False "not accessed" positive

[memeplex]

Environment data

• Language Server version: 2021.3.3
• OS and version: MacOS Big Sur 11.2
• Python version (& distribution if applicable, e.g. Anaconda): 3.8.6

Expected behaviour

In the following code fragment:

        hess = np.eye(n + 1)
        res = sp.optimize.minimize(
            lambda w: 0.5 * ((w - weights) ** 2).sum(),
            weights,
            jac=lambda w: w - weights,
            hess=lambda w: hess,
            constraints=[sp.minimize.LinearConstraint(A, lb, ub)],
        )
        if not res.success:
            raise OptimizationError(res)
        weights = iter(res.x)
        c.weight = next(weights)
        for g in c.groups:
            g.weight = next(weights)

pylance is showing "hess" is not accessed for the. first line. The problem is related to the way I access the results of minimize: as an iterator on res.x. This seems to make pylance believe I'm not using the result at all.

Actual behaviour

Since I'm actually using the results of minimize and hess is indeed part of the input of minimize, the error is spurious and should not be shown.

[memeplex]

Indeed, removing:

        weights = iter(res.x)
        c.weight = next(weights)
        for g in c.groups:
            g.weight = next(weights)

suppresses the spurious report. If I also remove:

        if not res.success:
            raise OptimizationError(res)

the report is now for res, which is all right since res is not used anymore.

I think the problem has to do with the assignment to weights in weights = iter(res.x). Using weights = res.x or assigning to a new variable like weights_ = iter(res.x) both remove the warning. It's really weird.

[memeplex]

Here is a reproducible example:

import numpy as np
import scipy as sp


def f():
    weights = np.ones(10)
    hess = np.eye(10)
    res = sp.optimize.minimize(
        lambda w: 0.5 * ((w - weights) ** 2).sum(),
        weights,
        jac=lambda w: w - weights,
        hess=lambda w: hess,
    )
    weights = iter(res.x)
    return weights

[erictraut]

Thanks for the bug report. This will be fixed in the next release.

[jakebailey]

This issue has been fixed in version 2021.3.4, which we've just released. You can find the changelog here: https://github.com/microsoft/pylance-release/blob/main/CHANGELOG.md#202134-31-march-2021

[memeplex]

Thanks!