Fix ValueError on unlisted CrateDB column type identifiers

CHANGES.rst

@@ -5,6 +5,14 @@ Changes for crate
 Unreleased
 ================
 
+- Breaking change: ``DefaultTypeConverter`` now decodes ``DataType.UUID``
+  columns to Python ``uuid.UUID`` objects instead of returning the raw string.
+
+- Added CrateDB column type identifiers to ``DataType``: ``INTERVAL`` (17), 
+  ``ROW`` (18), ``FLOAT_VECTOR`` (28), ``UUID`` (29), and ``REGTYPE`` (30).
+  Fixed ``Converter.get()`` raising ``ValueError`` for column type identifiers
+  it does not know. Unknown identifiers now fall back to the default converter.
+
 - Breaking change: ``connect()`` now raises ``ConnectionError`` immediately if
 no configured server node responds.
 

docs/by-example/cursor.rst

@@ -343,7 +343,7 @@ Python data type conversion
 
 The cursor object can optionally convert database types to native Python data
 types. Currently, this is implemented for the CrateDB data types ``IP``,
-``TIMESTAMP``, ``TIMETZ``, and ``BIT`` on behalf of the
+``TIMESTAMP``, ``TIMETZ``, ``BIT``, and ``UUID`` on behalf of the
 ``DefaultTypeConverter``.
 
     >>> cursor = connection.cursor(converter=DefaultTypeConverter())
@@ -403,6 +403,29 @@ Executing the query and fetching the decoded result:
     ['0110']
 
 
+CrateDB's ``UUID`` type is returned over HTTP as a string. It is decoded to a
+Python ``uuid.UUID`` object.
+
+    >>> cursor = connection.cursor(converter=DefaultTypeConverter())
+
+.. hide: set up the mocked response::
+
+    >>> connection.client.set_next_response({
+    ...     "col_types": [29],
+    ...     "rows":[ [ "a5b3c1e0-1b7f-4f3e-9a2d-6c4e8f0a1b2c" ] ],
+    ...     "cols":[ "id" ],
+    ...     "rowcount":1,
+    ...     "duration":1
+    ... })
+
+Executing the query and fetching the decoded result:
+
+    >>> cursor.execute("select 'a5b3c1e0-1b7f-4f3e-9a2d-6c4e8f0a1b2c'::uuid")
+
+    >>> cursor.fetchone()
+    [UUID('a5b3c1e0-1b7f-4f3e-9a2d-6c4e8f0a1b2c')]
+
+
 Custom data type conversion
 ===========================
 

src/crate/client/converter.py

@@ -27,6 +27,7 @@
 import datetime as dt
 import ipaddress
 import re
+import uuid
 from copy import deepcopy
 from enum import Enum
 from typing import Any, Callable, Dict, List, Optional, Union
@@ -89,6 +90,17 @@ def _to_bit_string(value: Optional[str]) -> Optional[str]:
     return match.group(1)
 
 
+def _to_uuid(value: Optional[str]) -> Optional[uuid.UUID]:
+    """
+    Convert a CrateDB UUID wire value to a Python ``uuid.UUID``.
+
+    https://docs.python.org/3/library/uuid.html
+    """
+    if value is None:
+        return None
+    return uuid.UUID(value)
+
+
 def _to_default(value: Optional[Any]) -> Optional[Any]:
     return value
 
@@ -113,6 +125,8 @@ class DataType(Enum):
     GEOSHAPE = 14
     TIMESTAMP_WITHOUT_TZ = 15
     UNCHECKED_OBJECT = 16
+    INTERVAL = 17
+    ROW = 18
     REGPROC = 19
     TIME = 20
     OIDVECTOR = 21
@@ -122,19 +136,33 @@ class DataType(Enum):
     BIT = 25
     JSON = 26
     CHARACTER = 27
+    FLOAT_VECTOR = 28
+    UUID = 29
+    REGTYPE = 30
     ARRAY = 100
 
 
 ConverterMapping = Dict[DataType, ConverterFunction]
 
 
+def _resolve(type_: Any) -> Optional[DataType]:
+    """
+    Map a wire type identifier to a `DataType`.
+    """
+    try:
+        return DataType(type_)
+    except ValueError:
+        return None
+
+
 # Map data type identifier to converter function.
 _DEFAULT_CONVERTERS: ConverterMapping = {
     DataType.IP: _to_ipaddress,
     DataType.TIMESTAMP_WITH_TZ: _to_datetime,
     DataType.TIMESTAMP_WITHOUT_TZ: _to_datetime,
     DataType.TIME: _to_time,
     DataType.BIT: _to_bit_string,
+    DataType.UUID: _to_uuid,
 }
 
 
@@ -149,9 +177,12 @@ def __init__(
 
     def get(self, type_: ColTypesDefinition) -> ConverterFunction:
         if isinstance(type_, int):
-            return self._mappings.get(DataType(type_), self._default)
+            data_type = _resolve(type_)
+            if data_type is None:
+                return self._default
+            return self._mappings.get(data_type, self._default)
         type_, inner_type = type_
-        if DataType(type_) is not DataType.ARRAY:
+        if _resolve(type_) is not DataType.ARRAY:
             raise ValueError(
                 f"Data type {type_} is not implemented as collection type"
             )