[Mediawiki-l] Merging databases and 'special:statistics' anomaly

[Charles Collis]

Hi,

I have a slight anomaly after merging two wikis in that the statistics
appear not to change after the merge. I have run various scripts in the
maintenance folder including rebuildall.php, updateArticleCount.php,
clear_stats.php and updateSpecialPages.php.

The combined wiki seemed to inherit identical stats from the wiki that acted
as a 'host' during the merge. Before the merge the 'host' claimed to have
3746 total pages and 407 legit pages. The wiki that got folded in claimed
7668 total and 616 legit. The combined wiki claimed 3746 total and 407 legit
(same as the 'host').

The only script that appeared to make a difference to those figures was
updateArticleCount.php. This bumped up the legit article count of the
combined wiki from 407 to 871, but oddly the total no. of pages stayed
resolutely at 3746.

I'll briefly mention how I did the merge in case it is relevant:

*Wikis being merged were old (1.5.2) so upgraded them individually to 1.9.3.
All seemed fine.
*Used dumpBackup.php to extract data from the wiki to be folded into the
'host' wiki
*Used importDump.php to import this data into 'host' wiki
*Ran rebuildall.php
*Copied images from merged wiki into image folder of 'host' wiki, then ran '
rebuildImages.php --missing' to recover image information in tables (as
image data in tables don't seem to transfer in dumpBackup / importDump
process).
*Ran rebuildall again as well and other scripts mentioned in opening
paragraph.

I certainly didn't expect the total pages to be the sum of the two wikis as
I there are a lot of shared pages both in the Mediawiki namespace and there
were some duplicates in the main namespace too.

All the page titles seem to list OK in the 'all articles' special page and
various pages I've manually visited from each contributing wiki seem to be
exist in the combined wiki which is heartening. I am just concerned that the
'total pages' stubbornly not reporting correctly may indicate data integrity
problems, and feeI slightly reluctant to make this combined wiki
operational, in case there are issues later on.

Any suggestions to get the statistics reporting correctly gratefully
received.

Regards,
Charles.