corrupt titles since 2.0.4 upgrade

[anarcat]

there's something weird about the titles in the database...

third line is in chinese (i think?), 阿纳猫-阿纳猫 is not the title of that page.

i don't believe that was a problem before...

[anarcat]

also you'll notice that some titles are okay... but not all of them, which is even weirder.

[arp242]

This happens when people use Google Translate or similar tools to translate a page, which also translates the title and sends it to GoatCounter.

I think you updated from 1.x to 2.x? This includes quite a few database changes which make things faster and take up less disk space. Previously, every pageview would store the title:

created_at           | path  | title   | ..
2020-01-01 15:00:00  | /path | Hello!  | ..
2020-01-01 15:10:00  | /path | Hello!  | ..

That's obviously quite redundant and really adds up on disk space after you've got millions of pageviews, so in 2.x it's:

created_at           | path_id
2020-01-01 15:00:00  | 1
2020-01-01 15:10:00  | 1

And then the paths table stores the path and title once.

The way this works is that it will use the first title it sees (for new pages), and then it will update to a new title if it sees a new title ten times in a row (cached in memory, so e.g. a restart will clear the counter on this).

The problem here is that when updating from 1.x to 2.x GoatCounter doesn't really "know" which title is the correct one. One reason I changed this (aside from performance/space) is that there's now one canonical title. I looked a bit at getting the "most frequently used title" during the migration, but that was incredibly slow, so it just uses the last one (or first one? I forgot). It's unfortunate because sometimes it gets the wrong title, but it was the best I could figure out without having migrations for larger sites take days, and it should correct itself eventually.

There isn't a facility to edit the titles manually, but you can change it in the database if you want (paths.title).

[anarcat]

On 2022-02-01 02:06:46, Martin Tournoij wrote: [...]

It's unfortunate because sometimes it gets the wrong title, but it was the best I could figure out without having migrations for larger sites take days, and it should correct itself eventually. There isn't a facility to edit the titles manually, but you can change it in the database if you want (`paths.title`).

ooh, i see. so this is not a bug, it's a feature! :) nice to see it will fix itself too. i guess this can be closed then, thanks!

…

-- Be curious. Read widely. Try new things. I think a lot of what people call intelligence boils down to curiosity. - Aaron Swartz