RolandUnger added a comment.

  It seems that it is a sudden spike. The computing time only for 
`addStatementUsage` jumped from about 600 ms to 5000 ms (all Lua stuff from 
about 4.5 sec to about 9 sec). Within the last two weeks I added and optimized 
the analysis of opening hours to our vCard template. As in the past I used the 
same technique to //prevent// retrieving data from Wikidata. We are using 
several tables like this: https://de.wikivoyage.org/wiki/Modul:Hours/i18n with 
prepared data. On this way there is usually only one getEntity call per 
template. The number of entities used increased within the last two weeks only 
by about 1. Of course I added some statements to the Wikidata datasets (mainly 
of museums) to test the display of opening hours. Within the last two weeks 
there was no significant change. The core functions were added already on Sep. 
26, 2020.
  
  There is another article with the same behavior: 
https://de.wikivoyage.org/wiki/Eisenbahnmuseen_in_Europa. For this article, I 
got today in the morning a runtime error that there was not enough time.
  
    Lua Profile:
        Scribunto_LuaSandboxCallback::addStatementUsage                 5580 ms 
      68.7%
        Scribunto_LuaSandboxCallback::addSiteLinksUsage                  800 ms 
       9.9%
        Scribunto_LuaSandboxCallback::callParserFunction                 720 ms 
       8.9%
        Scribunto_LuaSandboxCallback::getEntity                          320 ms 
       3.9%
        Scribunto_LuaSandboxCallback::addLabelUsage                      120 ms 
       1.5%
        Scribunto_LuaSandboxCallback::incrementStatsKey                  120 ms 
       1.5%
        Scribunto_LuaSandboxCallback::fullUrl                             80 ms 
       1.0%
        Scribunto_LuaSandboxCallback::gsub                                40 ms 
       0.5%
        init <Modul:CountryData/Geography>                                40 ms 
       0.5%
        ?                                                                 40 ms 
       0.5%
        [others] 
  
  I think there is a cache problem (see also T236749 
<https://phabricator.wikimedia.org/T236749>). And I assume that we exceeded a 
threshold with only minor additions. As already stated I think this bug is the 
same like T236485 <https://phabricator.wikimedia.org/T236485>. The caching bug 
was not really solved but only downplayed. Therefore it was only a question of 
time that this bug reoccured.
  
  There are only 166 entity calls (one per template) in the article of Halle 
(Saale) which is clearly less than the limit of 400. And I hope that we can use 
up to 250 Wikidata calls in future. Of course we use a lot of properties (in 
article mentioned about 50, maximally about 100) but we do not retrieve 
additional ones because of the tables mentioned. (I remember that you stated 
that Wikivoyage is the Wikimedia site which uses Wikidata in most extreme 
extent -- more than Commons. That is really true.)

TASK DETAIL
  https://phabricator.wikimedia.org/T265562

EMAIL PREFERENCES
  https://phabricator.wikimedia.org/settings/panel/emailpreferences/

To: RolandUnger
Cc: hoo, Lydia_Pintscher, RolandUnger, Aklapper, Akuckartz, darthmon_wmde, 
Nandana, lucamauri, Lahi, Gq86, Darkminds3113, GoranSMilovanovic, QZanden, 
LawExplorer, Vali.matei, _jensen, rosalieper, Scott_WUaS, Jonas, Volker_E, 
Wikidata-bugs, aude, GWicke, Dinoguy1000, Mbch331, Rxy, Jay8g
_______________________________________________
Wikidata-bugs mailing list
Wikidata-bugs@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikidata-bugs

Reply via email to