Use Literal type in io and _compression

[sobolevn]

No description provided.

[JelleZijlstra]

Just curious, what is your motivation for all these Literal changes? Is there a concrete use case this unlocks?

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[sobolevn]

Just curious, what is your motivation for all these Literal changes? Is there a concrete use case this unlocks?

Yes, I am working on better Literal type support in mypy. The first issue is already there: python/mypy#11992

[JelleZijlstra]

I saw that but I wasn't clear on the concrete use case there either.

[Akuli]

I can see myself staring at some code that works incorrectly with a chunk of data that is 10000 bytes long but fine with 5000 bytes, wondering what io.DEFAULT_BUFFER_SIZE is, and bringing my mouse over it to see a tooltip describing it. It would be convenient to just see the actual value.

[sobolevn]

@JelleZijlstra I think that this is how typing should be 🙂

It is just happens right now that we don't use Literal that much in Python, mostly because of the legacy. But, I think that at some point we can reach this behavior in mypy:

x = 1
reveal_type(x) # Literal[1]

I think that what pyright does at the moment.
And this will make literal math a very handy feature!

[a-reich]

There has been discussion of using literal arithmetic to type array operations (for the shape parameter).