TRIO115 false positive with with sleep(var) where var starts as 0

[mikenerone]

Ruff 0.2.1

The following snippet illustrates a TRIO115 false positive when sleep() is passed a variable whose initial value was 0, but changes later in a loop construct. Similar to #9934, this is already fixed in flake8-trio. This seems to be strongly indicating that recent fixes to flake8-trio need to be ported in general.

import trio

async def main() -> None:
    sleep = 0
    for _ in range(2):
        # ↓↓↓↓↓ TRIO115 [*] Use `trio.lowlevel.checkpoint()` instead of `trio.sleep(0)`
        await trio.sleep(sleep)
        sleep = 10

trio.run(main)

[charliermarsh]

Thanks. I don't think this has been "fixed" in flake8-trio -- rather, flake8-trio never flags with a non-constant, it only flags if you pass exactly 0. You can see the source here: https://github.com/Zac-HD/flake8-trio/blob/a652441fb3a9a92f14da3542bb2a0a1ba8f03f87/flake8_trio/visitors/visitors.py#L219.

[charliermarsh]

We could consider doing the same, or just check if the variable is ever re-bound to something else.

[mikenerone]

Thanks. I don't think this has been "fixed" in flake8-trio -- rather, flake8-trio never flags with a non-constant, it only flags if you pass exactly 0.

We could consider doing the same, or just check if the variable is ever re-bound to something else.

@charliermarsh Oh, I see - you're correct. Not flagging on a non-constant sounds reasonable to me. With a variable, even if not re-bound, it seems like the developer is signaling that it's considered changeable in at least some sense.

[charliermarsh]

Agreed.

[jakkdl]

flake8-trio dev here:
the "constant" in the source code is just how the AST labels literals, flake8-trio doesn't track the values of any variables (and has no straightforward way of doing so)

[robincaloudis]

As #10376 is merged, it looks like this issue could be closed.