feat(clickhouse): properly support native boolean types

Author: cpcloud

ibis/backends/clickhouse/__init__.py

@@ -81,9 +81,12 @@ class Options(ibis.config.Config):
         ----------
         temp_db : str
             Database to use for temporary objects.
+        bool_type : str
+            Type to use for boolean columns
         """
 
         temp_db: str = "__ibis_tmp"
+        bool_type: str = "Boolean"
 
     def __init__(self, *args, external_tables=None, **kwargs):
         super().__init__(*args, **kwargs)

ibis/backends/clickhouse/datatypes.py

@@ -4,6 +4,7 @@
 
 import parsy
 
+import ibis
 import ibis.expr.datatypes as dt
 from ibis.common.parsing import (
     COMMA,
@@ -19,6 +20,10 @@
 )
 
 
+def _bool_type():
+    return getattr(getattr(ibis.options, "clickhouse", None), "bool_type", "Boolean")
+
+
 def parse(text: str) -> dt.DataType:
     @parsy.generate
     def datetime():
@@ -35,8 +40,9 @@ def datetime():
         | spaceless_string("smallint", "int16", "int2").result(dt.Int16(nullable=False))
         | spaceless_string("date32", "date").result(dt.Date(nullable=False))
         | spaceless_string("time").result(dt.Time(nullable=False))
-        | spaceless_string("tinyint", "int8", "int1", "boolean", "bool").result(
-            dt.Int8(nullable=False)
+        | spaceless_string("tinyint", "int8", "int1").result(dt.Int8(nullable=False))
+        | spaceless_string("boolean", "bool").result(
+            getattr(dt, _bool_type())(nullable=False)
         )
         | spaceless_string("integer", "int32", "int4", "int").result(
             dt.Int32(nullable=False)
@@ -223,6 +229,11 @@ def _(ty: dt.DataType) -> str:
     return type(ty).__name__.capitalize()
 
 
+@serialize_raw.register(dt.Boolean)
+def _(_: dt.Boolean) -> str:
+    return _bool_type()
+
+
 @serialize_raw.register(dt.Array)
 def _(ty: dt.Array) -> str:
     return f"Array({serialize(ty.value_type)})"

ibis/backends/clickhouse/tests/conftest.py

@@ -23,9 +23,13 @@ class TestConf(UnorderedComparator, BackendTest, RoundHalfToEven):
     returned_timestamp_unit = 's'
     supported_to_timestamp_units = {'s'}
     supports_floating_modulus = False
-    bool_is_int = True
     supports_json = False
 
+    @property
+    def native_bool(self) -> bool:
+        [(value,)] = self.connection._client.execute("SELECT true")
+        return isinstance(value, bool)
+
     @staticmethod
     def _load_data(
         data_dir: Path,

ibis/backends/clickhouse/tests/test_operators.py

@@ -5,6 +5,7 @@
 import pandas as pd
 import pandas.testing as tm
 import pytest
+from pytest import param
 
 import ibis
 import ibis.expr.datatypes as dt
@@ -156,11 +157,17 @@ def test_field_in_literals(con, alltypes, translate, container):
     assert len(con.execute(expr))
 
 
-@pytest.mark.parametrize('column', ['int_col', 'float_col', 'bool_col'])
-def test_negate(con, alltypes, translate, column):
-    # clickhouse represent boolean as UInt8
+@pytest.mark.parametrize(
+    ("column", "operator"),
+    [
+        param("int_col", "-", id="int_col"),
+        param("float_col", "-", id="float_col"),
+        param("bool_col", "NOT ", id="bool_col"),
+    ],
+)
+def test_negate(con, alltypes, translate, column, operator):
     expr = -alltypes[column]
-    assert translate(expr.op()) == f'-{column}'
+    assert translate(expr.op()) == f"{operator}{column}"
     assert len(con.execute(expr))
 
 

ibis/backends/datafusion/tests/conftest.py

@@ -16,7 +16,7 @@ class TestConf(BackendTest, RoundAwayFromZero):
     # additional_skipped_operations = frozenset({ops.StringSQLLike})
     # supports_divide_by_zero = True
     # returned_timestamp_unit = 'ns'
-    bool_is_int = True
+    native_bool = False
     supports_structs = False
     supports_json = False
 

ibis/backends/mysql/tests/conftest.py

@@ -26,7 +26,7 @@ class TestConf(BackendTest, RoundHalfToEven):
     returned_timestamp_unit = 's'
     supports_arrays = False
     supports_arrays_outside_of_select = supports_arrays
-    bool_is_int = True
+    native_bool = False
     supports_structs = False
 
     def __init__(self, data_directory: Path) -> None:

ibis/backends/tests/base.py

@@ -72,7 +72,7 @@ class BackendTest(abc.ABC):
     returned_timestamp_unit = 'us'
     supported_to_timestamp_units = {'s', 'ms', 'us'}
     supports_floating_modulus = True
-    bool_is_int = False
+    native_bool = True
     supports_structs = True
     supports_json = True
     reduction_tolerance = 1e-7
@@ -163,7 +163,7 @@ def least(f: Callable[..., ir.Value], *args: ir.Value) -> ir.Value:
     @property
     def functional_alltypes(self) -> ir.Table:
         t = self.connection.table('functional_alltypes')
-        if self.bool_is_int:
+        if not self.native_bool:
             return t.mutate(bool_col=t.bool_col == 1)
         return t