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.

13 年之前
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611
  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 import logger
  17. from wiki.category import Category
  18. from wiki.constants import *
  19. from wiki.exceptions import *
  20. from wiki.page import Page
  21. from wiki.user import User
  22. class Site(object):
  23. """
  24. EarwigBot's Wiki Toolset: Site Class
  25. Represents a Site, with support for API queries and returning Pages, Users,
  26. and Categories. The constructor takes a bunch of arguments and you probably
  27. won't need to call it directly, rather tools.get_site() for returning Site
  28. instances, tools.add_site() for adding new ones to config, and
  29. tools.del_site() for removing old ones from config, should suffice.
  30. Public methods:
  31. name -- returns our name (or "wikiid"), like "enwiki"
  32. project -- returns our project name, like "wikipedia"
  33. lang -- returns our language code, like "en"
  34. domain -- returns our web domain, like "en.wikipedia.org"
  35. api_query -- does an API query with the given kwargs as params
  36. sql_query -- does an SQL query and yields its results
  37. get_replag -- returns the estimated database replication lag
  38. namespace_id_to_name -- given a namespace ID, returns associated name(s)
  39. namespace_name_to_id -- given a namespace name, returns associated id
  40. get_page -- returns a Page object for the given title
  41. get_category -- returns a Category object for the given title
  42. get_user -- returns a User object for the given username
  43. """
  44. def __init__(self, name=None, project=None, lang=None, base_url=None,
  45. article_path=None, script_path=None, sql=None,
  46. namespaces=None, login=(None, None), cookiejar=None,
  47. user_agent=None, assert_edit=None, maxlag=None):
  48. """Constructor for new Site instances.
  49. This probably isn't necessary to call yourself unless you're building a
  50. Site that's not in your config and you don't want to add it - normally
  51. all you need is tools.get_site(name), which creates the Site for you
  52. based on your config file. We accept a bunch of kwargs, but the only
  53. ones you really "need" are `base_url` and `script_path` - this is
  54. enough to figure out an API url. `login`, a tuple of
  55. (username, password), is highly recommended. `cookiejar` will be used
  56. to store cookies, and we'll use a normal CookieJar if none is given.
  57. First, we'll store the given arguments as attributes, then set up our
  58. URL opener. We'll load any of the attributes that weren't given from
  59. the API, and then log in if a username/pass was given and we aren't
  60. already logged in.
  61. """
  62. # Attributes referring to site information, filled in by an API query
  63. # if they are missing (and an API url can be determined):
  64. self._name = name
  65. self._project = project
  66. self._lang = lang
  67. self._base_url = base_url
  68. self._article_path = article_path
  69. self._script_path = script_path
  70. self._namespaces = namespaces
  71. # Attributes used for API queries:
  72. self._assert_edit = assert_edit
  73. self._maxlag = maxlag
  74. self._max_retries = 5
  75. # Attributes used for SQL queries:
  76. self._sql_data = sql
  77. self._sql_conn = None
  78. # Set up cookiejar and URL opener for making API queries:
  79. if cookiejar is not None:
  80. self._cookiejar = cookiejar
  81. else:
  82. self._cookiejar = CookieJar()
  83. if user_agent is None:
  84. user_agent = USER_AGENT # Set default UA from wiki.constants
  85. self._opener = build_opener(HTTPCookieProcessor(self._cookiejar))
  86. self._opener.addheaders = [("User-Agent", user_agent),
  87. ("Accept-Encoding", "gzip")]
  88. # Get all of the above attributes that were not specified as arguments:
  89. self._load_attributes()
  90. # If we have a name/pass and the API says we're not logged in, log in:
  91. self._login_info = name, password = login
  92. if name is not None and password is not None:
  93. logged_in_as = self._get_username_from_cookies()
  94. if logged_in_as is None or name != logged_in_as:
  95. self._login(login)
  96. def __repr__(self):
  97. """Returns the canonical string representation of the Site."""
