A Python robot that edits Wikipedia and interacts with people over IRC https://en.wikipedia.org/wiki/User:EarwigBot
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

site.py 24 KiB

13 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597
  1. # -*- coding: utf-8 -*-
  2. from cookielib import CookieJar
  3. from gzip import GzipFile
  4. from json import loads
  5. from os.path import expanduser
  6. from re import escape as re_escape, match as re_match
  7. from StringIO import StringIO
  8. from time import sleep
  9. from urllib import unquote_plus, urlencode
  10. from urllib2 import build_opener, HTTPCookieProcessor, URLError
  11. from urlparse import urlparse
  12. try:
  13. import oursql
  14. except ImportError:
  15. oursql = None
  16. from wiki.category import Category
  17. from wiki.constants import *
  18. from wiki.exceptions import *
  19. from wiki.page import Page
  20. from wiki.user import User
  21. class Site(object):
  22. """
  23. EarwigBot's Wiki Toolset: Site Class
  24. Represents a Site, with support for API queries and returning Pages, Users,
  25. and Categories. The constructor takes a bunch of arguments and you probably
  26. won't need to call it directly, rather tools.get_site() for returning Site
  27. instances, tools.add_site() for adding new ones to config, and
  28. tools.del_site() for removing old ones from config, should suffice.
  29. Public methods:
  30. name -- returns our name (or "wikiid"), like "enwiki"
  31. project -- returns our project name, like "wikipedia"
  32. lang -- returns our language code, like "en"
  33. domain -- returns our web domain, like "en.wikipedia.org"
  34. api_query -- does an API query with the given kwargs as params
  35. sql_query -- does an SQL query and yields its results
  36. namespace_id_to_name -- given a namespace ID, returns associated name(s)
  37. namespace_name_to_id -- given a namespace name, returns associated id
  38. get_page -- returns a Page object for the given title
  39. get_category -- returns a Category object for the given title
  40. get_user -- returns a User object for the given username
  41. """
  42. def __init__(self, name=None, project=None, lang=None, base_url=None,
  43. article_path=None, script_path=None, sql=None,
  44. namespaces=None, login=(None, None), cookiejar=None,
  45. user_agent=None, assert_edit=None, maxlag=None):
  46. """Constructor for new Site instances.
  47. This probably isn't necessary to call yourself unless you're building a
  48. Site that's not in your config and you don't want to add it - normally
  49. all you need is tools.get_site(name), which creates the Site for you
  50. based on your config file. We accept a bunch of kwargs, but the only
  51. ones you really "need" are `base_url` and `script_path` - this is
  52. enough to figure out an API url. `login`, a tuple of
  53. (username, password), is highly recommended. `cookiejar` will be used
  54. to store cookies, and we'll use a normal CookieJar if none is given.
  55. First, we'll store the given arguments as attributes, then set up our
  56. URL opener. We'll load any of the attributes that weren't given from
  57. the API, and then log in if a username/pass was given and we aren't
  58. already logged in.
  59. """
  60. # Attributes referring to site information, filled in by an API query
  61. # if they are missing (and an API url can be determined):
  62. self._name = name
  63. self._project = project
  64. self._lang = lang
  65. self._base_url = base_url
  66. self._article_path = article_path
  67. self._script_path = script_path
  68. self._namespaces = namespaces
  69. # Attributes used for API queries:
  70. self._assert_edit = assert_edit
  71. self._maxlag = maxlag
  72. self._max_retries = 5
  73. # Attributes used for SQL queries:
  74. self._sql_data = sql
  75. self._sql_conn = None
  76. # Set up cookiejar and URL opener for making API queries:
  77. if cookiejar is not None:
  78. self._cookiejar = cookiejar
  79. else:
  80. self._cookiejar = CookieJar()
  81. if user_agent is None:
  82. user_agent = USER_AGENT # Set default UA from wiki.constants
  83. self._opener = build_opener(HTTPCookieProcessor(self._cookiejar))
  84. self._opener.addheaders = [("User-Agent", user_agent),
  85. ("Accept-Encoding", "gzip")]
  86. # Get all of the above attributes that were not specified as arguments:
  87. self._load_attributes()
  88. # If we have a name/pass and the API says we're not logged in, log in:
  89. self._login_info = name, password = login
  90. if name is not None and password is not None:
  91. logged_in_as = self._get_username_from_cookies()
  92. if logged_in_as is None or name != logged_in_as:
  93. self._login(login)
  94. def __repr__(self):
  95. """Returns the canonical string representation of the Site."""
