Hi,
I am trying to understand the cast behavior for Arrow Array for nested
types. For example, PyArrow supports cast from list(int64) to list(int32):
>>> import pyarrow as pa
>>> a = pa.array([[1, 2], [3, 4, 5]])
>>> a.type
ListType(list<item: int64>)
>>> b = a.cast(pa.list_(pa.int32()))
>>> b.type
ListType(list<item: int32>)
However, it's not supported for struct(int64, int64) to struct(int32,
int32):
>>> a = pa.array( [(1, 2), (3, 4)] , type = pa.struct([('f1', pa.int64()) ,
('f2', pa.int64()) ] ) )
>>> a.type
>>> b = a.cast( pa.struct([('f1', pa.int32()) , ('f2', pa.int32()) ] ) )
......
pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 120, in pyarrow.lib.check_status
pyarrow.lib.ArrowNotImplementedError: Unsupported cast from struct<f1:
int64, f2: int64> to struct using function cast_struct
I am asking since both casts are supported in ANSI SQL. I am wondering if
CAST for struct is designed not supported in PyArrow? Thanks!
--
Best Regards,
Wenlei Xie
Email: [email protected]
