optimization/converters/integer_to_binary.py: OverflowError: cannot convert float infinity to integer

[HassanNaseri]

Information

• Qiskit Aqua version: 0.7.3
• Python version: 3.8.2
• Operating system: macOS Catalina

What is the current behavior?

I get the following error when converting a MIP to QUBO:

  File "xxx/.pyenv/versions/3.8.2/lib/python3.8/site-packages/qiskit/optimization/converters/integer_to_binary.py", line 114, in _encode_var
    power = int(np.log2(var_range))

OverflowError: cannot convert float infinity to integer

Steps to reproduce the problem

The InequalityToEquality() converter creates slack variables with same upperbund and lowerbound (both zero). This results in _encode_var() function crashing at following line as var_range equals zero.
power = int(np.log2(var_range))

This can be reproduced by trying to encode a least or equal constraint with all positive coefficients in LHS and zero in RHS, e.g. x[1] + x[2] <= 0, which can naturally happen in automatically converting large problems. This can be safely encoded as x[1] + x[2] = 0 without slack variables.

What is the expected behavior?

The expected behaviour is not to create such slack variables at first place.

Suggested solutions

This should be fixed at two places.

• The InequalityToEquality() converter should not create slack variables if both upperbound and lowerbound are same.
• The IntegerToBinary() converter should check the var_range before trying to implement binary encoding.

[t-imamichi]

Thank you for reporting the error. We will add a check of var_range.