Documentation of aggregation methods would be nice

philip 2015-10-10 12:00:26 UTC #24

Quick straw poll: is anybody currently relying on the 'jolly' (unary) behaviour currently exhibited by 'binary'?

adamwolf 2015-10-10 13:37:16 UTC #25

Not me.

insti 2015-10-11 14:24:28 UTC #26

yes. (Apart from the zero handing, which I'd prefer to work differently.)

adamwolf 2015-10-11 14:48:01 UTC #27

Ha! I thought he was asking "who likes binary with how zero works today"?

I use binary on fiveish goals, and if non-zero datapoints didn't count, they'd work better.

Adam Wolf

insti 2015-10-11 15:57:04 UTC #28

oh. In which case, no.

Now I'm just confused.

philip 2015-10-11 18:55:14 UTC #29

I ought to leave the straw polls to @dreev, apparently.

Other than the eponymous @jolly, seems as though everyone would welcome a redefinition of 'binary' to count the existence of any non-zero datapoint as aggregating to '1', and '0' if there are no datapoints or all datapoints are themselves zero.

All of which may have added to the confusion, of course. Sounds like a quick and easy #UVI.

drtall 2015-10-11 19:01:12 UTC #30

No, no, don't make them strings! 1 != '1' (yes I'm just teasing)

insti 2015-10-11 19:19:20 UTC #31

Is there any evidence that @jolly wanted 0 to be 1?

philip 2015-10-11 20:53:22 UTC #32

No but yes. (:

The 'jolly' aggregation method was intended in order to enforce frequency of measurement-taking, counting days on which something was entered, regardless of the measured value.

His particular items of interest would be unlikely to result in a zero reading, but the request was specifically about datapoints entered, yes.

That aggregation method has a valid use, so let's not change it. That use is just not the one that most of us seem to have expected when reading the code...

jolly 2015-10-12 15:22:48 UTC #33

None whatsoever. I was using this for a goal where I wanted to enforce me entering data, while ignoring the value of the data entered. And for that goal, 0 was never a value I would have entered...but it might be for someone else.

brax 2015-11-16 05:55:03 UTC #34

What was the upshot of this? I don't see 'jolly' in the list. Has 'binary" been changed to have the 'nonzero' behavior? Because the 'nonzero' behavior is actually exactly ideal for my exercise goal (though several others work fine too).

philip 2015-11-16 11:30:36 UTC #35

The 'jolly' function has been deprecated, and removed from the selectable list. Though it still works if that's what had been set.

Looks as though we haven't yet made the 'binary' vs 'unary' distinction in the code.

That should be a quick #UVI though, @dreev. I looked at the code and the only thing stopping me from making the changes is that my dev env is broken and I wouldn't be able to test it.

dreev 2015-11-18 21:51:51 UTC #36

Done! Made nonzero a new aggday method and updated the wikified post above.

brax 2015-11-19 02:37:53 UTC #37

This is super! But, I don't see it in the list on my exercise goal. Is there an overnight refresh or something?

dreev 2015-11-20 00:26:21 UTC #38

Sorry about that! I deployed it for the back end (and thus working via the API) but I forgot to update the dropdown in the UI. Done now!

brax 2015-11-20 03:52:58 UTC #39

Awesome. Two of my goals are now sporting it : )

brax 2015-11-22 13:12:19 UTC #40

Aaaaaand nonzero aggmode breaks pessimistic presumptions on the Do Less goals, apparently. One my goals is still sporting it though : D

dreev 2015-11-24 22:47:36 UTC #41

Can you point us to the problematic one? (Can move this to support@beeminder.com if it makes sense to.)

brax 2015-11-25 04:36:55 UTC #42

Sure thing - I emailed the deets to the support line with your name in the subject.

drtall 2016-02-01 21:50:34 UTC #43

Let me take another stab? If there was some possible universe in which I could get the comments for my aggday, I would have a good use for it.

My new Beelint goal is trying to track the number of goals that are in violation each day, and currently uses min to not fault you for violations that come up during the day. Sort of like how Gmail Zero works.

The problem is that I might go from A,B to B,C and it looks like those are both size 2, but C is new and shouldn't be counted. I'd like to just count the problems that have been present ALL day. So what I'd really like is an aggday that looks like

lambda comments: len(set.intersection(*[set(c.split(',') if c else set()) for c in comments]))

If there's no hope I can print/read a state file but I was hoping to keep my thingy stateless.