Improve error messages when using builtin cost functions

[HDembinski]

There are many ways to make mistakes and use the builtin cost functions incorrectly. The user has to provide the model and it is easy to return an array of incorrect size.

from iminuit import cost, Minuit
from numba_stats import norm
import numpy as np

n = [1, 2, 3]
edges = [0.1, 0.2, 0.3, 0.4]

def model(x, n, a, b):
    # oopsi, returned probability per bin instead of scaled cdf
    return n * np.diff(norm.cdf(x, a, b))
    # return n * norm.cdf(x, a, b)  # correct

c = cost.ExtendedBinnedNLL(n, edges, model)
c(1, 2, 3)

This produces a very confusing error message:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/var/folders/tl/pv6mt7z17tz0stm1fjfg01cc0000gn/T/ipykernel_28667/1963231075.py in <module>
     12 
     13 c = cost.ExtendedBinnedNLL(n, edges, model)
---> 14 c(1, 2, 3)

/usr/local/lib/python3.8/site-packages/iminuit/cost.py in __call__(self, *args)
    482         float
    483         """
--> 484         r = self._call(args)
    485         if self.verbose >= 1:
    486             print(args, "->", r)

/usr/local/lib/python3.8/site-packages/iminuit/cost.py in _call(self, args)
   1292         else:
   1293             n = self._masked
-> 1294         return poisson_chi2(n, mu)
   1295 
   1296 

/usr/local/lib/python3.8/site-packages/iminuit/cost.py in poisson_chi2(n, mu)
    345         n, mu = np.atleast_1d(n, mu)  # type:ignore
    346         if mu.dtype in (np.float32, np.float64):  # type:ignore
--> 347             return _poisson_chi2_nb(n, mu)
    348         # fallback to numpy for float128
    349         return _poisson_chi2_np(n, mu)

ValueError: unable to broadcast argument 1 to output array
File "/usr/local/lib/python3.8/site-packages/iminuit/cost.py", line 1, 

We need a better error message for this cost function and for others.

[amanmdesai]

Hi @HDembinski,

I would like to work on this issue
Could you please guide me through it?

Thanks
Aman

[HDembinski]

Ok, first check that you can reproduce this error message with the latest iminuit version. Then you need to setup your computor to develop iminuit, see CONTRIBUTING.md for details.

Once you are ready to develop, try to make sense of src/iminuit/cost.py, which contains all the cost functions. Try to identify the best place where a check can be made that the arrays have the correct expected size and types and if not, raise an informative ValueError at that location.

Note: the cost functions use a deep class hierarchy to avoid duplicating code. Every class that the user interacts with has a chain of base classes. This makes it a bit hard to understand what is happening where. I think that's the greatest challenge here, otherwise this is fairly simple to implement.

[amanmdesai]

It seems this issue did not close automatically after the merge of #863 (as i did not include the prefix "close/fix" in the message).

[HDembinski]

Yes, indeed, let's close it. Thanks again.