Last modified: 2014-06-16 22:07:43 UTC
Created attachment 15664 [details] users with most revisions with 0 userid per wiki (+ overall counts per wiki) The issue surfaced when edits per wiki per month per namespace were merged into one file for all 800 wikis. Lines in merged editor file: 55351997 Lines with user id 0: 603446 (1.1% of total). Often one user has revisions with userid 0 and revisions with the proper userid in same stub dump. In some wikis hundreds of user have revisions with userid 0. E.g.check historic stub dump for mediawikiwiki : http://dumps.wikimedia.org/mediawikiwiki/20140605/mediawikiwiki-20140605-stub-meta-history.xml.gz First field after user name is userid, for Erik Zachte sometimes 20226, sometimes 0. I checked the actual dump content: it is in the dumps. Erik Zachte 0 2004-01 wx mediawiki 100 3 Erik Zachte 0 2004-03 wx mediawiki 0 5 Erik Zachte 20226 2004-03 wx mediawiki 102 10 Erik Zachte 20226 2004-04 wx mediawiki 0 2 Erik Zachte 20226 2004-04 wx mediawiki 102 1 Erik Zachte 0 2004-05 wx mediawiki 0 3 Erik Zachte 20226 2004-05 wx mediawiki 102 1 Erik Zachte 20226 2004-06 wx mediawiki 102 25 Erik Zachte 20226 2004-06 wx mediawiki 103 7 Erik Zachte 0 2004-07 wx mediawiki 0 1 Erik Zachte 20226 2004-07 wx mediawiki 102 10 See also https://trello.com/c/3ecjp9aM/237-master-monthly-editor-activity-data
This can happen for edits imported by Special:Import prior to the user creating an account on the wiki. Although 1% seems much higher rate then I would expect for that situation.
