I just wanted to put() a simple page on a MediaWiki 1.16 instance, where I have to use screen scraping (use_api=False).

There is something strange however:

There is an API call invoked by _getBlocked:

/w/api.php?action=query&format=json&meta=userinfo&uiprop=blockinfo

Here's my backtrace:

File "pywikipedia/wikipedia.py", line 693, in get expandtemplates = expandtemplates)

File "pywikipedia/wikipedia.py", line 743, in _getEditPage return self._getEditPageOld(get_redirect, throttle, sysop, oldid, change_edit_time)

File "pywikipedia/wikipedia.py", line 854, in _getEditPageOld text = self.site().getUrl(path, sysop = sysop)

File "pywikipedia/wikipedia.py", line 5881, in getUrl self._getUserDataOld(text, sysop = sysop)

File "pywikipedia/wikipedia.py", line 6016, in _getUserDataOld blocked = self._getBlock(sysop = sysop)

File "pywikipedia/wikipedia.py", line 5424, in _getBlock data = query.GetData(params, self)

File "pywikipedia/query.py", line 146, in GetData jsontext = site.getUrl( path, retry=True, sysop=sysop, data=data)

getUrl(), which is also called from API, seems always to call _getUserDataOld(text) where text is ... API output so it tries to do strange things on that and gives warnings like

Note: this language does not allow global bots. WARNING: Token not found on wikipedia:pl. You will not be able to edit any page.

which is nonsense since the analyzed text is not HTML - only API output.

If getUrl() is supposed to be a low-level call, why call _getUserDataOld() there?

http://www.mediawiki.org/wiki/Special:Code/pywikipedia/7461

has introduced this call there.

It's easily reproducable by this:

import wikipedia import config config.use_api = False wikipedia.verbose = True s = wikipedia.getSite("pl", "wikipedia") p = wikipedia.Page(s, u"User:Saper") c = p.get() c += "<!-- test -->" p.put(c, u"Testing wiki", botflag=False)

//Saper