2 Simple, strightforward robot
4 Written by Oleg BroytMann. Copyright (C) 2000-2007 PhiloSoft Design.
9 from base64 import b64encode
10 from urlparse import urljoin
12 from m_lib.net.www.util import parse_time
13 from m_lib.md5wrapper import md5wrapper
15 from bkmk_objects import Robot
16 from parse_html import parse_html
19 class RedirectException(Exception):
25 def __init__(self, errcode, newurl):
26 Exception.__init__(self, "(%s) to %s" % (self.reloc_dict[errcode], newurl))
30 class MyURLopener(urllib.URLopener):
31 # Error 302 -- relocated (temporarily)
32 def http_error_302(self, url, fp, errcode, errmsg, headers, data=None):
33 if headers.has_key('location'):
34 newurl = headers['location']
35 elif headers.has_key('uri'):
36 newurl = headers['uri']
39 raise RedirectException(errcode, newurl)
41 # Error 301 -- also relocated (permanently)
42 http_error_301 = http_error_302
44 # Error 401 -- authentication required
45 def http_error_401(self, url, fp, errcode, errmsg, headers, data=None):
46 raise IOError, ('http error', errcode, "Authentication required ", headers)
49 urllib._urlopener = MyURLopener()
51 # Fake headers to pretend this is a real browser
52 _version = "Mozilla/5.0 (X11; U; Linux 2.6 i686; en) Gecko/20001221 Firefox/2.0.0"
53 urllib._urlopener.addheaders[0] = ('User-Agent', _version)
54 _version = "bookmarks_db (Python %d.%d.%d; urllib/%s)" % (
55 sys.version_info[0], sys.version_info[1], sys.version_info[2], urllib.__version__)
56 urllib._urlopener.addheader('X-User-Agent', _version)
57 urllib._urlopener.addheader('Referer', '')
59 urllib._urlopener.addheader('Connection', 'close')
60 urllib._urlopener.addheader('Accept', '*/*')
61 urllib._urlopener.addheader('Accept-Language', 'ru,en')
62 urllib._urlopener.addheader('Cache-Control', 'max-age=300')
66 if isinstance(msg, str):
72 s.append("'%s'" % str(i).replace('\n', "\\n"))
73 return "(%s)" % ' '.join(s)
76 urllib_ftpwrapper = urllib.ftpwrapper
79 class myftpwrapper(urllib_ftpwrapper):
80 def __init__(self, user, passwd, host, port, dirs):
81 urllib_ftpwrapper.__init__(self, user, passwd, host, port, dirs)
83 ftpcache_key = (user, host, port, '/'.join(dirs))
85 urllib.ftpwrapper = myftpwrapper
89 _welcome = urllib._urlopener.ftpcache[ftpcache_key].ftp.welcome
90 ftpcache_key = None # I am assuming there are no duplicate ftp URLs in db.
91 # If there are - ftpcache_key in prev line is invalid.
95 icons = {} # Icon cache; maps URL to a tuple (content type, data)
96 # or None if there is no icon.
98 class robot_simple(Robot):
99 def check_url(self, bookmark):
100 if not self.tempfname:
101 self.tempfname = bookmark.tempfname
105 self.start = int(time.time())
108 url_type, url_rest = urllib.splittype(bookmark.href)
109 url_host, url_path = urllib.splithost(url_rest)
110 url_path, url_tag = urllib.splittag(url_path)
112 # Set fake referer to the root of the site
113 urllib._urlopener.addheaders[2] = ('Referer', "%s://%s%s" % (url_type, url_host, url_path))
115 if bookmark.charset: urllib._urlopener.addheader('Accept-Charset', bookmark.charset)
116 fname, headers = urllib.urlretrieve("%s://%s%s" % (url_type, url_host, url_path), self.tempfname)
117 if bookmark.charset: del urllib._urlopener.addheaders[-1]
124 size = headers["Content-Length"]
129 last_modified = headers["Last-Modified"]
134 last_modified = parse_time(last_modified)
137 last_modified = str(int(last_modified))
139 last_modified = bookmark.last_visit
142 bookmark.last_modified = last_modified
145 if urllib._urlopener.type == "ftp": # Pass welcome message through MD5
146 md5.update(get_welcome())
148 md5.md5file(self.tempfname)
149 bookmark.md5 = str(md5)
153 content_type = headers["Content-Type"]
155 content_type, charset = content_type.split(';')
156 content_type = content_type.strip()
157 charset = charset.split('=')[1].strip()
158 self.log(" HTTP charset : %s" % charset)
159 except (ValueError, IndexError):
161 self.log(" no charset in Content-Type header")
162 if content_type == "text/html":
163 parser = parse_html(fname, charset, self.log)
164 bookmark.real_title = parser.title
166 refresh = parser.refresh
168 url = refresh.split('=', 1)[1]
172 timeout = float(refresh.split(';')[0])
173 except (IndexError, ValueError):
174 raise RedirectException("html", "Bad redirect to %s (%s)" % (url, refresh))
177 timeout = int(refresh.split(';')[0])
180 raise RedirectException("html", "%s (%s sec)" % (url, timeout))
185 icon = "/favicon.ico"
186 icon = urljoin("%s://%s%s" % (url_type, url_host, url_path), icon)
187 self.log(" looking for icon at: %s" % icon)
190 content_type, bookmark.icon = icons[icon]
191 self.log(" cached icon: %s" % content_type)
193 self.log(" cached icon: no icon")
199 fname, headers = urllib.urlretrieve(_icon)
200 except RedirectException, e:
202 self.log(" redirect to : %s" % _icon)
206 raise IOError("Too many redirects")
208 etype, emsg, tb = sys.exc_info()
209 self.log(" no icon : %s %s" % (etype, emsg))
215 content_type = headers["Content-Type"]
216 if content_type.startswith("image/"):
217 icon_file = open(fname, "rb")
218 icon = icon_file.read()
220 bookmark.icon = "data:%s;base64,%s" % (content_type, b64encode(icon))
221 self.log(" got icon : %s" % content_type)
222 icons[icon] = (content_type, bookmark.icon)
224 self.log(" no icon : bad content type '%s'" % content_type)
232 if (msg[0] == "http error") and (msg[1] == -1):
233 bookmark.no_error = "The server did not return any header - it is not an error, actually"
235 bookmark.error = get_error(msg)
238 bookmark.error = "Unexpected EOF (FTP server closed connection)"
240 except RedirectException, msg:
241 bookmark.moved = str(msg)
243 except KeyboardInterrupt:
247 self.finish_check_url(bookmark)
253 def finish_check_url(self, bookmark):
254 # Calculate these attributes even in case of an error
255 if os.path.exists(self.tempfname):
256 size = str(os.path.getsize(self.tempfname))
262 bookmark.last_tested = str(start)
264 now = int(time.time())
265 bookmark.test_time = str(now - start)