  98. res = ", ".join((
  99. "Site(name={_name!r}", "project={_project!r}", "lang={_lang!r}",
  100. "base_url={_base_url!r}", "article_path={_article_path!r}",
  101. "script_path={_script_path!r}", "assert_edit={_assert_edit!r}",
  102. "maxlag={_maxlag!r}", "sql={_sql!r}", "login={0}",
  103. "user_agent={2!r}", "cookiejar={1})"
  104. ))
  105. name, password = self._login_info
  106. login = "({0}, {1})".format(repr(name), "hidden" if password else None)
  107. cookies = self._cookiejar.__class__.__name__
  108. try:
  109. cookies += "({0!r})".format(self._cookiejar.filename)
  110. except AttributeError:
  111. cookies += "()"
  112. agent = self._opener.addheaders[0][1]
  113. return res.format(login, cookies, agent, **self.__dict__)
  114. def __str__(self):
  115. """Returns a nice string representation of the Site."""
  116. res = "<Site {0} ({1}:{2}) at {3}>"
  117. return res.format(self.name(), self.project(), self.lang(),
  118. self.domain())
  119. def _api_query(self, params, tries=0, wait=5):
  120. """Do an API query with `params` as a dict of parameters.
  121. This will first attempt to construct an API url from self._base_url and
  122. self._script_path. We need both of these, or else we'll raise
  123. SiteAPIError.
  124. We'll encode the given params, adding format=json along the way, as
  125. well as &assert= and &maxlag= based on self._assert_edit and _maxlag.
  126. We make the request through self._opener, which has built-in cookie
  127. support via self._cookiejar, a User-Agent (wiki.constants.USER_AGENT),
  128. and Accept-Encoding set to "gzip".
  129. Assuming everything went well, we'll gunzip the data (if compressed),
  130. load it as a JSON object, and return it.
  131. If our request failed for some reason, we'll raise SiteAPIError with
  132. details. If that reason was due to maxlag, we'll sleep for a bit and
  133. then repeat the query until we exceed self._max_retries.
  134. There's helpful MediaWiki API documentation at
  135. <http://www.mediawiki.org/wiki/API>.
  136. """
  137. if self._base_url is None or self._script_path is None:
  138. e = "Tried to do an API query, but no API URL is known."
  139. raise SiteAPIError(e)
  140. url = ''.join((self._base_url, self._script_path, "/api.php"))
  141. params["format"] = "json" # This is the only format we understand
  142. if self._assert_edit: # If requested, ensure that we're logged in
  143. params["assert"] = self._assert_edit
  144. if self._maxlag: # If requested, don't overload the servers
  145. params["maxlag"] = self._maxlag
  146. data = urlencode(params)
  147. logger.debug("{0} -> {1}".format(url, data))
  148. try:
  149. response = self._opener.open(url, data)
  150. except URLError as error:
  151. if hasattr(error, "reason"):
  152. e = "API query failed: {0}.".format(error.reason)
  153. elif hasattr(error, "code"):
  154. e = "API query failed: got an error code of {0}."
  155. e = e.format(error.code)
  156. else:
  157. e = "API query failed."
  158. raise SiteAPIError(e)
  159. result = response.read()
  160. if response.headers.get("Content-Encoding") == "gzip":
  161. stream = StringIO(result)
  162. gzipper = GzipFile(fileobj=stream)
  163. result = gzipper.read()
  164. try:
  165. res = loads(result) # Parse as a JSON object
  166. except ValueError:
  167. e = "API query failed: JSON could not be decoded."
  168. raise SiteAPIError(e)
  169. try:
  170. code = res["error"]["code"]
  171. info = res["error"]["info"]
  172. except (TypeError, KeyError):
  173. return res
  174. if code == "maxlag":
  175. if tries >= self._max_retries:
  176. e = "Maximum number of retries reached ({0})."
