Deprecated usecols with out of bounds indices in read_csv with c engine

doc/source/whatsnew/v1.3.0.rst

@@ -611,6 +611,7 @@ Deprecations
 - Deprecated the ``level`` keyword for :class:`DataFrame` and :class:`Series` aggregations; use groupby instead (:issue:`39983`)
 - The ``inplace`` parameter of :meth:`Categorical.remove_categories`, :meth:`Categorical.add_categories` is deprecated and will be removed in a future version (:issue:`37643`)
 - Deprecated :func:`merge` producing duplicated columns through the ``suffixes`` keyword  and already existing columns (:issue:`22818`)
+- Deprecated using ``usecols`` with out of bounds indices for ``read_csv`` with ``engine="c"`` (:issue:`25623`)
 
 .. ---------------------------------------------------------------------------
 

pandas/_libs/parsers.pyx

@@ -941,6 +941,17 @@ cdef class TextReader:
                               f"{self.table_width - self.leading_cols} "
                               f"and found {num_cols}")
 
+        if (self.usecols is not None and not callable(self.usecols) and
+                all(isinstance(u, int) for u in self.usecols)):
+            missing_usecols = [col for col in self.usecols if col >= num_cols]
+            if missing_usecols:
+                warnings.warn(
+                    "Defining usecols with out of bounds indices is deprecated "
+                    "and will raise a ParserError in a future version.",
+                    FutureWarning,
+                    stacklevel=6,
+                )
+
         results = {}
         nused = 0
         for i in range(self.table_width):

pandas/tests/io/parser/test_c_parser_only.py

@@ -680,3 +680,16 @@ def test_float_precision_options(c_parser_only):
 
     with pytest.raises(ValueError, match=msg):
         parser.read_csv(StringIO(s), float_precision="junk")
+
+
+def test_usecols_indices_out_of_bounds(c_parser_only):
+    # GH#25623
+    parser = c_parser_only
+    data = """
+a,b
+1,2
+    """
+    with tm.assert_produces_warning(FutureWarning):
+        result = parser.read_csv(StringIO(data), usecols=[0, 2])
+    expected = DataFrame({"a": [1], "b": None})
+    tm.assert_frame_equal(result, expected)