Activity analytics

[Minoru]

This is feature two of the three that @ForNeVeR originally envisioned.

Gestalt should monitor commits to the repositories belonging to the organization and report per-user and per-repo statistics (probably with weekly and monthly averages).

[ForNeVeR]

Well, I don't like this overall [ { "type": "user_stats" }, { "type": "repo_stats" } ] structure. It was useful for streaming data such as news, but it will introduce more problems for any unsorted data. I'd suggest the following:

{ "repos": [ { "name": "world-domination",
               "url": "https://github.com/codingteam/world-domination",
               "day_average": 15,
               "week_average": 1050,
               "month_average": 100500 } ],
  "users": [ { "name": "Minoru", 
               "url": "https://github.com/Minoru",
               "day_average": 15,
               "week_average": 1050,
               "month_average": 100500 } ],
}

And maybe we will want to introduce separate URLs for every user and repo later (if we'll add more time-expensive metrics and/or if the existing metrics suddenly become expensive).

Also I think there was a problem with "repo-url" key - you seem to be using underscores in keys, let's stick with them. (And I've fixed that in proposed format by replacing it with "url" anyway.)

[Minoru]

Yes, using the array was a bad idea; dictionary is better.

Your proposed format seems to be missing the point of the feature, though — we want something like "average number of commits over some periods of time", and it's nowhere to be found in your dicts.

repo-url was a typo, thank you. Fixed.

It also occurred to me that we should specify how exactly the commits are counted. With DVCS it's possible for a developer to code away for a week without pushing anything, and then suddenly we have all those commits pushed to all those repositories. How should we handle this? I see two approaches:

• use pushes' dates. That's simple enough: the user above will have 0 average for a week but a big number for today;
• use commits' dates. That's slighly harder (we need to crawl the commits and extract the dates) and might be counterintuitive to less adept DVCS users (people who weren't even in the top 10 suddenly jumping to top places).

Even though, I'd vote for the second approach as more "fair".

[ForNeVeR]

I just haven't shown some keys from the dictionaries. I've updated my examples to show them.

Regarding the commit counting - I'd say we don't care at this moment. We can use only data provided by Github API and if it provides both kinds of dates - that's fine; if not - that's fine too.

[Minoru]

I've replaced the format in the draft with the one you suggested. Thanks!

I also added a sentence clarifying how we count commits.

[Minoru]

Closing the issue due to the stabilization of version 1 of the API.