Calculating the Truck-Factor of git Applications

[GeorgLink]

The CII Census refers to the truck factor tool for git. - Maybe this can help us.

[howderek]

It looks like this lab also made http://gittrends.io/ which looks really great. I'll definitely check out their approach

[howderek]

Here is the paper related to their approach

[howderek]

After reading the paper, I found a couple of issues and a couple of good insights!

Issues:

• This approach doesn't take time into account, so core developers who have already left are still counted in the bus factor
• This approach requires manual review of the repositories to filter out files that should not be looked at when measuring how much a developer contributed

Insights:

• They mentioned the difficulty in measuring the bus factor of projects like Linux and Homebrew, who have massive numbers of developers but not as many core developers because they are extensible projects
• Libraries added to a project should not be considered contributions (I maybe disagree, I think that large changes such as that are indicative of being a core developer)
• Their survey suggested that documentation and tests should influence the bus factor

[GeorgLink]

Those are great insights.

…

On Mar 3, 2017 09:50, "Derek Howard" ***@***.***> wrote: After reading the paper, I found a couple of issues and a couple of good insights! Issues: - This approach doesn't take time into account, so core developers who have already left are still counted in the bus factor - This approach requires manual review of the repositories to filter out files that should not Insights: - They mentioned the difficulty in measuring the bus factor of projects like Linux and Homebrew, who have massive numbers of developers but not as many core developers because they are extensible projects - Libraries added to a project should not be considered contributions (I maybe disagree, I think that large changes such as that are indicative of being a core developer) - Their survey suggested that documentation and tests should influence the bus factor — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#27 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIc5RdLEivDG3m7t9rNx83r7mk0KS_KWks5riDbRgaJpZM4MSXS6> .

[GeorgLink]

I had requested gittrends.io analysis of the HealthIndicators rep: http://gittrends.io/#/repos/OSSHealth/HealthIndicators

[ChristianCme]

@sgoggins @howderek and I discussed truck factor and we determined that we would use the total lines of code as part of the calculation. We originally were going to get this value from the GitHub API, however, we cannot get lines of code from the API, but I can get number of additions and deletions in each file by each user. Would that still be useful for this metric?

[GeorgLink]

We are downloading the repository for other metrics. Can that be combined by calculating lines of code when we look at the repository and have that metric ready for the truck factor? Lines of code, while not a health or sustainability metric, might be interesting as it changes over time.

[sgoggins]

Leveraging some of my prior work and what we know about truck factor, we imagined the following inputs having some logic for reasonable inclusion in a comprehensive metric; or category of metrics.

I think additions and deletions can be helpful, as can total lines of code. They can also be somewhat misleading when repositories are refactored.

Our notes from the discussion are below.

[abuhman]

I have added a folder with files for example of getting number of lines in a repo and percentage of a repo written by each user using git blame.

[germonprez]

Thanks!

…

On Mon, Apr 3, 2017 at 4:15 PM, abuhman ***@***.***> wrote: I have added a folder with files for example of getting number of lines in a repo and percentage of a repo written by each user using git blame. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#27 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAoDUJU8BmE62av2_M8KcbBd75fpEf4sks5rsWGFgaJpZM4MSXS6> .

-- Mutual of Omaha Associate Professor Information Systems College of Information Science & Technology University of Nebraska Omaha http://faculty.ist.unomaha.edu/mgermonprez/cv.txt

[sgoggins]

we have a "bus factor" calculation that can be incorporated into the end of year release, along with documentation.

[GeorgLink]

we have a "bus factor" calculation that can be incorporated into the end of year release, along with documentation.

Great! I look forward to it.

[sgoggins]

We are pushing this because we're prioritizing new visualization types to finish the cold fusion release.