  177. raise SiteAPIError(e.format(self._max_retries))
  178. tries += 1
  179. msg = 'Server says: "{0}". Retrying in {1} seconds ({2}/{3}).'
  180. logger.info(msg.format(info, wait, tries, self._max_retries))
  181. sleep(wait)
  182. return self._api_query(params, tries=tries, wait=wait*3)
  183. else:
  184. e = 'API query failed: got error "{0}"; server says: "{1}".'
  185. error = SiteAPIError(e.format(code, info))
  186. error.code, error.info = code, info
  187. raise error
  188. def _load_attributes(self, force=False):
  189. """Load data about our Site from the API.
  190. This function is called by __init__() when one of the site attributes
  191. was not given as a keyword argument. We'll do an API query to get the
  192. missing data, but only if there actually *is* missing data.
  193. Additionally, you can call this with `force=True` to forcibly reload
  194. all attributes.
  195. """
  196. # All attributes to be loaded, except _namespaces, which is a special
  197. # case because it requires additional params in the API query:
  198. attrs = [self._name, self._project, self._lang, self._base_url,
  199. self._article_path, self._script_path]
  200. params = {"action": "query", "meta": "siteinfo"}
  201. if not self._namespaces or force:
  202. params["siprop"] = "general|namespaces|namespacealiases"
  203. result = self._api_query(params)
  204. self._load_namespaces(result)
  205. elif all(attrs): # Everything is already specified and we're not told
  206. return # to force a reload, so do nothing
  207. else: # We're only loading attributes other than _namespaces
  208. params["siprop"] = "general"
  209. result = self._api_query(params)
  210. res = result["query"]["general"]
  211. self._name = res["wikiid"]
  212. self._project = res["sitename"].lower()
  213. self._lang = res["lang"]
  214. self._base_url = res["server"]
  215. self._article_path = res["articlepath"]
  216. self._script_path = res["scriptpath"]
  217. def _load_namespaces(self, result):
  218. """Fill self._namespaces with a dict of namespace IDs and names.
  219. Called by _load_attributes() with API data as `result` when
  220. self._namespaces was not given as an kwarg to __init__().
  221. """
  222. self._namespaces = {}
  223. for namespace in result["query"]["namespaces"].values():
  224. ns_id = namespace["id"]
  225. name = namespace["*"]
  226. try:
  227. canonical = namespace["canonical"]
  228. except KeyError:
  229. self._namespaces[ns_id] = [name]
  230. else:
  231. if name != canonical:
  232. self._namespaces[ns_id] = [name, canonical]
  233. else:
  234. self._namespaces[ns_id] = [name]
  235. for namespace in result["query"]["namespacealiases"]:
  236. ns_id = namespace["id"]
  237. alias = namespace["*"]
  238. self._namespaces[ns_id].append(alias)
  239. def _get_cookie(self, name, domain):
  240. """Return the named cookie unless it is expired or doesn't exist."""
  241. for cookie in self._cookiejar:
  242. if cookie.name == name and cookie.domain == domain:
  243. if cookie.is_expired():
  244. break
  245. return cookie
  246. def _get_username_from_cookies(self):
  247. """Try to return our username based solely on cookies.
  248. First, we'll look for a cookie named self._name + "Token", like
  249. "enwikiToken". If it exists and isn't expired, we'll assume it's valid
  250. and try to return the value of the cookie self._name + "UserName" (like
  251. "enwikiUserName"). This should work fine on wikis without single-user
  252. login.
  253. If `enwikiToken` doesn't exist, we'll try to find a cookie named
  254. `centralauth_Token`. If this exists and is not expired, we'll try to
  255. return the value of `centralauth_User`.
  256. If we didn't get any matches, we'll return None. Our goal here isn't to
  257. return the most likely username, or what we *want* our username to be
  258. (for that, we'd do self._login_info[0]), but rather to get our current
  259. username without an unnecessary ?action=query&meta=userinfo API query.