  96. res = ", ".join((
  97. "Site(name={_name!r}", "project={_project!r}", "lang={_lang!r}",
  98. "base_url={_base_url!r}", "article_path={_article_path!r}",
  99. "script_path={_script_path!r}", "assert_edit={_assert_edit!r}",
  100. "maxlag={_maxlag!r}", "sql={_sql!r}", "login={0}",
  101. "user_agent={2!r}", "cookiejar={1})"
  102. ))
  103. name, password = self._login_info
  104. login = "({0}, {1})".format(repr(name), "hidden" if password else None)
  105. cookies = self._cookiejar.__class__.__name__
  106. try:
  107. cookies += "({0!r})".format(self._cookiejar.filename)
  108. except AttributeError:
  109. cookies += "()"
  110. agent = self._opener.addheaders[0][1]
  111. return res.format(login, cookies, agent, **self.__dict__)
  112. def __str__(self):
  113. """Returns a nice string representation of the Site."""
  114. res = "<Site {0} ({1}:{2}) at {3}>"
  115. return res.format(self.name(), self.project(), self.lang(),
  116. self.domain())
  117. def _api_query(self, params, tries=0, wait=5):
  118. """Do an API query with `params` as a dict of parameters.
  119. This will first attempt to construct an API url from self._base_url and
  120. self._script_path. We need both of these, or else we'll raise
  121. SiteAPIError.
  122. We'll encode the given params, adding format=json along the way, as
  123. well as &assert= and &maxlag= based on self._assert_edit and _maxlag.
  124. We make the request through self._opener, which has built-in cookie
  125. support via self._cookiejar, a User-Agent (wiki.constants.USER_AGENT),
  126. and Accept-Encoding set to "gzip".
  127. Assuming everything went well, we'll gunzip the data (if compressed),
  128. load it as a JSON object, and return it.
  129. If our request failed for some reason, we'll raise SiteAPIError with
  130. details. If that reason was due to maxlag, we'll sleep for a bit and
  131. then repeat the query until we exceed self._max_retries.
  132. There's helpful MediaWiki API documentation at
  133. <http://www.mediawiki.org/wiki/API>.
  134. """
  135. if self._base_url is None or self._script_path is None:
  136. e = "Tried to do an API query, but no API URL is known."
  137. raise SiteAPIError(e)
  138. url = ''.join((self._base_url, self._script_path, "/api.php"))
  139. params["format"] = "json" # This is the only format we understand
  140. if self._assert_edit: # If requested, ensure that we're logged in
  141. params["assert"] = self._assert_edit
  142. if self._maxlag: # If requested, don't overload the servers
  143. params["maxlag"] = self._maxlag
  144. data = urlencode(params)
  145. print url, data # debug code
  146. try:
  147. response = self._opener.open(url, data)
  148. except URLError as error:
  149. if hasattr(error, "reason"):
  150. e = "API query failed: {0}.".format(error.reason)
  151. elif hasattr(error, "code"):
  152. e = "API query failed: got an error code of {0}."
  153. e = e.format(error.code)
  154. else:
  155. e = "API query failed."
