A Glob I Would Not Have Written
sortbench is a repo I made in March to time three sorting approaches against each other. I got my answer in an afternoon and never opened it again. It has eleven commits. The eleventh is from last Tuesday.
The message is chore: tidy. Lowercase, conventional prefix, no body. That is how I write a commit like that.
The diff touches three files. It deletes out.txt and out2.txt, benchmark dumps I committed by accident on the first day and never cleaned up, and it adds one line to .gitignore. Nothing else. The tree is cleaner afterward and nothing about the benchmark changed.
The author is me. Name and email exactly as I have them configured. Author date and committer date match to the second, which is what a commit typed by hand at a terminal looks like and not what most tooling produces.
So I did it half-asleep. I have been at the keyboard at bad hours more often than I would like to defend, and a two-file cleanup in a dead repo is exactly the sort of thing I would do while waiting for something else to finish and then fail to remember. My shell history has no git commit that night, but I run several terminals and history does not reliably flush from all of them, so that is weak evidence at best.
The part I keep going back to is the .gitignore line.
I would have written out.txt and out2.txt. Those are the two files in front of me. They are the problem, so I delete them and I ignore them by name. What is in the file is out*.txt. That is a statement about files that do not exist yet, made by someone thinking about the next run instead of this one.
I noticed because of the next run. Three days ago I opened the repo to check a number, ran the benchmark once, and it wrote out3.txt. git status came back clean. The file had already been handled by a decision made a week before it existed.
The commit timestamp is 03:26. I have looked at that number enough lately that I no longer find it interesting, which is probably a mistake, but it is where I am.
I thought about reverting it. It would take a second and it would settle whether I actually care. I have not, because the revert puts out.txt and out2.txt back into a repo that is better without them, and I would be undoing correct work to prove a point about who did it.
sortbench has eleven commits and a clean tree.