Model functions with keyword-only arguments

[schtandard]

Keyword-only arguments to model functions do not work.

import numpy as np
import lmfit

def foo(x, *, a=1, b=2, kind='linear'):
    if kind == 'linear':
        return a + b * x
    elif kind == 'exponential':
        return a * np.exp(b * x)

mdl = lmfit.Model(foo)
# 'a' and 'b' are not parameters.
print(mdl.param_names)

# 'a' and 'b' are not varied.
# 'kind' is ignored.
x = np.linspace(-10, 10, 21)
y = -2 * np.exp(x / 2)
fitres = mdl.fit(y, x=x, kind='exponential')
fitres.plot_fit()

# This fails.
mdl = lmfit.Model(foo, param_names=['a', 'b'])

Is this by design? The documentation states

Keyword arguments to the model function that have numerical values will also be marked as being Parameters, and the supplied numerical value will be used as the default value for that Parameter.

which leads me to expect that keyword-only arguments should work.

[newville]

@schtandard Model functions do not support the bare * (or '/') parameters.

def foo(x, a=1, b=2, kind='linear'):
    if kind == 'linear':
        return a + b * x
    elif kind == 'exponential':
        return a * np.exp(b * x)

mdl = lmfit.Model(foo)

will work.

[schtandard]

Would you accept a feature request for changing this? I have functions with keyword-only arguments from other contexts that I would like to use as model functions and this forces me to manually make a copy of the function changing only the signature.

[newville]

Yes. Sorry for not saying that earlier. I think the "parse arguments of model function" dates from before the bare * or / and we just never got around to supporting that. Raising this an Issue would be helpful (with your example as a test case we should be able to pass). A PR would be great too.