Take working days into account when calculating average

[harmenjanssen]

Problem: currently, doing a little work on a Sunday will throw off your average, whereas it should actually beef up your average.
Proposal: do not use total days to calculate the average, but only use the "official" working days. Example: working 8 hours for 4 days and doing a 1 hour job on Sunday should show an average of 8,25 hours instead of 6,6 hours.

I found that the following API contains a person's working days: https://api.productive.io/api/v2/timesheet_reports?filter%5Bafter%5D=2021-06-21&filter%5Bbefore%5D=2021-06-27&filter%5Bperson_id%5D=41757&page=1&per_page=200

Example output:

"attributes": {
    "week": "2021/W25",
    "monday_time": 579,
    "monday_capacity": 480,
    "monday_available": 450,
    "tuesday_time": 488,
    "tuesday_capacity": 480,
    "tuesday_available": 450,
    "wednesday_time": 394,
    "wednesday_capacity": 480,
    "wednesday_available": 450,
    "thursday_time": 452,
    "thursday_capacity": 480,
    "thursday_available": 450,
    "friday_time": 11,
    "friday_capacity": 0,
    "friday_available": 0,
    "saturday_time": 0,
    "saturday_capacity": 0,
    "saturday_available": 0,
    "sunday_time": 0,
    "sunday_capacity": 0,
    "sunday_available": 0
}

I think the _capacity or _available keys are significant here, although I don't understand the difference.
But either of them shows that I am not working on Friday for instance. I think we can use that to gather a person's working days from the given date and come up with the correct average.

I would work on this myself but I don't have local development figured out yet. (see other issue)

[martijngastkemper]

That would be a very nice improvement. Only downside of an extra API call is time. Slash commands in Slack require a response within 3 seconds. So when adding this a limit on the amount of days could solve it. Probably 150 days (200 minus non working days), because 200 days are fetched in one API call to time_reports.

Slack offers a way to have more time, but that requires more work. Immediately send a response back and invoke another Lambda function (something like this). The payload send by Slack contains a response_url to send messages to a channel for 30 minutes: Slack docs.

[harmenjanssen]

A limit of 200 days doesn't sound unreasonable. 🙂
Or make it a year, that's a nice period. If we move some API calls to parallel operations it might work!