ZeroDivisionError: division by zero

Python ZeroDivisionError has an important interaction with floating-point arithmetic that surprises developers from other languages. Native Python: 1 / 0 and 1.0 / 0.0 both raise ZeroDivisionError — Python does not return float('inf') as IEEE 754 arithmetic technically allows. NumPy behaves differently: numpy.float64(1.0) / numpy.float64(0.0) returns inf and emits a RuntimeWarning: divide by zero encountered but does NOT raise an exception. This inconsistency between native Python floats and NumPy floats is a frequent source of confusion in mixed code. The float('nan') value is another edge case. NaN (Not a Number) is not equal to anything, including itself — float('nan') == float('nan') is False. This means checking if divisor == float('nan') always evaluates False. Use math.isnan(divisor) instead. Division by NaN does not raise ZeroDivisionError — it propagates NaN into the result. For data pipelines, the safe pattern is: result = (numerator / denominator) if denominator != 0 else default_value, combined with math.isnan(denominator) and math.isinf(denominator) checks for full float validation. NumPy's numpy.errstate(divide='raise') context manager converts NumPy's division-by-zero warning into a Python exception, matching native Python behavior when you want consistent error handling across both types.