Anonymous function versus File in matlabfunction

I am creating several large functions from symbolic expressions (jacobians used in optimization).

I've noticed that it takes some time to create the file output using

matlabFunction(jacob,'File','jacobianFunctinName','Vars',{listOfSymbolicVars})

Using an anonymous function, it is MUCH faster:

jacfun=matlabFunction(jacob,'Vars',{listOfSymbolicVars})

Is there a reason for this (profiling the code points out matlabWrite)? I know that anonymous functions were slower in past releases; should I pay the price up front to make the file, or just use the anonymous function?

Thanks,

Brad

Answer 1

[Walter Roberson]

https://www.mathworks.com/help/symbolic/matlabfunction.html#input_argument_namevalue_Optimize

"By default, matlabFunction with the File argument generates a file containing optimized code. Code optimization means that intermediate variables are used to simplify or speed up the code. MATLAB generates intermediate variables as a lowercase letter t followed by an automatically generated number, for example t32.

matlabFunction without the File argument (or with a file path specified by an empty character vector) creates a function handle. In this case, the code is not optimized. If you try to enforce code optimization by setting Optimize to true, then matlabFunction throws an error."