jorisvandenbossche commented on a change in pull request #11076:
URL: https://github.com/apache/arrow/pull/11076#discussion_r713182216
##########
File path: cpp/src/arrow/python/inference.cc
##########
@@ -354,12 +365,15 @@ class TypeInferrer {
*keep_going = make_unions_;
} else if (PyArray_CheckAnyScalarExact(obj)) {
RETURN_NOT_OK(VisitDType(PyArray_DescrFromScalar(obj), keep_going));
- } else if (PyList_Check(obj)) {
- RETURN_NOT_OK(VisitList(obj, keep_going));
+ } else if (PySet_Check(obj) || (Py_TYPE(obj) == &PyDictValues_Type)) {
+ RETURN_NOT_OK(VisitSet(obj, keep_going));
} else if (PyArray_Check(obj)) {
RETURN_NOT_OK(VisitNdarray(obj, keep_going));
} else if (PyDict_Check(obj)) {
RETURN_NOT_OK(VisitDict(obj));
+ } else if (PyList_Check(obj) || PyTuple_Check(obj) ||
+ PyObject_IsInstance(obj, deque_type_.obj())) {
Review comment:
I personally agree it doesn't seem needed to add special case code in
C++ to handle deques.
I would actually even doubt whether we should handle sets (they are
unordered, so not necessarily directly mapping to an array, the order you get
from iteration / conversion to list is not deterministic/guaranteed AFAIK). But
since we already support sets in non-nested situations, I understand that this
is making that consistent.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
