Refactor random tests and include explicit broadcasting tests

aesara/tensor/random/basic.py

@@ -29,8 +29,7 @@ except AttributeError:
     from numpy.lib.stride_tricks import _broadcast_shape
 
     def broadcast_shapes(*shapes):
-        arrays = [np.empty(x, dtype=[]) for x in shapes]
-        return _broadcast_shape(arrays)
+        return _broadcast_shape(*[np.empty(x, dtype=[]) for x in shapes])
 
 
 class ScipyRandomVariable(RandomVariable):

aesara/tensor/random/op.py

@@ -330,10 +330,6 @@ class RandomVariable(Op):
     def make_node(self, rng, size, dtype, *dist_params):
         """Create a random variable node.
 
-        XXX: Unnamed/non-keyword arguments are considered distribution
-        parameters!  If you want to set `size`, `rng`, and/or `name`, use their
-        keywords.
-
         Parameters
         ----------
         rng: RandomGeneratorType or RandomStateType

tests/unittest_tools.py

@@ -10,6 +10,7 @@ import aesara
 from aesara.compile.debugmode import str_diagnostic
 from aesara.configdefaults import config
 from aesara.gradient import verify_grad as orig_verify_grad
+from aesara.tensor.basic import as_tensor_variable
 from aesara.tensor.math import _allclose
 from aesara.tensor.math import add as aet_add
 
@@ -392,3 +393,10 @@ def assertFailure_fast(f):
         return test_with_assert
     else:
         return f
+
+
+def create_aesara_param(param_value):
+    """Create a `Variable` from a value and set its test value."""
+    p_aet = as_tensor_variable(param_value).type()
+    p_aet.tag.test_value = param_value
+    return p_aet