reportConstantRedefinition warn for constant declaration

[Vanerio]

Hi!

I like to keep all my module-level constants declared at the top of a file so it's easy to see the overall structure.
Sometimes these "constants" cannot be assigned immediately because their values depend on functions that are defined later, so I only declare their types at the top and assign them inside an initialize() function.

However, Pyright (in strict mode) reports a warning when I assign a value to an ALL_CAPS variable that was only declared but not defined.

Example:

# pyright: strict

A = 1.301030
B: float   # only a declaration, not a definition

def calculate(x: float) -> float:
    return x

def initialize() -> None:
    global B
    B = calculate(A)

initialize()

Pyright reports:

"B" is constant (because it is uppercase) and cannot be redefined (reportConstantRedefinition)

My questions are:

1. Is this behavior expected?

2. Was this always the behavior?
   I'm under the impression that older versions did not warn in this situation.

3. Is there a recommended way to declare (but not define) module-level variables that will be initialized later, while keeping them grouped with the constants?

Thank you very much!!

[erictraut]

Yes, this is working as expected. As far as I can remember, this behavior hasn't changed.

The technique you're using is fine for variables but not for constants. An all-caps symbol is assumed to be a constant. The recommended way is to use lowercase names for variables.

[Vanerio]

Thank you very much for the answer!

[Vanerio]

I have been thinking a little bit more about this. I’ve noticed some differences (please allow me the comparison):

Pyright:
Does not allow definition if a declaration is present.
Does warn for (supposed) redefinition of constant.

Mypy:
Explicitly requires declaration (for example, if you remove the line B: float in my code, a warning appears).
Does not warn for constant redefinition.

In my opinion, a middle ground would be more coherent:
Allow declarations (because they are not definitions, have no real effect on them, and help provide context).
Warn for constant redefinition.

This restriction can obscure the intended meaning and context behind the annotation.

I understand that definitions using CAPS are interpreted as new constants. That is expected and good.

But a declaration is still not a definition, even when written in CAPS, so perhaps it shouldn’t be treated the same way.

I’d be happy to hear your thoughts on this.
Thank you!