Pass size to specialized truncated dispatch

pymc/distributions/truncated.py

@@ -53,7 +53,7 @@ def update(self, node: Node):
 
 
 @singledispatch
-def _truncated(op: Op, lower, upper, *params):
+def _truncated(op: Op, lower, upper, size, *params):
     """Return the truncated equivalent of another `RandomVariable`."""
     raise NotImplementedError(f"{op} does not have an equivalent truncated version implemented")
 
@@ -150,7 +150,7 @@ def rv_op(cls, dist, lower, upper, max_n_steps, size=None):
 
         # Try to use specialized Op
         try:
-            return _truncated(dist.owner.op, lower, upper, *dist.owner.inputs)
+            return _truncated(dist.owner.op, lower, upper, size, *dist.owner.inputs)
         except NotImplementedError:
             pass
 
@@ -339,7 +339,7 @@ def truncated_logprob(op, values, *inputs, **kwargs):
 
 
 @_truncated.register(NormalRV)
-def _truncated_normal(op, lower, upper, rng, size, dtype, mu, sigma):
+def _truncated_normal(op, lower, upper, size, rng, old_size, dtype, mu, sigma):
     return TruncatedNormal.dist(
         mu=mu,
         sigma=sigma,

pymc/tests/distributions/test_truncated.py

@@ -53,12 +53,31 @@ def _icdf_not_implemented(*args, **kwargs):
     raise NotImplementedError()
 
 
-def test_truncation_specialized_op():
+@pytest.mark.parametrize("shape_info", ("shape", "dims", "observed"))
+def test_truncation_specialized_op(shape_info):
     rng = aesara.shared(np.random.default_rng())
     x = at.random.normal(0, 10, rng=rng, name="x")
 
-    xt = Truncated.dist(x, lower=5, upper=15, shape=(100,))
+    with Model(coords={"dim": range(100)}) as m:
+        if shape_info == "shape":
+            xt = Truncated("xt", dist=x, lower=5, upper=15, shape=(100,))
+        elif shape_info == "dims":
+            xt = Truncated("xt", dist=x, lower=5, upper=15, dims=("dim",))
+        elif shape_info == "observed":
+            xt = Truncated(
+                "xt",
+                dist=x,
+                lower=5,
+                upper=15,
+                observed=np.empty(
+                    100,
+                ),
+            )
+        else:
+            raise ValueError(f"Not a valid shape_info parametrization: {shape_info}")
+
     assert isinstance(xt.owner.op, TruncatedNormalRV)
+    assert xt.shape.eval() == (100,)
 
     # Test RNG is not reused
     assert xt.owner.inputs[0] is not rng