lines per commit

Stream: general

[Johan Commelin]

Can someone please generate a plot for the number of lines per commit to mathlib master? My hypothesis is that commits have been getting much smaller compared to two years ago (which is perfectly fine, and explained by the split-your-PRs policy).
I would just like to see the data, but I don't have time to dig this out of the logs myself. If someone could do it for me, that would be awesome!

[Yaël Dillies]

Oh yes, I have seen the monster PR from when mathlib was starting!

[Patrick Johnson]

You mean mathlib3, right? Here is the plot starting from commit deb16814 and ending with commit 4d56716d:

001.png

X-axis represents commits from older (left) to newer (right). Blue is how many files are changed, green and red are insertions and deletions. You can find raw data here.

[Patrick Johnson]

Here are only insertions and deletions, taking average of each 40 consecutive points:
002.png

[Johan Commelin]

aah, that shows the trend a lot better

[Johan Commelin]

Merci!

[Julian Berman]

In the programming world there are discussions and research papers trying to make suggestions about optimal lengths for pull requests

[Julian Berman]

They center around things like the attention span of people, how long it takes them to read lines of code on average, and number of defects they tend to uncover during that period

[Julian Berman]

I'd still be curious what alternate metrics or considerations there are for this world