Incorrect values for .a and .b (min and max) for scipy.stats.uniform with non-default args

My issue is about the definition of the attributes a and b (corresponding to min and max) for the scipy.stats.uniform distribution. These appear to always return 0 and 1 respectively, corresponding to the default values, regardless of the arguments.

Reproducing code example:

import scipy.stats as ss
x = ss.uniform(loc=5, scale=5)
print(x.a)  # 0.0
print(x.b)  # 1.0

The correct values should be 5.0 and 10.0.

Scipy/Numpy/Python version information:

1.5.4 1.18.2 sys.version_info(major=3, minor=6, micro=5, releaselevel='final', serial=0)

[mdhaber]

@reinhold-willcox it sounds like a and b were really supposed to be private attributes used only by developers. Since the intended behavior is not actually documented, and because the way they work now is acceptable to the developers, the proposed fix in gh-13228 is not going to be merged, so I will close this issue, too. Sorry I couldn’t be of more help! Please use the support method in new code.

[mdhaber]

The funny thing is that the initializer of the frozen distribution calls:

class rv_frozen(object):

    def __init__(self, dist, *args, **kwds):
        self.args = args
        self.kwds = kwds

        # create a new instance
        self.dist = dist.__class__(**dist._updated_ctor_param())

        shapes, _, _ = self.dist._parse_args(*args, **kwds)
        self.a, self.b = self.dist._get_support(*shapes)

but

    def _get_support(self, *args):
        """Return the support of the (unscaled, unshifted) distribution.

I wonder if this is intentional to call _get_support or if it should be calling support, e.g.

self.a, self.b = self.dist.support(*self.args, **self.kwds)

Well, I’ll submit a PR and see what the pros think.