  260. """
  261. domain = self.domain()
  262. name = ''.join((self._name, "Token"))
  263. cookie = self._get_cookie(name, domain)
  264. if cookie is not None:
  265. name = ''.join((self._name, "UserName"))
  266. user_name = self._get_cookie(name, domain)
  267. if user_name is not None:
  268. return user_name.value
  269. name = "centralauth_Token"
  270. for cookie in self._cookiejar:
  271. if cookie.domain_initial_dot is False or cookie.is_expired():
  272. continue
  273. if cookie.name != name:
  274. continue
  275. # Build a regex that will match domains this cookie affects:
  276. search = ''.join(("(.*?)", re_escape(cookie.domain)))
  277. if re_match(search, domain): # Test it against our site
  278. user_name = self._get_cookie("centralauth_User", cookie.domain)
  279. if user_name is not None:
  280. return user_name.value
  281. def _get_username_from_api(self):
  282. """Do a simple API query to get our username and return it.
  283. This is a reliable way to make sure we are actually logged in, because
  284. it doesn't deal with annoying cookie logic, but it results in an API
  285. query that is unnecessary in some cases.
  286. Called by _get_username() (in turn called by get_user() with no
  287. username argument) when cookie lookup fails, probably indicating that
  288. we are logged out.
  289. """
  290. params = {"action": "query", "meta": "userinfo"}
  291. result = self._api_query(params)
  292. return result["query"]["userinfo"]["name"]
  293. def _get_username(self):
  294. """Return the name of the current user, whether logged in or not.
  295. First, we'll try to deduce it solely from cookies, to avoid an
  296. unnecessary API query. For the cookie-detection method, see
  297. _get_username_from_cookies()'s docs.
  298. If our username isn't in cookies, then we're probably not logged in, or
  299. something fishy is going on (like forced logout). In this case, do a
  300. single API query for our username (or IP address) and return that.
  301. """
  302. name = self._get_username_from_cookies()
  303. if name is not None:
  304. return name
  305. return self._get_username_from_api()
  306. def _save_cookiejar(self):
  307. """Try to save our cookiejar after doing a (normal) login or logout.
  308. Calls the standard .save() method with no filename. Don't fret if our
  309. cookiejar doesn't support saving (CookieJar raises AttributeError,
  310. FileCookieJar raises NotImplementedError) or no default filename was
  311. given (LWPCookieJar and MozillaCookieJar raise ValueError).
  312. """
  313. try:
  314. self._cookiejar.save()
  315. except (AttributeError, NotImplementedError, ValueError):
  316. pass
  317. def _login(self, login, token=None, attempt=0):
  318. """Safely login through the API.
  319. Normally, this is called by __init__() if a username and password have
  320. been provided and no valid login cookies were found. The only other
  321. time it needs to be called is when those cookies expire, which is done
  322. automatically by api_query() if a query fails.
  323. Recent versions of MediaWiki's API have fixed a CSRF vulnerability,
  324. requiring login to be done in two separate requests. If the response
  325. from from our initial request is "NeedToken", we'll do another one with
  326. the token. If login is successful, we'll try to save our cookiejar.
  327. Raises LoginError on login errors (duh), like bad passwords and
  328. nonexistent usernames.
  329. `login` is a (username, password) tuple. `token` is the token returned
  330. from our first request, and `attempt` is to prevent getting stuck in a
  331. loop if MediaWiki isn't acting right.
  332. """
  333. name, password = login
  334. params = {"action": "login", "lgname": name, "lgpassword": password}
  335. if token is not None:
  336. params["lgtoken"] = token
  337. result = self._api_query(params)
  338. res = result["login"]["result"]
  339. if res == "Success":
  340. self._save_cookiejar()
  341. elif res == "NeedToken" and attempt == 0:
  342. token = result["login"]["token"]
  343. return self._login(login, token, attempt=1)
  344. else:
  345. if res == "Illegal":
  346. e = "The provided username is illegal."
