefox/2.0.0"
30 _x_user_agent = "bookmarks_db (Python %d.%d.%d)" % sys.version_info[:3]
34 'Accept-Language': 'ru,en',
35 'Cache-Control': 'max-age=300',
36 'Connection': 'close',
38 'User-Agent': _user_agent,
39 'X-User-Agent': _x_user_agent,
54 if isinstance(e, str):
60 s.append("'%s'" % str(i).replace('\n', "\\n"))
61 return "(%s)" % ' '.join(s)
64 # Icon cache; maps URL to a tuple (content type, data)
65 # or None if there is no icon.
69 class robot_base(Robot):
72 def __init__(self, *args, **kw):
73 Robot.__init__(self, *args, **kw)
74 socket.setdefaulttimeout(int(self.timeout))
76 def check_url(self, bookmark):
78 self.start = int(time.time())
81 error, redirect_code, redirect_to, headers, content = \
82 self.get(bookmark, bookmark.href, True)
85 bookmark.error = error
89 self.set_redirect(bookmark, redirect_code, redirect_to)
97 size = headers["Content-Length"]
102 last_modified = headers["Last-Modified"]
107 last_modified = parse_time(last_modified)
109 if not size: # Could be None from headers
113 last_modified = str(int(last_modified))
115 last_modified = bookmark.last_visit
118 bookmark.last_modified = last_modified
123 content_type = headers["Content-Type"]
124 self.log(" Content-Type : %s" % content_type)
125 if content_type is None:
126 if b'html' in content.lower():
127 content_type = 'text/html'
129 content_type = 'text/plain'
130 self.log(" Set Content-Type to: %s"
133 # extract charset from
134 # "text/html; charset=UTF-8, foo; bar"
135 content_type, charset = content_type.split(';', 1)
136 content_type = content_type.strip()
137 charset = charset.split('=')[1].strip().split(',')[0]
138 self.log(" HTTP charset : %s" % charset)
139 except (ValueError, IndexError):
141 self.log(" no charset in Content-Type header")
143 for ctype in ("text/html", "application/xhtml+xml"):
144 if content_type.startswith(ctype):
147 content_stripped = content.strip()
148 if content_stripped and charset:
150 content_stripped = content_stripped.decode(
154 self.log(" unknown charset "
155 "in Content-Type header")
156 if content_stripped and is_html:
158 content_stripped, charset, self.log)
160 bookmark.charset = charset
161 elif parser and parser.meta_charset:
162 bookmark.charset = parser.meta_charset
164 bookmark.real_title = parser.title
169 icon = "/favicon.ico"
170 icon_url = urljoin(bookmark.href, icon)
171 self.log(" looking for icon at: %s" % icon_url)
172 if icon_url in icons:
174 bookmark.icon_href = icon_url
175 content_type, bookmark.icon = icons[icon_url]
176 self.log(" cached icon : %s"
179 self.log(" cached icon : no icon")
180 elif icon_url.startswith('data:'):
181 content_type, icon_data = \
182 icon_url[len('data:'):].split(',', 1)
183 bookmark.icon_href = bookmark.icon = icon_url
184 self.log(" got data icon : %s" % content_type)
185 icons[icon_url] = (content_type, icon_url)
190 error, icon_redirect_code, \
191 icon_redirect_to, icon_headers, \
193 self.get(bookmark, _icon_url)
194 if icon_redirect_code:
195 _icon_url = icon_redirect_to
196 self.log(" redirect to : %s"
199 if icon_data is None:
200 raise IOError("No icon")
203 raise IOError("Too many redirects")
205 etype, emsg, _ = sys.exc_info()
206 self.log(" no icon : %s %s"
208 etype = emsg = _ = None
209 icons[icon_url] = None
211 content_type = icon_headers["Content-Type"]
212 if content_type and (
213 content_type.startswith("application/")
214 or content_type.startswith("image/")
215 or content_type.startswith("text/plain")
217 bookmark.icon_href = icon_url
218 self.log(" got icon : %s"
221 content_type.startswith("application/")
222 or content_type.startswith(
225 self.log(" non-image content type,"
227 content_type = 'image/x-icon'
228 if not isinstance(icon_data, bytes):
229 icon_data = icon_data.encode('latin1')
230 bookmark.icon = "data:%s;base64,%s" \
231 % (content_type, b64encode(icon_data))
232 icons[icon_url] = (content_type,
236 self.log(" no icon : "
237 "bad content type '%s'"
240 icons[icon_url] = None
241 if parser and parser.refresh:
242 refresh = parser.refresh
244 url = refresh.split('=', 1)[1]
248 timeout = float(refresh.split(';')[0])
249 except (IndexError, ValueError):
250 self.set_redirect(bookmark, "html",
251 "Bad redirect to %s (%s)"
256 timeout = int(refresh.split(';')[0])
259 self.set_redirect(bookmark, "html",
264 bookmark.charset = charset
266 if not content_stripped:
267 self.log(" empty response, no content")
269 self.log(" not html")
270 except KeyError as key:
271 self.log(" no header: %s" % key)
274 if bookmark.href.startswith("ftp://"):
275 # Pass welcome message through MD5
276 ftp_welcome = self.get_ftp_welcome()
277 if not isinstance(ftp_welcome, bytes):
278 ftp_welcome = ftp_welcome.encode(charset or 'utf-8')
279 md5.update(ftp_welcome)
281 if isinstance(content, bytes):
284 md5.update(content.encode(charset or 'utf-8'))
285 bookmark.md5 = str(md5)
288 bookmark.error = "Unexpected EOF (FTP server closed connection)"
289 self.log(' EOF: %s' % bookmark.error)
291 except KeyboardInterrupt:
292 self.log("Keyboard interrupt (^C)")
295 except socket.error as e:
296 bookmark.error = get_error(e)
297 self.log(bookmark.error)
301 traceback.print_exc()
302 bookmark.error = "Exception!"
303 self.log(' Exception: %s' % bookmark.error)
306 self.finish_check_url(bookmark)
311 def set_redirect(self, bookmark, errcode, newurl):
312 bookmark.moved = moved = "(%s) to %s" % (reloc_dict[errcode], newurl)
314 moved.encode('ascii')
315 except UnicodeEncodeError:
317 moved = moved.encode(bookmark.charset)
318 except (LookupError, TypeError, UnicodeEncodeError):
319 moved = moved.encode('utf-8')
320 self.log(' Moved: %s' % moved)
322 def finish_check_url(self, bookmark):
324 bookmark.last_tested = str(start)
325 now = int(time.time())
326 bookmark.test_time = str(now - start)