Update activity score calculation algorithm

[3cpt]

This PR contains changes to the Repository Activity Score algorithm.

• replace watchers_count by subscribers_count - because stars are the same as watchers, so I just kept stargazers and replaced watchers by subscribers (read more about this here).
• fix slice number - I'm not a JS pro, but seems that we just want to slice the 13 (~3 months)
• simplify check - some validations can be made by just add the ?
• repo JSON access standardization - JS versatility allow using json['prop'] or json.prop. I picked one for better reading.

[spier]

Thanks for the contribution @3cpt!

Very cool that you are reviewing the pseudo code here.
I am not even sure if it was meant to represent JS but you are probably right.

You mentioned elsewhere that you are planning to add a score calculation function to one of the reference implementations of the crawlers, right?

Great catch with the watchers/stargazers/subscribers topic. I need to learn more about this. I am a bit confused after reading the article that you referenced above. Maybe I just got to read it again. :)

I left one inline comment for you, to see if we could improve readability of the score calculation even further.

Not sure how we could go about validating the pseudo code here.

@Michadelic you will find this one interesting. Any feedback and ideas what we could use to validation the algorithm?

[3cpt]

You mentioned elsewhere that you are planning to add a score calculation function to one of the reference implementations of the crawlers, right?

Yep. That's why I start analyzing the algorithm, I'm implementing a Python version to integrate in the zkoppert/innersource-crawler. Should we add a section in the pattern to implementations of the algorithm?

Great catch with the watchers/stargazers/subscribers topic. I need to learn more about this. I am a bit confused after reading the article that you referenced above. Maybe I just got to read it again. :)

As far as I understood, they changed watchers to stars and that created a small problem (More and Other)

Not sure how we could go about validating the pseudo code here.

Agree. What about use the mock data from the planet's repo.json and generate the score instead of mock it too?

[spier]

@3cpt just confirming that I didn't forget about your PR. Would just like to wait for @Michadelic for feedback.
(it might still be holiday season in some places so let's give him some time :)).

[3cpt]

@spier No worries. I replicated the same code in Python and C# to validate the consistency of the numbers.
So if someone uses the algorithm to gather the score, please, share it! Real data is always good in this case.

[Michadelic]

Hi @3cpt @spier, indeed there is holiday season over here so please excuse my late response.
Thank you very much for your great findings, will take a look in detail now and provide feedback in the code directly.

I took this code directly from our running crawler on GitHub enterprise, so it is a working JavaScript implemention, but for sure it can be further tweaked and adjusted. If you spot more issues just let me know.

Michael

[Michadelic]

Great simplifications, just the slice change should be reverted (see my comment)

[spier]

@Michadelic would you mind reviewing/approving this PR once it is in a state that is ready to be merged?
I don't feel knowledgeable enough myself about the activity score to do this myself. Cheers.

[3cpt]

During my last check I verified that the + 1 in the L64 and L70 (in the previous version) lead the result of the multiplier to > 1, when the rule in the comment says that should be between 0...1.
So, if, in the worst case scenario we get 0, 0 / x = 0, if it is x / x = 1.

Please, your feedback is welcome, and I could be doing my math wrongly.

[Michadelic]

please keep the "1+" as explained in the comment, other changes for readability are appreciated 👍

[spier]

@Michadelic would you mind reviewing the latest status of this PR again, and then approving the PR when everything looks good?

To me it looks like @3cpt implemented the change that you asked for (thanks for that!) for but I am not 100% sure so I figured I rather ask you for confirmation.

[Michadelic]

Thank you for the improvements, looks good now!

[Michadelic]

@spier PR can be merged now, looking good

[spier]

Thank you for this contribution @3cpt !!! Much appreciated fix.

Also we are looking forward to see your reference implementation.
Once that is live, then we could also link to that from this pattern.