Incompatibility of cppad_py.a_double with float in python

[zhengp0]

Here is a simple test program where the issue happens.

Test case:

import cppad_py
import numpy

# initialize dimensions
n_dep = 1
n_ind = 1

# declare independent variables
x  = numpy.ones(n_ind, dtype=float)
ax = cppad_py.independent(x)

# declare dependent variables (quadratic function)
ay = numpy.zeros(n_dep, dtype=cppad_py.a_double)
ay[0] = 0.5*ax[0]*ax[0]

# define corresponding function
af = cppad_py.a_fun(ax, ay)

# calculate the gradient
g  = af.jacobian(x)

Error report:

Traceback (most recent call last):
  File "test.py", line 14, in <module>
    ay[0] = 0.5*ax[0]*ax[0]
TypeError: unsupported operand type(s) for *: 'float' and 'a_double'

Thanks!

[bradbell]

This will take some time for me to get to and fix.
My current plan is to create a python type called a_float that wraps a_double and accomplishes this.

For now, you can use
ay[0] = a_double(0.5) * ax[0] * ax[0]

[bradbell]

I am getting the following error message with your test above:

(Pdb) 
TypeError: __init__() takes 2 positional arguments but 3 were given
> /home/bradbell/repo/cppad_py.git/junk.py(21)<module>()
-> af = cppad_py.a_fun(ax, ay)

[bradbell]

The problem is you are calling a_fun instead of d_fun; see
https://bradbell.github.io/cppad_py/doc/py_fun_ctor.htm
I need to get a better error message for this case.

There is a python __init__ function for the a_fun class in the lib/python/cppad_py/fun.py file. This is the function that the error message below is for:
TypeError: __init__() takes 2 positional arguments but 3 were given