Why does MatLab has inputParse function but does not use it in its own functions

Why does MatLab has inputParse function but does not use it in its own functions, for example, in interp1 which accepts arguments in name-value style?

[Stephen Cobeldick]

• Because inputParse is slow: code is faster with custom input checking.
• Because many toolbox functions call a compiled function to check the inputs, which is specialized for the family of functions that use it.

[Rostislav Teryaev]

Okay, it makes sense. But why do they offer it to users?

[Stephen Cobeldick]

• Demand.
• Convenience.
• For many users speed is not critical.

Answer 1

[Jan]

The total time to solve a problem consists of:

T_total = T_design + T_programming + T_debugging + T_runtime

inputParser can reduce the programming time remarkably. It might waste many milliseconds, but if the function is called some hundred times only, this does not matter. Using an hour to write and test a variety of input arguments can be inefficient in comparison, even if the code consumes microseconds only. And if you have a bug in manual input parsing, which was not detected before delivering the software, the costs can be substantial.

The shorter the function and the more frequently it is called, the worse is the slowdown by inputParser. But even interp1 and polyfit need more time with checking the inputs than with the actual computations - without inputParser ...