[Python] Cannot cast nested nullable field to not-nullable

[asfimport]

Casting from nullable field to not-nullable works provided all values are present. So for example this is a valid cast:

table = pa.table({'column_1': pa.array([1, 2 ,3])})table.cast(
    pa.schema([
        f.with_nullable(False) for f in table.schema
    ])
)

But it doesn't work for nested field. Here's an example:

import pyarrow as pa

record = {"nested_int": 1}

data_type = pa.struct(
    [
        pa.field("nested_int", pa.int32(), nullable=True),
    ]
)

data_type_after = pa.struct(
    [
        pa.field("nested_int", pa.int32(), nullable=False),
    ]
)

table = pa.table({"column_1": pa.array([record], data_type)})

table.cast(pa.schema([pa.field("column_1", data_type_after)])) 

Throws:

pyarrow.lib.ArrowTypeError: cannot cast nullable field to non-nullable field: struct<nested_int: int32> struct<nested_int: int32 not null> 

This is somewhat related to https://github.com/apache/arrow/issues/13177 and https://issues.apache.org/jira/browse/ARROW-16603 

Reporter: &res / @0x26res

_{Note: This issue was originally created as ARROW-18430. Please see the migration documentation for further details.}

[jorisvandenbossche]

Just ran into this as well:

>>> arr = pa.array([{'x': 1.0, 'y': 2.0}, {'x': 2.0, 'y': 3.0}])
>>> arr.type
StructType(struct<x: double, y: double>)

>>> arr.cast(pa.struct([pa.field("x", pa.float64(), nullable=False), pa.field("y", pa.float64(), nullable=False)]))
...
ArrowTypeError: cannot cast nullable field to non-nullable field: struct<x: double, y: double> struct<x: double not null, y: double not null>
/home/joris/scipy/repos/arrow/cpp/src/arrow/compute/exec.cc:920  kernel_->exec(kernel_ctx_, input, &output)
/home/joris/scipy/repos/arrow/cpp/src/arrow/compute/function.cc:277  executor->Execute(input, &listener)

[raulcd]

I am moving this to 17.0.0. Let me know if it should be part of 16.0.0

[wabu]

Also ran into this, would be great to address this soonish.

[NickCrews]

Also just ran into this. Not familiar at all with arrows codebase, but if this isn't that hard and someone can give me pointers, I can take a stab at this. Opened #43782 as a first attempt, I would love some help with it if you can.

[pitrou]

Issue resolved by pull request 43782 https://github.com/apache/arrow/pull/43782