Add a type guard for intX (pymc-devs#4569)

* add type guard for inX
* fix test for pandas
* fix posterior test, ints passed for float data

Closes pymc-devs#4279

Author: ferrine

pymc/aesaraf.py

@@ -444,6 +444,9 @@ def intX(X):
     """
     Convert a aesara tensor or numpy array to aesara.tensor.int32 type.
     """
+    # check value is already int, do nothing in this case
+    if (hasattr(X, "dtype") and "int" in str(X.dtype)) or isinstance(X, int):
+        return X
     intX = _conversion_map[aesara.config.floatX]
     try:
         return X.astype(intX)

pymc/tests/test_aesaraf.py

@@ -244,7 +244,7 @@ def test_convert_observed_data(input_dtype):
     assert isinstance(aesara_output, Variable)
     npt.assert_allclose(aesara_output.eval(), aesara_graph_input.eval())
     intX = pm.aesaraf._conversion_map[aesara.config.floatX]
-    if dense_input.dtype == intX or dense_input.dtype == aesara.config.floatX:
+    if "int" in str(dense_input.dtype) or dense_input.dtype == aesara.config.floatX:
         assert aesara_output.owner is None  # func should not have added new nodes
         assert aesara_output.name == input_name
     else:
@@ -254,7 +254,8 @@ def test_convert_observed_data(input_dtype):
     if "float" in input_dtype:
         assert aesara_output.dtype == aesara.config.floatX
     else:
-        assert aesara_output.dtype == intX
+        # only cast floats, leave ints as is
+        assert aesara_output.dtype == input_dtype
 
     # Check function behavior with generator data
     generator_output = func(square_generator)

pymc/tests/test_data.py

@@ -87,7 +87,7 @@ def test_sample_posterior_predictive_after_set_data(self):
             )
         # Predict on new data.
         with model:
-            x_test = [5, 6, 9]
+            x_test = [5.0, 6.0, 9.0]
             pm.set_data(new_data={"x": x_test})
             y_test = pm.sample_posterior_predictive(trace)
 
@@ -111,7 +111,7 @@ def test_sample_posterior_predictive_after_set_data_with_coords(self):
             )
         # Predict on new data.
         with model:
-            x_test = [5, 6]
+            x_test = [5.0, 6.0]
             pm.set_data(new_data={"x": x_test}, coords={"obs_id": ["a", "b"]})
             pm.sample_posterior_predictive(idata, extend_inferencedata=True, predictions=True)