From 6f9744df75a16f6174c025dcedcc92d9b0b6e3d7 Mon Sep 17 00:00:00 2001
From: nkwork9999 <143652584+nkwork9999@users.noreply.github.com>
Date: Tue, 30 Jun 2026 22:56:23 +0900
Subject: [PATCH] Fix seed boolean null handling

Signed-off-by: nkwork9999 <143652584+nkwork9999@users.noreply.github.com>
---
 sqlmesh/core/model/definition.py |  5 ++---
 tests/dbt/test_transformation.py | 25 +++++++++++++++++++++++++
 2 files changed, 27 insertions(+), 3 deletions(-)

diff --git a/sqlmesh/core/model/definition.py b/sqlmesh/core/model/definition.py
index a059fa4e87..4936e9b176 100644
--- a/sqlmesh/core/model/definition.py
+++ b/sqlmesh/core/model/definition.py
@@ -1706,6 +1706,7 @@ def render(
 
     def render_seed(self) -> t.Iterator[QueryOrDF]:
         import numpy as np
+        import pandas as pd
 
         self._ensure_hydrated()
 
@@ -1746,8 +1747,6 @@ def render_seed(self) -> t.Iterator[QueryOrDF]:
 
             # convert all date/time types to native pandas timestamp
             for column in [*date_columns, *datetime_columns]:
-                import pandas as pd
-
                 df[column] = pd.to_datetime(df[column], infer_datetime_format=True, errors="ignore")  # type: ignore
 
             # extract datetime.date from pandas timestamp for DATE columns
@@ -1763,7 +1762,7 @@ def render_seed(self) -> t.Iterator[QueryOrDF]:
                     )
 
             for column in bool_columns:
-                df[column] = df[column].apply(lambda i: str_to_bool(str(i)))
+                df[column] = df[column].apply(lambda i: None if pd.isna(i) else str_to_bool(str(i)))
 
             df.loc[:, string_columns] = df[string_columns].mask(
                 cond=lambda x: x.notna(),  # type: ignore
diff --git a/tests/dbt/test_transformation.py b/tests/dbt/test_transformation.py
index fe6073dfad..934b8f64b8 100644
--- a/tests/dbt/test_transformation.py
+++ b/tests/dbt/test_transformation.py
@@ -947,6 +947,31 @@ def test_seed_single_whitespace_is_na(tmp_path):
     assert df["col_b"].to_list() == [1, None]
 
 
+def test_seed_boolean_nulls_are_preserved(tmp_path):
+    seed_csv = tmp_path / "seed.csv"
+    with open(seed_csv, "w", encoding="utf-8") as fd:
+        fd.write("id,test_ind\n")
+        fd.write("1,null\n")
+        fd.write("2,false\n")
+        fd.write("3,true\n")
+        fd.write("4,null\n")
+
+    seed = SeedConfig(
+        name="test_model",
+        package="foo",
+        path=Path(seed_csv),
+        column_types={"test_ind": "boolean"},
+    )
+
+    context = DbtContext()
+    context.project_name = "foo"
+    context.target = DuckDbConfig(name="target", schema="test")
+    sqlmesh_seed = seed.to_sqlmesh(context)
+
+    df = next(sqlmesh_seed.render_seed())
+    assert df["test_ind"].to_list() == [None, False, True, None]
+
+
 def test_seed_partial_column_inference(tmp_path):
     seed_csv = tmp_path / "seed.csv"
     with open(seed_csv, "w", encoding="utf-8") as fd: