REGR: Avoid error when defining non existing usecols with python engine (#41244)

Author: phofl

pandas/io/parsers/python_parser.py

@@ -298,10 +298,10 @@ def _exclude_implicit_index(self, alldata):
             # error: Cannot determine type of 'index_col'
             offset = len(self.index_col)  # type: ignore[has-type]
 
-        if self._col_indices is not None and len(names) != len(self._col_indices):
-            names = [names[i] for i in sorted(self._col_indices)]
-
-        return {name: alldata[i + offset] for i, name in enumerate(names)}, names
+        len_alldata = len(alldata)
+        return {
+            name: alldata[i + offset] for i, name in enumerate(names) if i < len_alldata
+        }, names
 
     # legacy
     def get_chunk(self, size=None):
@@ -472,7 +472,12 @@ def _infer_columns(self):
                     self._handle_usecols(columns, names)
                 else:
                     num_original_columns = len(names)
-                columns = [names]
+                if self._col_indices is not None and len(names) != len(
+                    self._col_indices
+                ):
+                    columns = [[names[i] for i in sorted(self._col_indices)]]
+                else:
+                    columns = [names]
             else:
                 columns = self._handle_usecols(columns, columns[0])
         else:

pandas/tests/io/parser/usecols/test_usecols_basic.py

@@ -373,3 +373,18 @@ def test_usecols_subset_names_mismatch_orig_columns(all_parsers, usecols):
     result = parser.read_csv(StringIO(data), header=0, names=names, usecols=usecols)
     expected = DataFrame({"A": [1, 5], "C": [3, 7]})
     tm.assert_frame_equal(result, expected)
+
+
+@pytest.mark.parametrize("names", [None, ["a", "b"]])
+def test_usecols_indices_out_of_bounds(all_parsers, names):
+    # GH#25623
+    parser = all_parsers
+    data = """
+a,b
+1,2
+    """
+    result = parser.read_csv(StringIO(data), usecols=[0, 2], names=names, header=0)
+    expected = DataFrame({"a": [1], "b": [None]})
+    if names is None and parser.engine == "python":
+        expected = DataFrame({"a": [1]})
+    tm.assert_frame_equal(result, expected)