  156. raise SiteAPIError(e)
  157. result = response.read()
  158. if response.headers.get("Content-Encoding") == "gzip":
  159. stream = StringIO(result)
  160. gzipper = GzipFile(fileobj=stream)
  161. result = gzipper.read()
  162. try:
  163. res = loads(result) # Parse as a JSON object
  164. except ValueError:
  165. e = "API query failed: JSON could not be decoded."
  166. raise SiteAPIError(e)
  167. try:
  168. code = res["error"]["code"]
  169. info = res["error"]["info"]
  170. except (TypeError, KeyError):
  171. return res
  172. if code == "maxlag":
  173. if tries >= self._max_retries:
  174. e = "Maximum number of retries reached ({0})."
  175. raise SiteAPIError(e.format(self._max_retries))
  176. tries += 1
  177. msg = 'Server says: "{0}". Retrying in {1} seconds ({2}/{3}).'
  178. print msg.format(info, wait, tries, self._max_retries)
  179. sleep(wait)
  180. return self._api_query(params, tries=tries, wait=wait*3)
  181. else:
  182. e = 'API query failed: got error "{0}"; server says: "{1}".'
  183. error = SiteAPIError(e.format(code, info))
  184. error.code, error.info = code, info
  185. raise error
  186. def _load_attributes(self, force=False):
  187. """Load data about our Site from the API.
  188. This function is called by __init__() when one of the site attributes
  189. was not given as a keyword argument. We'll do an API query to get the
  190. missing data, but only if there actually *is* missing data.
  191. Additionally, you can call this with `force=True` to forcibly reload
  192. all attributes.
  193. """
  194. # All attributes to be loaded, except _namespaces, which is a special
  195. # case because it requires additional params in the API query:
  196. attrs = [self._name, self._project, self._lang, self._base_url,
  197. self._article_path, self._script_path]
  198. params = {"action": "query", "meta": "siteinfo"}
  199. if not self._namespaces or force:
  200. params["siprop"] = "general|namespaces|namespacealiases"
  201. result = self._api_query(params)
  202. self._load_namespaces(result)
  203. elif all(attrs): # Everything is already specified and we're not told
  204. return # to force a reload, so do nothing
  205. else: # We're only loading attributes other than _namespaces
  206. params["siprop"] = "general"
  207. result = self._api_query(params)
  208. res = result["query"]["general"]
  209. self._name = res["wikiid"]
  210. self._project = res["sitename"].lower()
  211. self._lang = res["lang"]
  212. self._base_url = res["server"]
  213. self._article_path = res["articlepath"]
  214. self._script_path = res["scriptpath"]
  215. def _load_namespaces(self, result):
  216. """Fill self._namespaces with a dict of namespace IDs and names.
  217. Called by _load_attributes() with API data as `result` when
  218. self._namespaces was not given as an kwarg to __init__().
  219. """
  220. self._namespaces = {}
  221. for namespace in result["query"]["namespaces"].values():
  222. ns_id = namespace["id"]
  223. name = namespace["*"]
  224. try:
  225. canonical = namespace["canonical"]
  226. except KeyError:
  227. self._namespaces[ns_id] = [name]
  228. else:
  229. if name != canonical:
  230. self._namespaces[ns_id] = [name, canonical]
  231. else:
  232. self._namespaces[ns_id] = [name]
  233. for namespace in result["query"]["namespacealiases"]:
  234. ns_id = namespace["id"]
  235. alias = namespace["*"]
  236. self._namespaces[ns_id].append(alias)
  237. def _get_cookie(self, name, domain):
  238. """Return the named cookie unless it is expired or doesn't exist."""
  239. for cookie in self._cookiejar:
  240. if cookie.name == name and cookie.domain == domain:
  241. if cookie.is_expired():
  242. break
  243. return cookie
  244. def _get_username_from_cookies(self):
  245. """Try to return our username based solely on cookies.
  246. First, we'll look for a cookie named self._name + "Token", like
  247. "enwikiToken". If it exists and isn't expired, we'll assume it's valid
  248. and try to return the value of the cookie self._name + "UserName" (like
  249. "enwikiUserName"). This should work fine on wikis without single-user
  250. login.
