I'm always a bit embarassed when I look at my commit history, for personal projects but also for projects at work. The simple commit that just fixes a bug or adds a feature in a few lines is very rare. It's often that a bug could be fixed on a single line, but this "actual" issue is drowned in a sea of read and green in the commit diff. Normally I cannot resist the urge to improve something. Sometimes it's a structural improvement. Often it's only cosmetic changes or moving stuff around that might make things clearer or not, I'm not always sure. At least my mind is working on the problem.
When I look at commits from some other devs that often do these localized fixes, it almost always seems like the codebase deteriotes a little bit on a global scale for example because some knowledge is duplicated from some other location 500 lines away.
I think I lean on the side of focusing on the end product (the tip of the development branch) at the cost of having not the most beautiful history ever. I'm also aware that my style of developing makes it hard to develop in large teams / a product with many release lines. For my small projects that have very little collaboration, my approach seems to work, though.
Maybe there's only two ways to have a nice history: Have a shitty end product, or be a domain expert. The latter seems unrealistic to me, because working on something you already know how to do is not very satisfying.
When I look at commits from some other devs that often do these localized fixes, it almost always seems like the codebase deteriotes a little bit on a global scale for example because some knowledge is duplicated from some other location 500 lines away.
I think I lean on the side of focusing on the end product (the tip of the development branch) at the cost of having not the most beautiful history ever. I'm also aware that my style of developing makes it hard to develop in large teams / a product with many release lines. For my small projects that have very little collaboration, my approach seems to work, though.
Maybe there's only two ways to have a nice history: Have a shitty end product, or be a domain expert. The latter seems unrealistic to me, because working on something you already know how to do is not very satisfying.