That makes sense to me. Thank you! :)
CCing @daniel for sanity check

Makes sense, and should be fine.

However, purging all these pages is quite expensive. We could calculate the number and write it to page_props directly. That's a little bit hackish, but a lot faster, and not too horrible. It does needs more code to be written, but that shouldn't take more than a day or two. Worth it? I don't know.

In T163475#3208210, @daniel wrote:

Makes sense, and should be fine.

However, purging all these pages is quite expensive. We could calculate the number and write it to page_props directly. That's a little bit hackish, but a lot faster, and not too horrible. It does needs more code to be written, but that shouldn't take more than a day or two. Worth it? I don't know.

Is it worth the effort? Anyone available to actually do this? Sounds nice, but if it means waiting for many months because this has no priority, I rather just run a purge bot.

Yeah in that case let's just run the bot.

@Multichill You can go ahead with the bot run. The feature is deployed.

Given codfw database being under pressure right now, I suggest starting the process tomorrow. Especially if it's going to be intense purging. If the speed is low or you are not doing it in peak time, I'd say go ahead.

I haven't done anything yet, but quite a few pages already have the page_prop:

MariaDB [wikidatawiki_p]> SELECT COUNT(*) FROM page_props WHERE pp_propname='wb-identifiers' LIMIT 1;
+----------+

+----------+

+----------+
1 row in set (0.66 sec)

Some heavy job running that is updating a lot of pages? I'll take it easy with the run.

Running https://github.com/multichill/toollabs/blob/master/bot/wikidata/page_props_purge.py now.

The bot is slow purging the items. We're now around the 7 million done. No load issues or any other problems detected. I could speed up, but right now it's stable and I don't want to risk any issues.