  251. If `enwikiToken` doesn't exist, we'll try to find a cookie named
  252. `centralauth_Token`. If this exists and is not expired, we'll try to
  253. return the value of `centralauth_User`.
  254. If we didn't get any matches, we'll return None. Our goal here isn't to
  255. return the most likely username, or what we *want* our username to be
  256. (for that, we'd do self._login_info[0]), but rather to get our current
  257. username without an unnecessary ?action=query&meta=userinfo API query.
  258. """
  259. domain = self.domain()
  260. name = ''.join((self._name, "Token"))
  261. cookie = self._get_cookie(name, domain)
  262. if cookie is not None:
  263. name = ''.join((self._name, "UserName"))
  264. user_name = self._get_cookie(name, domain)
  265. if user_name is not None:
  266. return user_name.value
  267. name = "centralauth_Token"
  268. for cookie in self._cookiejar:
  269. if cookie.domain_initial_dot is False or cookie.is_expired():
  270. continue
  271. if cookie.name != name:
  272. continue
  273. # Build a regex that will match domains this cookie affects:
  274. search = ''.join(("(.*?)", re_escape(cookie.domain)))
  275. if re_match(search, domain): # Test it against our site
  276. user_name = self._get_cookie("centralauth_User", cookie.domain)
  277. if user_name is not None:
  278. return user_name.value
  279. def _get_username_from_api(self):
  280. """Do a simple API query to get our username and return it.
  281. This is a reliable way to make sure we are actually logged in, because
  282. it doesn't deal with annoying cookie logic, but it results in an API
  283. query that is unnecessary in some cases.
  284. Called by _get_username() (in turn called by get_user() with no
  285. username argument) when cookie lookup fails, probably indicating that
  286. we are logged out.
  287. """
  288. params = {"action": "query", "meta": "userinfo"}
  289. result = self._api_query(params)
  290. return result["query"]["userinfo"]["name"]
  291. def _get_username(self):
  292. """Return the name of the current user, whether logged in or not.
  293. First, we'll try to deduce it solely from cookies, to avoid an
  294. unnecessary API query. For the cookie-detection method, see
  295. _get_username_from_cookies()'s docs.
  296. If our username isn't in cookies, then we're probably not logged in, or
  297. something fishy is going on (like forced logout). In this case, do a
  298. single API query for our username (or IP address) and return that.
  299. """
  300. name = self._get_username_from_cookies()
  301. if name is not None:
  302. return name
  303. return self._get_username_from_api()
  304. def _save_cookiejar(self):
  305. """Try to save our cookiejar after doing a (normal) login or logout.
  306. Calls the standard .save() method with no filename. Don't fret if our
  307. cookiejar doesn't support saving (CookieJar raises AttributeError,
  308. FileCookieJar raises NotImplementedError) or no default filename was
  309. given (LWPCookieJar and MozillaCookieJar raise ValueError).
  310. """
  311. try:
  312. self._cookiejar.save()
  313. except (AttributeError, NotImplementedError, ValueError):
  314. pass
  315. def _login(self, login, token=None, attempt=0):
  316. """Safely login through the API.
  317. Normally, this is called by __init__() if a username and password have
  318. been provided and no valid login cookies were found. The only other
  319. time it needs to be called is when those cookies expire, which is done
  320. automatically by api_query() if a query fails.
  321. Recent versions of MediaWiki's API have fixed a CSRF vulnerability,
  322. requiring login to be done in two separate requests. If the response
  323. from from our initial request is "NeedToken", we'll do another one with
  324. the token. If login is successful, we'll try to save our cookiejar.
  325. Raises LoginError on login errors (duh), like bad passwords and
  326. nonexistent usernames.
