Defined __repr__ for tf/ss [sf#8] about python-control

[murrayrm]

*Commented by anonymous on 2014-04-01 16:09 UTC
Is there anything I can do to help get these patches accepted? I would like to be able to share my work with others without directing them to download my private git repo. Here is my current notebook, it goes from first principles to a full LQR based lateral/longitudinal controller in one IPython notebook using control/scipy/numpy/sympy.

[murrayrm]

*Commented by anonymous on 2014-04-01 16:12 UTC
Note that the current aerodynamic model is a bit off, I haven't put reasonable coefficients in yet. I just wanted to get something working. At the angle of attack it is at, it should be stalling but I didn't model stall. I have done a CFD analysis of a UAV in XFLR5 and will put those coefficients in today.

[murrayrm]

Commented by scottlivingston on 2014-04-01 16:44 UTC
You are abusing __repr__ by setting it to __str__. This patch qualifies as an unnecessary hack for use in IPython. Cf.
https://docs.python.org/2.7/reference/datamodel.html#object.__repr__

[murrayrm]

*Commented by anonymous on 2014-04-01 16:59 UTC
I think setting repr to something would be useful. You could make it a call to the ctor string. The current address of the object that is default is useless. Also, can you take a look at the other patches. There are required changes there for my code to work. Like state-space array access.

[murrayrm]

Commented by scottlivingston on 2014-04-01 17:15 UTC
Anonymous: Are you James Goppert? Please log-in so that your username is archived in these comments.

[murrayrm]

Commented by jgoppert on 2014-04-01 17:36 UTC
Yes, sorry about that, was using my cell.

[murrayrm]

Commented by murrayrm on 2014-04-02 20:37 UTC
I'll try to take a look at the various patches and incorporate them this coming weekend. For this particular patch, my plan is to follow the python guidelines.

[murrayrm]

*Commented by anonymous on 2014-04-02 21:45 UTC
Thanks!

[murrayrm]

Commented by jgoppert on 2014-04-02 21:48 UTC
I'm currently working on packaging slycot for pypi. Kind of a headache, but I'm getting there.

[cwrowley]

Included in 71cfb32. But we should probably redefine __repr__ to give an expression that will recreate the object, rather than duplicating __str__, as mentioned above.

[slivingston]

I traced the integration of commit 71cfb32 through pull request #28 . Neither here (#16) nor there was the concern about abuse of __repr__ addressed. Or rather, it was decided as something that can be changed later. For comparison and a suggestion as to why this may matter,

from __future__ import print_function
import numpy as np

x = np.random.rand(3)
print(A.__str__())
print(A.__repr__())

whereas

from __future__ import print_function
import control

S = control.StateSpace([[-1.]], [[1.]], [[1.]], [[0.]])
print(S.__str__())
print(S.__repr__())

As an example of how we may create a __repr__ method that is consistent with the Python data model, consider the following (continuing the previous code snippet):

def example_repr(S):
    return "StateSpace("+repr(S.A)+", "+repr(S.B)+", "+repr(S.C)+", "+repr(S.D)+")"

print(example_repr(S))

This representation string is easily used in eval(), after ensuring the appropriate names are available.

from control import StateSpace
from numpy import matrix

S2 = eval(example_repr(S))

where S2 is an instance of StateSpace.

[murrayrm]

Changed milestone on this closed release to improve tracking.