  347. elif res == "NotExists":
  348. e = "The provided username does not exist."
  349. elif res == "EmptyPass":
  350. e = "No password was given."
  351. elif res == "WrongPass" or res == "WrongPluginPass":
  352. e = "The given password is incorrect."
  353. else:
  354. e = "Couldn't login; server says '{0}'.".format(res)
  355. raise LoginError(e)
  356. def _logout(self):
  357. """Safely logout through the API.
  358. We'll do a simple API request (api.php?action=logout), clear our
  359. cookiejar (which probably contains now-invalidated cookies) and try to
  360. save it, if it supports that sort of thing.
  361. """
  362. params = {"action": "logout"}
  363. self._api_query(params)
  364. self._cookiejar.clear()
  365. self._save_cookiejar()
  366. def _sql_connect(self, **kwargs):
  367. """Attempt to establish a connection with this site's SQL database.
  368. oursql.connect() will be called with self._sql_data as its kwargs,
  369. which is usually config.wiki["sites"][self.name()]["sql"]. Any kwargs
  370. given to this function will be passed to connect() and will have
  371. precedence over the config file.
  372. Will raise SQLError() if the module "oursql" is not available. oursql
  373. may raise its own exceptions (e.g. oursql.InterfaceError) if it cannot
  374. establish a connection.
  375. """
  376. if not oursql:
  377. e = "Module 'oursql' is required for SQL queries."
  378. raise SQLError(e)
  379. args = self._sql_data
  380. for key, value in kwargs.iteritems():
  381. args[key] = value
  382. if "read_default_file" not in args and "user" not in args and "passwd" not in args:
  383. args["read_default_file"] = expanduser("~/.my.cnf")
  384. self._sql_conn = oursql.connect(**args)
  385. def name(self):
  386. """Returns the Site's name (or "wikiid" in the API), like "enwiki"."""
  387. return self._name
  388. def project(self):
  389. """Returns the Site's project name in lowercase, like "wikipedia"."""
  390. return self._project
  391. def lang(self):
  392. """Returns the Site's language code, like "en" or "es"."""
  393. return self._lang
  394. def domain(self):
  395. """Returns the Site's web domain, like "en.wikipedia.org"."""
  396. return urlparse(self._base_url).netloc
  397. def api_query(self, **kwargs):
  398. """Do an API query with `kwargs` as the parameters.
  399. See _api_query()'s documentation for details.
  400. """
  401. return self._api_query(kwargs)
  402. def sql_query(self, query, params=(), plain_query=False, dict_cursor=False,
  403. cursor_class=None, show_table=False):
  404. """Do an SQL query and yield its results.
  405. If `plain_query` is True, we will force an unparameterized query.
  406. Specifying both params and plain_query will cause an error.
  407. If `dict_cursor` is True, we will use oursql.DictCursor as our cursor,
  408. otherwise the default oursql.Cursor. If `cursor_class` is given, it
  409. will override this option.
  410. If `show_table` is True, the name of the table will be prepended to the
  411. name of the column. This will mainly affect a DictCursor.
  412. Example:
  413. >>> query = "SELECT user_id, user_registration FROM user WHERE user_name = ?"
  414. >>> params = ("The Earwig",)
  415. >>> result1 = site.sql_query(query, params)
  416. >>> result2 = site.sql_query(query, params, dict_cursor=True)
  417. >>> for row in result1: print row
  418. (7418060L, '20080703215134')
  419. >>> for row in result2: print row
  420. {'user_id': 7418060L, 'user_registration': '20080703215134'}
  421. See _sql_connect() for information on how a connection is acquired.
  422. <http://packages.python.org/oursql> has helpful documentation on the
  423. oursql module.
