From 4a484b59b4a6ea2a2c7d6b5790c001634280236e Mon Sep 17 00:00:00 2001
From: Gil Forsyth <gil@forsyth.dev>
Date: Tue, 5 Sep 2023 14:39:32 -0400
Subject: [PATCH] feat(duckdb): use new TypeMapper class for parsing dtypes

---
 ibis/backends/duckdb/__init__.py  | 23 +++++++------
 ibis/backends/duckdb/datatypes.py | 56 -------------------------------
 2 files changed, 12 insertions(+), 67 deletions(-)

diff --git a/ibis/backends/duckdb/__init__.py b/ibis/backends/duckdb/__init__.py
index ea33944458233..6c6aeb0b53484 100644
--- a/ibis/backends/duckdb/__init__.py
+++ b/ibis/backends/duckdb/__init__.py
@@ -11,30 +11,32 @@
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Iterable, Iterator, Mapping, MutableMapping
 
-import duckdb
-import pyarrow as pa
-import sqlglot as sg
-import toolz
-from packaging.version import parse as vparse
-
 import ibis
 import ibis.common.exceptions as exc
 import ibis.expr.datatypes as dt
 import ibis.expr.operations as ops
 import ibis.expr.schema as sch
 import ibis.expr.types as ir
+import pyarrow as pa
+import sqlglot as sg
+import toolz
 from ibis import util
 from ibis.backends.base import CanCreateSchema
 from ibis.backends.base.sql import BaseBackend
+from ibis.backends.base.sqlglot.datatypes import DuckDBType
 from ibis.backends.duckdb.compiler import translate
-from ibis.backends.duckdb.datatypes import parse, serialize, DuckDBType
+from ibis.backends.duckdb.datatypes import serialize
 from ibis.expr.operations.relations import PandasDataFrameProxy
 from ibis.expr.operations.udf import InputType
-from ibis.formats.pyarrow import PyArrowData
 from ibis.formats.pandas import PandasData
+from ibis.formats.pyarrow import PyArrowData
+from packaging.version import parse as vparse
+
+import duckdb
 
 if TYPE_CHECKING:
     from collections.abc import Iterable, Iterator, Mapping, MutableMapping
+
     import ibis.expr.operations as ops
     import pandas as pd
     import torch
@@ -247,7 +249,7 @@ def get_schema(self, table_name: str, database: str | None = None) -> sch.Schema
         names, types, *_ = results.fetch_arrow_table()
         names = names.to_pylist()
         types = types.to_pylist()
-        return sch.Schema(dict(zip(names, map(parse, types))))
+        return sch.Schema(dict(zip(names, map(DuckDBType.from_string, types))))
 
     def list_databases(self, like: str | None = None) -> list[str]:
         result = self.raw_sql("PRAGMA database_list;")
@@ -1146,8 +1148,7 @@ def _metadata(self, query: str) -> Iterator[tuple[str, dt.DataType]]:
             map(as_py, rows["column_type"]),
             map(as_py, rows["null"]),
         ):
-            ibis_type = parse(type)
-            # ibis_type = DuckDBType.from_string(type, nullable=nullable)
+            ibis_type = DuckDBType.from_string(type, nullable=nullable)
             yield name, ibis_type.copy(nullable=null.lower() == "yes")
 
     def _register_in_memory_tables(self, expr: ir.Expr) -> None:
diff --git a/ibis/backends/duckdb/datatypes.py b/ibis/backends/duckdb/datatypes.py
index 80102f5b53b42..f8abd82bd5319 100644
--- a/ibis/backends/duckdb/datatypes.py
+++ b/ibis/backends/duckdb/datatypes.py
@@ -2,63 +2,7 @@
 
 import functools
 
-import sqlalchemy.dialects.postgresql as psql
-
 import ibis.expr.datatypes as dt
-from ibis.backends.base.sql.alchemy.datatypes import AlchemyType
-from ibis.backends.base.sqlglot.datatypes import DuckDBType as SqlglotDuckdbType
-
-# _from_duckdb_types = {
-#     psql.BYTEA: dt.Binary,
-#     psql.UUID: dt.UUID,
-#     ducktypes.TinyInteger: dt.Int8,
-#     ducktypes.SmallInteger: dt.Int16,
-#     ducktypes.Integer: dt.Int32,
-#     ducktypes.BigInteger: dt.Int64,
-#     ducktypes.HugeInteger: dt.Decimal(38, 0),
-#     ducktypes.UInt8: dt.UInt8,
-#     ducktypes.UTinyInteger: dt.UInt8,
-#     ducktypes.UInt16: dt.UInt16,
-#     ducktypes.USmallInteger: dt.UInt16,
-#     ducktypes.UInt32: dt.UInt32,
-#     ducktypes.UInteger: dt.UInt32,
-#     ducktypes.UInt64: dt.UInt64,
-#     ducktypes.UBigInteger: dt.UInt64,
-# }
-
-# _to_duckdb_types = {
-#     dt.UUID: psql.UUID,
-#     dt.Int8: ducktypes.TinyInteger,
-#     dt.Int16: ducktypes.SmallInteger,
-#     dt.Int32: ducktypes.Integer,
-#     dt.Int64: ducktypes.BigInteger,
-#     dt.UInt8: ducktypes.UTinyInteger,
-#     dt.UInt16: ducktypes.USmallInteger,
-#     dt.UInt32: ducktypes.UInteger,
-#     dt.UInt64: ducktypes.UBigInteger,
-# }
-
-
-class DuckDBType(AlchemyType):
-    dialect = "duckdb"
-
-    @classmethod
-    def to_ibis(cls, typ, nullable=True):
-        if dtype := _from_duckdb_types.get(type(typ)):
-            return dtype(nullable=nullable)
-        else:
-            return super().to_ibis(typ, nullable=nullable)
-
-    @classmethod
-    def from_ibis(cls, dtype):
-        if typ := _to_duckdb_types.get(type(dtype)):
-            return typ
-        else:
-            return super().from_ibis(dtype)
-
-    @classmethod
-    def from_string(cls, type_string, nullable=True):
-        return SqlglotDuckdbType.from_string(type_string, nullable=nullable)
 
 
 @functools.singledispatch