Performance of the hours example

[Byron]

A first example has been added which is a pretty faithful clone of git-hours, a tool I could never actually run as node decided it won't build certain extensions or… what do I know, so there won't be a comparison to it here.

Instead, we race directly against git shortlog on the biggest repository I have, the linux kernel.

linux git:(master) ✗ time hours . master -p
Getting all commits…
Found 1002691 commits and extracted their data in 6.616651791s (151541 commits/s)
Getting all commit data…
Obtained 1002691 commits in 277.658583ms (3611237 commits/s)
total hours: 968161.81
total 8h days: 121020.23
total commits = 1002691
hours . master -p  8.18s user 0.99s system 121% cpu 7.570 total

As a fun stat, one would need more than 331 years of working every day for 8 hours in order to write the linux kernel, based on the estimation algorithm.

➜  linux git:(master) ✗ time git shortlog -s -n >/dev/null
git shortlog -s -n > /dev/null  6.53s user 0.21s system 99% cpu 6.753 total

Git shortlog does a little less work I believe and ends up being a little faster, delivering very impressive performance for sure. gitoxide can't parallelize the commit traversal where most of the time is spent, but does parallelize the data extraction part to get a 2.5x speedup.

[Byron]

Adding git-estimate (v1.1) for comparison, written in Golang.

/usr/bin/time -lp ~/Downloads/git-estimate-darwin-amd64 >/dev/null
real        51.01
user        43.45
sys         15.92

It went up to using about 4516MB of RAM, whereas hours peaks at about 995MB, whereas git shortlog peaks at 1172MB.