Add warning if observed in DensityDist is dict (#6292)

* 🚸 add warning if observed is dict
* Improve validation
* 👌 update error message

Author: symeneses

pymc/distributions/distribution.py

@@ -577,6 +577,14 @@ def random(mu, rng=None, size=None):
 
     def __new__(cls, name, *args, **kwargs):
         kwargs.setdefault("class_name", name)
+        if isinstance(kwargs.get("observed", None), dict):
+            raise TypeError(
+                "Since ``v4.0.0`` the ``observed`` parameter should be of type"
+                " ``pd.Series``, ``np.array``, or ``pm.Data``."
+                " Previous versions allowed passing distribution parameters as"
+                " a dictionary in ``observed``, in the current version these "
+                "parameters are positional arguments."
+            )
         return super().__new__(cls, name, *args, **kwargs)
 
     @classmethod

pymc/tests/distributions/test_distribution.py

@@ -145,9 +145,29 @@ def test_density_dist_with_random(self, size):
                 random=lambda mu, rng=None, size=None: rng.normal(loc=mu, scale=1, size=size),
                 observed=np.random.randn(100, *size),
             )
-
         assert obs.eval().shape == (100,) + size
 
+    def test_density_dist_with_random_invalid_observed(self):
+        with pytest.raises(
+            TypeError,
+            match=(
+                "Since ``v4.0.0`` the ``observed`` parameter should be of type"
+                " ``pd.Series``, ``np.array``, or ``pm.Data``."
+                " Previous versions allowed passing distribution parameters as"
+                " a dictionary in ``observed``, in the current version these "
+                "parameters are positional arguments."
+            ),
+        ):
+            size = (3,)
+            with pm.Model() as model:
+                mu = pm.Normal("mu", 0, 1)
+                pm.DensityDist(
+                    "density_dist",
+                    mu,
+                    random=lambda mu, rng=None, size=None: rng.normal(loc=mu, scale=1, size=size),
+                    observed={"values": np.random.randn(100, *size)},
+                )
+
     def test_density_dist_without_random(self):
         with pm.Model() as model:
             mu = pm.Normal("mu", 0, 1)