Github user HyukjinKwon commented on the issue: https://github.com/apache/spark/pull/22858 Yup, I think strictly we should change. Looks there are two occurrences at `udf` and `pands_udf` `isinstance(..., str)`. Another problem at PySpark is, inconsistent type comparison like type(...) == t` vs `isinstance(..., t)`. For instance, `type(...) == dict` vs `isinstance(..., dict)` - the former does not allow `OrderedDict` but the later allows. Another problem is, some types like `bool` at Python inherits `int`. In this case, `isinstance(...)` might produce unexpected results, for instance, ```python >>> isinstance(True, int) True ``` I was nervous about the cases above and didn't fix those changes so far.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org