  327. `login` is a (username, password) tuple. `token` is the token returned
  328. from our first request, and `attempt` is to prevent getting stuck in a
  329. loop if MediaWiki isn't acting right.
  330. """
  331. name, password = login
  332. params = {"action": "login", "lgname": name, "lgpassword": password}
  333. if token is not None:
  334. params["lgtoken"] = token
  335. result = self._api_query(params)
  336. res = result["login"]["result"]
  337. if res == "Success":
  338. self._save_cookiejar()
  339. elif res == "NeedToken" and attempt == 0:
  340. token = result["login"]["token"]
  341. return self._login(login, token, attempt=1)
  342. else:
  343. if res == "Illegal":
  344. e = "The provided username is illegal."
  345. elif res == "NotExists":
  346. e = "The provided username does not exist."
  347. elif res == "EmptyPass":
  348. e = "No password was given."
  349. elif res == "WrongPass" or res == "WrongPluginPass":
  350. e = "The given password is incorrect."
  351. else:
  352. e = "Couldn't login; server says '{0}'.".format(res)
  353. raise LoginError(e)
  354. def _logout(self):
  355. """Safely logout through the API.
  356. We'll do a simple API request (api.php?action=logout), clear our
  357. cookiejar (which probably contains now-invalidated cookies) and try to
  358. save it, if it supports that sort of thing.
  359. """
  360. params = {"action": "logout"}
  361. self._api_query(params)
  362. self._cookiejar.clear()
  363. self._save_cookiejar()
  364. def _sql_connect(self, **kwargs):
  365. """Attempt to establish a connection with this site's SQL database.
  366. oursql.connect() will be called with self._sql_data as its kwargs,
  367. which is usually config.wiki["sites"][self.name()]["sql"]. Any kwargs
  368. given to this function will be passed to connect() and will have
  369. precedence over the config file.
  370. Will raise SQLError() if the module "oursql" is not available. oursql
  371. may raise its own exceptions (e.g. oursql.InterfaceError) if it cannot
  372. establish a connection.
  373. """
  374. if not oursql:
  375. e = "Module 'oursql' is required for SQL queries."
  376. raise SQLError(e)
  377. args = self._sql_data
  378. for key, value in kwargs.iteritems():
  379. args[key] = value
  380. if "read_default_file" not in args and "user" not in args and "passwd" not in args:
  381. args["read_default_file"] = expanduser("~/.my.cnf")
  382. self._sql_conn = oursql.connect(**args)
  383. def name(self):
  384. """Returns the Site's name (or "wikiid" in the API), like "enwiki"."""
  385. return self._name
  386. def project(self):
  387. """Returns the Site's project name in lowercase, like "wikipedia"."""
  388. return self._project
  389. def lang(self):
  390. """Returns the Site's language code, like "en" or "es"."""
  391. return self._lang
  392. def domain(self):
  393. """Returns the Site's web domain, like "en.wikipedia.org"."""
  394. return urlparse(self._base_url).netloc
  395. def api_query(self, **kwargs):
  396. """Do an API query with `kwargs` as the parameters.
  397. See _api_query()'s documentation for details.
  398. """
  399. return self._api_query(kwargs)
  400. def sql_query(self, query, params=(), plain_query=False, dict_cursor=False,
  401. cursor_class=None, show_table=False):
  402. """Do an SQL query and yield its results.
  403. If `plain_query` is True, we will force an unparameterized query.
  404. Specifying both params and plain_query will cause an error.
  405. If `dict_cursor` is True, we will use oursql.DictCursor as our cursor,
  406. otherwise the default oursql.Cursor. If `cursor_class` is given, it
  407. will override this option.
  408. If `show_table` is True, the name of the table will be prepended to the
  409. name of the column. This will mainly affect a DictCursor.
  410. Example:
  411. >>> query = "SELECT user_id, user_registration FROM user WHERE user_name = ?"
