Antony–22 raised a question about accounting for "new articles" that are moved from other namespaces to article space. For the purposes of counting total articles, I'm guessing that these are properly accounted for as deltas to the total, even if they're not considered new articles for the purpose of NPP under Special:NewPages. Is that correct?
Thanks, Pine
Recent conversations on this mailing list are leading me to a new definition of "fuzzy math". (: Thanks Nemo.
Pine
On Fri, Oct 30, 2015 at 12:49 AM, Federico Leva (Nemo) nemowiki@gmail.com wrote:
https://meta.wikimedia.org/wiki/Article_counts_revisited
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
Le 30/10/2015 06:23, Pine W a écrit :
Antony–22 raised a question about accounting for "new articles" that are moved from other namespaces to article space. For the purposes of counting total articles, I'm guessing that these are properly accounted for as deltas to the total, even if they're not considered new articles for the purpose of NPP under Special:NewPages. Is that correct?
The article count is vastly inaccurate. Mostly because operations on pages are not atomic with the update count operation and, all such operations are not entirely determined nor do they have an associated required hit count action.
I can't remember the code, but I am pretty sure we had a few race conditions. An example could be:
count: 100 create article delete article concurrently delete article action updates counter (100 - 1 => 99) create article updates count based on old reference (100 + 1 => 101)
It should be a zero sum, but since the creation still referred to the old count, the counter ends up gaining an article.
Don't quote me. The code has probably changed a lot and might not be subject to race conditions any more.
I did ran updateArticleCounts.php on purpose on frwiki just before it hits a milestone (maybe 100k). That resulted either in a regression (back to 98k) or a bump (hey we already are at 102k).
That confused a lot of people and we spent a good chunk of time to determine which article actually has hit the milestone (ended up picking a nice one created in that time window).
updateArticleCount.php runs automatically on the 21th of each month https://phabricator.wikimedia.org/T68867