apacheGH-37650: [Python] Check filter inputs in FilterMetaFunction (a…

…pache#38075) ### Rationale for this change Prevent a segfault upon passing a non-(chunked_)array object into `Table.filter`. See apache#37650. ### What changes are included in this PR? 1. Check filter Datum kind to make sure that it is an array or a chunked array 2. test that attempting to filter a table with another table raises a not implemented error ### Are these changes tested? In PyArrow, yes ### Are there any user-facing changes? Raises an error if a non-array or non-chunked_array object is passed into `Table.filter` * Closes: apache#37650 Lead-authored-by: Patrick Clarke <[email protected]> Co-authored-by: Antoine Pitrou <[email protected]> Signed-off-by: Antoine Pitrou <[email protected]>
loicalleyne · Nov 13, 2023 · 8ff1dbd · 8ff1dbd
1 parent 8148a27
commit 8ff1dbd
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 1 deletion.
diff --git a/cpp/src/arrow/compute/kernels/vector_selection_filter_internal.cc b/cpp/src/arrow/compute/kernels/vector_selection_filter_internal.cc
@@ -942,7 +942,7 @@ Result<std::shared_ptr<Table>> FilterTable(const Table& table, const Datum& filt
       inputs.back() = filter.chunked_array()->chunks();
       break;
     default:
-      return Status::NotImplemented("Filter should be array-like");
+      return Status::TypeError("Filter should be array-like");
   }
 
   // Rechunk inputs to allow consistent iteration over their respective chunks
@@ -998,6 +998,10 @@ class FilterMetaFunction : public MetaFunction {
   Result<Datum> ExecuteImpl(const std::vector<Datum>& args,
                             const FunctionOptions* options,
                             ExecContext* ctx) const override {
+    if (args[1].kind() != Datum::ARRAY && args[1].kind() != Datum::CHUNKED_ARRAY) {
+      return Status::TypeError("Filter should be array-like");
+    }
+
     const auto& filter_type = *args[1].type();
     const bool filter_is_plain_bool = filter_type.id() == Type::BOOL;
     const bool filter_is_ree_bool =

diff --git a/python/pyarrow/tests/test_compute.py b/python/pyarrow/tests/test_compute.py
@@ -1378,6 +1378,11 @@ def test_filter_errors():
                            match="must all be the same length"):
             obj.filter(mask)
 
+    scalar = pa.scalar(True)
+    for filt in [batch, table, scalar]:
+        with pytest.raises(TypeError):
+            table.filter(filt)
+
 
 def test_filter_null_type():
     # ARROW-10027