  412. >>> params = ("The Earwig",)
  413. >>> result1 = site.sql_query(query, params)
  414. >>> result2 = site.sql_query(query, params, dict_cursor=True)
  415. >>> for row in result1: print row
  416. (7418060L, '20080703215134')
  417. >>> for row in result2: print row
  418. {'user_id': 7418060L, 'user_registration': '20080703215134'}
  419. See _sql_connect() for information on how a connection is acquired.
  420. <http://packages.python.org/oursql> has helpful documentation on the
  421. oursql module.
  422. This may raise SQLError() or one of oursql's exceptions
  423. (oursql.ProgrammingError, oursql.InterfaceError, ...) if there were
  424. problems with the query.
  425. """
  426. if not self._sql_conn:
  427. self._sql_connect()
  428. if not cursor_class:
  429. if dict_cursor:
  430. cursor_class = oursql.DictCursor
  431. else:
  432. cursor_class = oursql.Cursor
  433. with self._sql_conn.cursor(cursor_class, show_table=show_table) as cur:
  434. cur.execute(query, params, plain_query)
  435. for result in cur:
  436. yield result
  437. def namespace_id_to_name(self, ns_id, all=False):
  438. """Given a namespace ID, returns associated namespace names.
  439. If all is False (default), we'll return the first name in the list,
  440. which is usually the localized version. Otherwise, we'll return the
  441. entire list, which includes the canonical name.
  442. For example, returns u"Wikipedia" if ns_id=4 and all=False on enwiki;
  443. returns [u"Wikipedia", u"Project"] if ns_id=4 and all=True.
  444. Raises NamespaceNotFoundError if the ID is not found.
  445. """
  446. try:
  447. if all:
  448. return self._namespaces[ns_id]
  449. else:
  450. return self._namespaces[ns_id][0]
  451. except KeyError:
  452. e = "There is no namespace with id {0}.".format(ns_id)
  453. raise NamespaceNotFoundError(e)
  454. def namespace_name_to_id(self, name):
  455. """Given a namespace name, returns the associated ID.
  456. Like namespace_id_to_name(), but reversed. Case is ignored, because
  457. namespaces are assumed to be case-insensitive.
  458. Raises NamespaceNotFoundError if the name is not found.
  459. """
  460. lname = name.lower()
  461. for ns_id, names in self._namespaces.items():
  462. lnames = [n.lower() for n in names] # Be case-insensitive
  463. if lname in lnames:
  464. return ns_id
  465. e = "There is no namespace with name '{0}'.".format(name)
  466. raise NamespaceNotFoundError(e)
  467. def get_page(self, title, follow_redirects=False):
  468. """Returns a Page object for the given title (pagename).
  469. Will return a Category object instead if the given title is in the
  470. category namespace. As Category is a subclass of Page, this should not
  471. cause problems.
  472. Note that this doesn't do any direct checks for existence or
  473. redirect-following - Page's methods provide that.
  474. """
  475. prefixes = self.namespace_id_to_name(NS_CATEGORY, all=True)
  476. prefix = title.split(":", 1)[0]
  477. if prefix != title: # Avoid a page that is simply "Category"
  478. if prefix in prefixes:
  479. return Category(self, title, follow_redirects)
  480. return Page(self, title, follow_redirects)
  481. def get_category(self, catname, follow_redirects=False):
  482. """Returns a Category object for the given category name.
  483. `catname` should be given *without* a namespace prefix. This method is
  484. really just shorthand for get_page("Category:" + catname).
  485. """
  486. prefix = self.namespace_id_to_name(NS_CATEGORY)
  487. pagename = ':'.join((prefix, catname))
  488. return Category(self, pagename, follow_redirects)
  489. def get_user(self, username=None):
  490. """Returns a User object for the given username.
  491. If `username` is left as None, then a User object representing the
  492. currently logged-in (or anonymous!) user is returned.
  493. """
  494. if username is None:
  495. username = self._get_username()
  496. return User(self, username)