GH-44500: [Python][Parquet] Map Parquet logical types to Arrow extension types by default (#46772)

### Rationale for this change

The Parquet C++ implementation now supports reading four logical types (JSON, UUID, Geometry, Geography) as Arrow extension types; however, users have to opt-in to avoid loosing the logical type on read.

### What changes are included in this PR?

This PR sets the default value of `arrow_extensions_enabled` to `True` (in Python).

### Are these changes tested?

Yes, the behaviour of `arrow_extensions_enabled` was already tested (and tests were updated to reflect the new default value).

### Are there any user-facing changes?

**This PR includes breaking changes to public APIs.**

Reading Parquet files that contained a JSON or UUID logical type will now have an extension type rather than string or fixed size binary, respectively. Python users that were relying on the previous behaviour would have to explicitly cast to storage or use `read_table(..., arrow_extensions_enabled=False)` after this PR:

```python
import uuid
import pyarrow as pa

json_array = pa.array(['{"k": "v"}'], pa.json_())
json_array.cast(pa.string())
#> [
#>   "{"k": "v"}"
#> ]

uuid_array = pa.array([uuid.uuid4().bytes], pa.uuid())
uuid_array.cast(pa.binary(16))
#> <pyarrow.lib.FixedSizeBinaryArray object at 0x11e42b1c0>
#> [
#>   746C1022AB434A97972E1707EC3EE8F4
#> ]
```
* GitHub Issue: #44500

Authored-by: Dewey Dunnington <dewey@wherobots.com>
Signed-off-by: AlenkaF <frim.alenka@gmail.com>

Author: paleolimbot

python/pyarrow/parquet/core.py

@@ -261,7 +261,7 @@ class ParquetFile:
         it will be parsed as an URI to determine the filesystem.
     page_checksum_verification : bool, default False
         If True, verify the checksum for each page read from the file.
-    arrow_extensions_enabled : bool, default False
+    arrow_extensions_enabled : bool, default True
         If True, read Parquet logical types as Arrow extension types where possible,
         (e.g., read JSON as the canonical `arrow.json` extension type or UUID as
         the canonical `arrow.uuid` extension type).
@@ -314,7 +314,7 @@ def __init__(self, source, *, metadata=None, common_metadata=None,
                  coerce_int96_timestamp_unit=None,
                  decryption_properties=None, thrift_string_size_limit=None,
                  thrift_container_size_limit=None, filesystem=None,
-                 page_checksum_verification=False, arrow_extensions_enabled=False):
+                 page_checksum_verification=False, arrow_extensions_enabled=True):
 
         self._close_source = getattr(source, 'closed', True)
 
@@ -1321,7 +1321,7 @@ class ParquetDataset:
     sufficient for most Parquet files.
 page_checksum_verification : bool, default False
     If True, verify the page checksum for each page read from the file.
-arrow_extensions_enabled : bool, default False
+arrow_extensions_enabled : bool, default True
     If True, read Parquet logical types as Arrow extension types where possible,
     (e.g., read JSON as the canonical `arrow.json` extension type or UUID as
     the canonical `arrow.uuid` extension type).
@@ -1339,7 +1339,7 @@ def __init__(self, path_or_paths, filesystem=None, schema=None, *, filters=None,
                  decryption_properties=None, thrift_string_size_limit=None,
                  thrift_container_size_limit=None,
                  page_checksum_verification=False,
-                 arrow_extensions_enabled=False):
+                 arrow_extensions_enabled=True):
         import pyarrow.dataset as ds
 
         # map format arguments
@@ -1739,7 +1739,7 @@ def partitioning(self):
     sufficient for most Parquet files.
 page_checksum_verification : bool, default False
     If True, verify the checksum for each page read from the file.
-arrow_extensions_enabled : bool, default False
+arrow_extensions_enabled : bool, default True
     If True, read Parquet logical types as Arrow extension types where possible,
     (e.g., read JSON as the canonical `arrow.json` extension type or UUID as
     the canonical `arrow.uuid` extension type).
@@ -1839,7 +1839,7 @@ def read_table(source, *, columns=None, use_threads=True,
                decryption_properties=None, thrift_string_size_limit=None,
                thrift_container_size_limit=None,
                page_checksum_verification=False,
-               arrow_extensions_enabled=False):
+               arrow_extensions_enabled=True):
 
     try:
         dataset = ParquetDataset(

python/pyarrow/tests/parquet/test_data_types.py

@@ -569,14 +569,15 @@ def test_json_extension_type(storage_type):
     _check_roundtrip(
         table,
         pa.table({"ext": pa.array(data, pa.string())}),
+        {"arrow_extensions_enabled": False},
         store_schema=False)
 
     # With arrow_extensions_enabled=True on read, we get a arrow.json back
     # (but with string() storage)
     _check_roundtrip(
         table,
         pa.table({"ext": pa.array(data, pa.json_(pa.string()))}),
-        read_table_kwargs={"arrow_extensions_enabled": True},
+        {"arrow_extensions_enabled": True},
         store_schema=False)
 
 
@@ -594,11 +595,13 @@ def test_uuid_extension_type():
     _check_roundtrip(
         table,
         pa.table({"ext": pa.array(data, pa.binary(16))}),
+        {"arrow_extensions_enabled": False},
         store_schema=False)
     _check_roundtrip(
         table,
         table,
-        {"arrow_extensions_enabled": True}, store_schema=False)
+        {"arrow_extensions_enabled": True},
+        store_schema=False)
 
 
 def test_undefined_logical_type(parquet_test_datadir):