Odd pop expected at least 1 argument, got 0 error

[wshanks]

I just noticed that in one test of another project I got the following error:

      File "D:\a\qiskit-experiments\qiskit-experiments\qiskit_experiments\curve_analysis\utils.py", line 101, in analysis_result_to_repr
    if result.value.std_dev is not None and np.isfinite(result.value.std_dev):
       ^^^^^^^^^^^^^^^^^^^^

      File "D:\a\qiskit-experiments\qiskit-experiments\.tox\qiskit-main\Lib\site-packages\uncertainties\core.py", line 511, in std_dev
    return float(sqrt(sum(delta**2 for delta in self.error_components().values())))
                                                ^^^^^^^^^^^^^^^^^^^^^^^

      File "D:\a\qiskit-experiments\qiskit-experiments\.tox\qiskit-main\Lib\site-packages\uncertainties\core.py", line 476, in error_components
    for variable, derivative in self.derivatives.items():
                                ^^^^^^^^^^^^^^^^

      File "D:\a\qiskit-experiments\qiskit-experiments\.tox\qiskit-main\Lib\site-packages\uncertainties\core.py", line 450, in derivatives
    self._linear_part.expand()

      File "D:\a\qiskit-experiments\qiskit-experiments\.tox\qiskit-main\Lib\site-packages\uncertainties\core.py", line 311, in expand
    (main_factor, main_expr) = self.linear_combo.pop()
                               ^^^^^^^^^^^^^^^^^^^^^^^

    TypeError: pop expected at least 1 argument, got 0

That was from this test run in the other project. The result is not reproducible so it is hard to say what caused it right now, but I wanted to report it in case it comes up in other places. The surrounding code in expand and derivatives has not changed recently as far as I can tell. linear_combo there can either be a dict or a list. derivatives checks expanded which checks if linear_combo is a list and then calls expand when it is and expand converts the list to a dict. The TypeError comes from expand calling pop() on a dict instead of a list. I haven't been able to see how the code could end up with linear_combo already a dict and expand getting called.

[jagerber48]

Interesting. After a quick look this change pops out at me:

a2cb45e#diff-d2e4fa0edf80ad20b9f3c6a2f2d3cfffebccc092c73570e50c3e278bbfdbb6eaR544-R545

line 544, 545 in core.py

[wshanks]

That is a more recent change that most of the related code. I still don't see how it could lead to the exception because that was supposed to be just moving where the conversion to LinearCombination happens (inside AffineScalarFunc.__init__ vs before passing to AffineScalarFunc.__init__) but it is something to think about.