There's always something new to learn about managing time in software. I've been working a lot with timestamps recently, and wrote up some surprising-to-me things from that work.
@ehmatthes FYI there is a much more elegant way to find out how many of a given time unit there is in a timedelta, which is to divide it by another timedelta, like so:
```
>>> timedelta(days=3) / timedelta(hours=1)
72
```
In fact all the talk about normalization feels a bit misleading. The root of your original bug isn't really normalization so much as the fact that you are accessing a component when what you wanted was a view onto the entire duration.
@pganssle That was an interesting read! It's nice to know how not alone I was in misusing `.seconds`!
@pganssle Thanks for responding, I always appreciate your insights about these libraries.
I went into this work without an understanding of how timedeltas are implemented, and reading about normalization helped me think about them much more clearly. It also brings up a number of ideas that a lot of readers aren't aware of, but are ready to make sense of.
@ehmatthes It would be nice if this ends up implemented in ruff, since the majority of the time accessing one component of a timedelta is not something you want to do anyway: https://github.com/astral-sh/ruff/issues/13598