  424. This may raise SQLError() or one of oursql's exceptions
  425. (oursql.ProgrammingError, oursql.InterfaceError, ...) if there were
  426. problems with the query.
  427. """
  428. if not self._sql_conn:
  429. self._sql_connect()
  430. if not cursor_class:
  431. if dict_cursor:
  432. cursor_class = oursql.DictCursor
  433. else:
  434. cursor_class = oursql.Cursor
  435. with self._sql_conn.cursor(cursor_class, show_table=show_table) as cur:
  436. cur.execute(query, params, plain_query)
  437. for result in cur:
  438. yield result
  439. def get_replag(self):
  440. """Return the estimated database replication lag in seconds.
  441. Requires SQL access. This function only makes sense on a replicated
  442. database (e.g. the Wikimedia Toolserver) and on a wiki that receives a
  443. large number of edits (ideally, at least one per second), or the result
  444. may be larger than expected.
  445. """
  446. query = "SELECT NOW() - MAX(rev_timestamp) FROM revision"
  447. result = list(self.sql_query(query))
  448. return result[0][0]
  449. def namespace_id_to_name(self, ns_id, all=False):
  450. """Given a namespace ID, returns associated namespace names.
  451. If all is False (default), we'll return the first name in the list,
  452. which is usually the localized version. Otherwise, we'll return the
  453. entire list, which includes the canonical name.
  454. For example, returns u"Wikipedia" if ns_id=4 and all=False on enwiki;
  455. returns [u"Wikipedia", u"Project"] if ns_id=4 and all=True.
  456. Raises NamespaceNotFoundError if the ID is not found.
  457. """
  458. try:
  459. if all:
  460. return self._namespaces[ns_id]
  461. else:
  462. return self._namespaces[ns_id][0]
  463. except KeyError:
  464. e = "There is no namespace with id {0}.".format(ns_id)
  465. raise NamespaceNotFoundError(e)
  466. def namespace_name_to_id(self, name):
  467. """Given a namespace name, returns the associated ID.
  468. Like namespace_id_to_name(), but reversed. Case is ignored, because
  469. namespaces are assumed to be case-insensitive.
  470. Raises NamespaceNotFoundError if the name is not found.
  471. """
  472. lname = name.lower()
  473. for ns_id, names in self._namespaces.items():
  474. lnames = [n.lower() for n in names] # Be case-insensitive
  475. if lname in lnames:
  476. return ns_id
  477. e = "There is no namespace with name '{0}'.".format(name)
  478. raise NamespaceNotFoundError(e)
  479. def get_page(self, title, follow_redirects=False):
  480. """Returns a Page object for the given title (pagename).
  481. Will return a Category object instead if the given title is in the
  482. category namespace. As Category is a subclass of Page, this should not
  483. cause problems.
  484. Note that this doesn't do any direct checks for existence or
  485. redirect-following - Page's methods provide that.
  486. """
  487. prefixes = self.namespace_id_to_name(NS_CATEGORY, all=True)
  488. prefix = title.split(":", 1)[0]
  489. if prefix != title: # Avoid a page that is simply "Category"
  490. if prefix in prefixes:
  491. return Category(self, title, follow_redirects)
  492. return Page(self, title, follow_redirects)
  493. def get_category(self, catname, follow_redirects=False):
  494. """Returns a Category object for the given category name.
  495. `catname` should be given *without* a namespace prefix. This method is
  496. really just shorthand for get_page("Category:" + catname).
  497. """
  498. prefix = self.namespace_id_to_name(NS_CATEGORY)
  499. pagename = ':'.join((prefix, catname))
  500. return Category(self, pagename, follow_redirects)
  501. def get_user(self, username=None):
  502. """Returns a User object for the given username.
  503. If `username` is left as None, then a User object representing the
  504. currently logged-in (or anonymous!) user is returned.
  505. """
  506. if username is None:
  507. username = self._get_username()
  508. return User(self, username)