Page MenuHomePhabricator

Compute (cumulative) count of 'new' uploaders per day
Closed, ResolvedPublic

Description

It would be helpful for monitoring the health of a competition instance to see how many uploaders joined the contest each day. (This is different from newbie counting)

Event Timeline

Ok, smth must be off, because it’s fairly unlikely that on https://tools.wmflabs.org/wikiloves/monuments/2017/Germany, 39 images could be uploaded on the first day by 0 joiners :p

The numbers given for the UK at https://tools.wmflabs.org/wikiloves/monuments/2018/United%20Kingdom don't match the actual daily uploads which we categorise as they happen in Commons cats Category:WLM-UK 2018 unfiltered 09-xx where xx is the day of the month. You can see the numbers in those cats here: https://commons.wikimedia.org/wiki/Category:Images_from_Wiki_Loves_Monuments_2018_in_the_United_Kingdom.

I suspect the tool may be incorrectly working to UTC rather than the local midnight-to-midnight times.

@MichaelMaggs The tool does not take local time into account (except for calculating the begin- and endtime), and can only be considered an approximation. This is why several countries have uploads on 31 August. It is not designed as the 'truth' but rather as a way to analyze and monitor how the competition is progressing. It may deal differently with deletions and several other edge cases.

The numbers given for the UK at https://tools.wmflabs.org/wikiloves/monuments/2018/United%20Kingdom don't match the actual daily uploads which we categorise as they happen in Commons cats Category:WLM-UK 2018 unfiltered 09-xx where xx is the day of the month. You can see the numbers in those cats here: https://commons.wikimedia.org/wiki/Category:Images_from_Wiki_Loves_Monuments_2018_in_the_United_Kingdom.

I suspect the tool may be incorrectly working to UTC rather than the local midnight-to-midnight times.

Sure, as @Effeietsanders says, this is by design. I’m happy to be proven otherwise but I really cannot think of any use case where it really matters that days follow local time − after all, midnight is a somehow arbitrary limit too (uploaders might have uploading sessions spanning 11PM to 1AM). Also, countries may be spanning several time zones too ;-)

If someone can present a very strong use case for having this data follow local time, I’d look into it, but otherwise this is a minefield I’d rather not venture into :-)

Going to resolve this, but will break out a task for the visualization component.