Physical-type correctness in scientific Python (1807.07643v3)

Abstract: The representation of units and dimensions in informatics systems is barely codified and often ignored. For instance, the major languages used in scientific computing (Fortran, C and Python), have no type for dimension or unit, and so physical quantities are represented in a program by variables of type real, resulting in the possibility of unit or dimensional errors. In view of this danger, many authors have proposed language schemes for unit-checking and conversion. However, since many physical quantities have the same units, it is possible for a block of code to be unit-compatible, but still physically meaningless. We demonstrate the limitations of three Python unit-libraries and present a justification and method for checking kind-of-quantity.