Writing UUID using PyArrow does not set the UUID logical type on Parquet

[Fokko]

Describe the bug, including details regarding any error messages, version, and platform.

Consider the following path:

In [1]: import pyarrow as pa
   ...: import uuid
   ...: 
   ...: schema = pa.schema(
   ...:    [
   ...:        pa.field("uuid", pa.uuid(), nullable=False),
   ...:    ]
   ...: )
   ...: 
   ...: arr_table = pa.Table.from_pydict(
   ...:     {
   ...:         "uuid": [
   ...:             uuid.UUID("00000000-0000-0000-0000-000000000000").bytes,
   ...:             uuid.UUID("11111111-1111-1111-1111-111111111111").bytes,
   ...:         ],
   ...:     },
   ...:     schema=schema,
   ...: )
   ...: 
   ...: import pyarrow.parquet as pq
   ...: 
   ...: with pq.ParquetWriter("/tmp/some-parquet-with-uuid.parquet", schema=schema) as writer:
   ...:    writer.write(arr_table)
   ...: 

> parq /tmp/some-parquet-with-uuid.parquet -s                                                                                                                    

 # Schema 
 <pyarrow._parquet.ParquetSchema object at 0x105cd0480>
required group field_id=-1 schema {
  required fixed_len_byte_array(16) field_id=-1 uuid;
}

Example one that has been created by Iceberg (Java):

parq /var/folders/h0/wqtwn1ks0m3bksc8n7mp4_lr0000gp/T/hive12405183144450566152/table/data/uuid_bucket=7/00000-1-daaf21e3-fbee-4954-88d8-ea0371f62a6a-0-00001.parquet -s     

 # Schema 
 <pyarrow._parquet.ParquetSchema object at 0x1058ce840>
required group field_id=-1 table {
  required fixed_len_byte_array(16) field_id=1 uuid (UUID);
}

The (UUID) indicates the logical type annotation many readers rely on.

Component(s)

Parquet

[Fokko]

Looping in @rok since he drove the adoption of UUID as the extension type.

[wgtmac]

IIUC, this has been fixed by @paleolimbot in #45866 and we just need to wait for 21.0.0.

[Fokko]

Ohh nice, thanks @wgtmac 🙌