Fix empty xtensor indexing

pytensor/xtensor/indexing.py

@@ -153,8 +153,18 @@ class Index(XOp):
         if len(idxs) > x_ndim:
             raise IndexError("Too many indices")
 
+        # Remove useless trailing slice(None) indices
+        # starting at -1 ensures we handle the case of no useful indices correctly with idxs[:0]
+        last_useful_index = -1
+        for i, idx in enumerate(idxs):
+            if isinstance(idx, slice) and idx == slice(None):
+                continue
+            last_useful_index = i
+
+        # Convert (useful) indices to symbolic variables
         idxs = [
-            as_idx_variable(idx, dim) for idx, dim in zip(idxs, x_dims, strict=False)
+            as_idx_variable(idx, dim)
+            for idx, dim in zip(idxs[: last_useful_index + 1], x_dims, strict=False)
         ]
 
         for i, idx in enumerate(idxs):
@@ -174,7 +184,9 @@ class Index(XOp):
                     idx_dim_shape = idx.type.shape[idx_dims.index(idx_dim)]
                     combine_dim_info(idx_dim, idx_dim_shape)
 
-        for dim_i, shape_i in zip(x_dims[i + 1 :], x_shape[i + 1 :]):
+        for dim_i, shape_i in zip(
+            x_dims[last_useful_index + 1 :], x_shape[last_useful_index + 1 :]
+        ):
             # Add back any unindexed dimensions
             if dim_i not in out_dims:
                 # If the dimension was not indexed, we keep it as is

pytensor/xtensor/type.py

@@ -488,9 +488,8 @@ class XTensorVariable(Variable[_XTensorTypeType, OptionalApplyType]):
                 )
             indexers = indexers_kwargs
 
-        if not indexers:
-            # No-op
-            return self
+        elif indexers is None:
+            indexers = {}
 
         if missing_dims not in {"raise", "warn", "ignore"}:
             raise ValueError(

tests/xtensor/test_indexing.py

@@ -12,6 +12,7 @@ from xarray import DataArray
 
 from pytensor.tensor import tensor
 from pytensor.xtensor import xtensor
+from tests.unittest_tools import assert_equal_computations
 from tests.xtensor.util import (
     xr_arange_like,
     xr_assert_allclose,
@@ -511,3 +512,33 @@ def test_diff(dim, n):
     else:
         expected_res = x_test.diff(dim, n=n)
     xr_assert_allclose(res, expected_res)
+
+
+def test_empty_index():
+    x = xtensor("x", shape=(5, 5), dims=("a", "b"))
+    out1 = x[()]
+    out2 = x[...]
+    out3 = x.isel({})
+    out4 = x.isel({"c": 0}, missing_dims="ignore")
+    assert_equal_computations([out1], [out2])
+    assert_equal_computations([out1], [out3])
+    assert_equal_computations([out1], [out4])
+
+    fn = xr_function([x], out1)
+    x_test = xr_random_like(x)
+    xr_assert_allclose(fn(x_test), x_test)
+
+
+def test_empty_update_index():
+    x = xtensor("x", shape=(5, 5), dims=("a", "b"))
+    out1 = x[()].inc(1)
+    out2 = x[...].inc(1)
+    out3 = x.isel({}).inc(1)
+    out4 = x.isel({"c": 0}, missing_dims="ignore").inc(1)
+    assert_equal_computations([out1], [out2])
+    assert_equal_computations([out1], [out3])
+    assert_equal_computations([out1], [out4])
+
+    fn = xr_function([x], out1)
+    x_test = xr_random_like(x)
+    xr_assert_allclose(fn(x_test), x_test + 1)