fix(memtables): disallow duplicate column names when constructing memtables

Author: cpcloud

ibis/expr/api.py

@@ -7,6 +7,7 @@
 import functools
 import numbers
 import operator
+from collections import Counter
 from typing import TYPE_CHECKING, Any, NamedTuple, overload
 
 import ibis.expr.builders as bl
@@ -476,6 +477,16 @@ def _memtable_from_dataframe(
             (f"col{i:d}" for i in builtins.range(len(cols))),
         )
         df = df.rename(columns=dict(zip(cols, newcols)))
+
+    # verify that the DataFrame has no duplicate column names because ibis
+    # doesn't allow that
+    cols = df.columns
+    dupes = [name for name, count in Counter(cols).items() if count > 1]
+    if dupes:
+        raise IbisInputError(
+            f"Duplicate column names found in DataFrame when constructing memtable: {dupes}"
+        )
+
     op = ops.InMemoryTable(
         name=name if name is not None else util.gen_name("pandas_memtable"),
         schema=sch.infer(df) if schema is None else schema,

ibis/expr/tests/test_api.py

@@ -11,7 +11,7 @@
 import ibis.expr.datatypes as dt
 import ibis.expr.schema as sch
 from ibis import _
-from ibis.common.exceptions import IntegrityError
+from ibis.common.exceptions import IbisInputError, IntegrityError
 
 
 def test_schema_from_names_types():
@@ -117,3 +117,10 @@ def test_repr_deferred_with_exprs(f, sol):
     expr = f(t)
     res = repr(expr)
     assert res == sol
+
+
+def test_duplicate_columns_in_memtable_not_allowed():
+    df = pd.DataFrame([[1, 2], [3, 4]], columns=["a", "a"])
+
+    with pytest.raises(IbisInputError, match="Duplicate column names"):
+        ibis.memtable(df)