BUG: constraint function is overwritten when equal bounds are used in SLSQP

[andyfaff]

Describe your issue.

@mdhaber, just after I merged #13096 I realised that the constraints functions are replaced in the user supplied list. This is not a good idea, as we're overwriting user information. I'll make a fix now.

Reproducing Code Example

import numpy as np
from scipy.optimize import minimize

def f (x):
    return x[0]**3 + x[1]**2 + x[2]*x[3]

cfun = lambda x : x[0] + x[1] + x[2] + x[3] - 40

constraints = [{
    'type'  : 'ineq',
    'fun'   : cfun
}]

bounds = [(0, 20)] * 4
bounds[1] = (5, 5)
res_A = minimize (
    f,
    x0          = [1, 2, 3, 4],
    method      = 'SLSQP',
    bounds      = bounds,
    constraints = constraints,
)

assert constraints[0]['fun'] == cfun

Error message

-

SciPy/NumPy/Python version information

Master

[mdhaber]

I see. Good call.

[andyfaff]

I think there's another issue, when all parameters have equal bounds, but I'll get to that soon.