Fix callable detection to support functools.partial and all callable objects

[ssam18]

Fixes #2008

Problem

The find_max_global and find_min_global functions failed with an AttributeError when used with functools.partial objects and other callable objects that don't have a __code__ attribute.

Example of the issue:

from functools import partial
import dlib

def f(arg0):
    return 0

# This would fail with: AttributeError: 'functools.partial' object has no attribute '__code__'
dlib.find_max_global(partial(f, 2), [0.], [1.], 100)

Solution

Changed the num_function_arguments function to use Python's inspect.signature() which properly works with all callable objects:

• Regular functions
• Lambda functions
• functools.partial objects
• Class methods
• Any object with a __call__ method

The implementation includes a fallback to the original __code__ method for backward compatibility if inspect.signature() fails for any reason.

Testing

I've included a test script (test_callable_fix.py) that verifies the fix works with:

• functools.partial objects
• Regular functions
• Lambda functions

Let me know if you'd like me to add any additional tests or make any changes

[ssam18]

I ran into this issue while using dlib for optimization in one of my projects. I was trying to use functools.partial to partially apply some parameters to my objective function before passing it to find_max_global, and kept getting the AttributeError about the missing __code__ attribute.

After digging into the source code, I realized the issue was with how callable detection was implemented. Since inspect.signature() is the standard way to introspect callable objects in Python and works with all types of callables, I figured this would be a good fix that maintains backward compatibility.

Hope this helps others who might run into the same issue

[davisking]

Nice, thanks for the PR :)