fix: address reviewer feedback for Db2 CI isolation and error handling

- Skip test_sqlglot_extended_correctly when db2-sqlglot-dialect is not installed
- Use static exp.DataType.Type enums in SCHEMA_DIFFER_KWARGS in db2.py to allow safe doc-tests collection without plugin
- Revert dead db2 migration code in migration.py to match upstream
- Raise explicit ConfigError in Db2ConnectionConfig._engine_adapter when plugin is missing
- Add plugin disclaimer and SCD Type 2 limitation note in db2.md docs
- Add pyproject.toml to db2.yaml path filters
- Clean up commented-out test configurations in config.yaml and test_integration.py

Signed-off-by: IBM Db2 Eco System <Hdm-dev-persona-db2-eco-system@ibm.com>

Author: hdm-db2-eco-system

.github/workflows/db2.yaml

@@ -10,6 +10,7 @@ on:
       - 'tests/core/engine_adapter/integration/test_integration_db2.py'
       - 'tests/core/engine_adapter/integration/docker/compose.db2.yaml'
       - '.github/workflows/db2.yaml'
+      - 'pyproject.toml'
   pull_request:
     branches:
       - main
@@ -19,6 +20,7 @@ on:
       - 'tests/core/engine_adapter/integration/test_integration_db2.py'
       - 'tests/core/engine_adapter/integration/docker/compose.db2.yaml'
       - '.github/workflows/db2.yaml'
+      - 'pyproject.toml'
 
 concurrency:
   group: db2-${{ github.event.pull_request.number || github.sha }}

docs/integrations/engines/db2.md

@@ -5,6 +5,8 @@ This page provides information about how to use SQLMesh with [IBM Db2](https://w
 !!! info
     The Db2 engine adapter is a community contribution. Due to this, only limited community support is available.
 
+    This adapter depends on the external `db2-sqlglot-dialect` package, which pins its own SQLGlot version range. It isn't covered by SQLMesh's core CI and may temporarily break or fail to install after SQLMesh upgrades SQLGlot, until the plugin catches up.
+
 ## Local/Built-in Scheduler
 
 **Engine Adapter Type**: `db2`
@@ -63,6 +65,7 @@ physical_table_naming_convention: hash_md5
 
 ## Limitations
 
+- **SCD Type 2 unsupported**: SCD Type 2 models are currently not supported on Db2.
 - **Single catalog only**: Db2 operates in single-catalog mode; cross-catalog queries are not supported.
 - **No inline column comments**: Column-level comments cannot be set inline during table creation.
 - **No atomic table replacement**: Db2 does not support `CREATE OR REPLACE TABLE`, so full model refreshes are not atomic. There is a brief window during which the table may be empty or partially populated.

sqlmesh/core/config/connection.py

@@ -2643,11 +2643,13 @@ def _connection_kwargs_keys(self) -> t.Set[str]:
 
     @property
     def _engine_adapter(self) -> t.Type[EngineAdapter]:
-        # DB2 adapter requires Python 3.10+ for db2-sqlglot-dialect
-        # Use getattr to avoid mypy errors on Python 3.9
-        return t.cast(
-            t.Type[EngineAdapter], getattr(engine_adapter, "Db2EngineAdapter", EngineAdapter)
-        )
+        adapter = getattr(engine_adapter, "Db2EngineAdapter", None)
+        if adapter is None:
+            raise ConfigError(
+                "The Db2 engine adapter requires Python 3.10+ and the 'db2' extra. "
+                "Please install it with: pip install 'sqlmesh[db2]'"
+            )
+        return t.cast(t.Type[EngineAdapter], adapter)
 
     def get_catalog(self) -> t.Optional[str]:
         """Db2 stores catalog names in uppercase; normalise here so the default_catalog

sqlmesh/core/engine_adapter/db2.py

@@ -65,21 +65,21 @@ class Db2EngineAdapter(
     SCHEMA_DIFFER_KWARGS = {
         "parameterized_type_defaults": {
             # DECIMAL without precision defaults to (5, 0)
-            exp.DataType.build("DECIMAL", dialect=DIALECT).this: [(5, 0), (0,)],
+            exp.DataType.Type.DECIMAL: [(5, 0), (0,)],
             # CHAR without length defaults to 1
-            exp.DataType.build("CHAR", dialect=DIALECT).this: [(1,)],
+            exp.DataType.Type.CHAR: [(1,)],
             # VARCHAR without length defaults to 1
-            exp.DataType.build("VARCHAR", dialect=DIALECT).this: [(1,)],
+            exp.DataType.Type.VARCHAR: [(1,)],
             # TIMESTAMP defaults to 6 digits of fractional seconds
-            exp.DataType.build("TIMESTAMP", dialect=DIALECT).this: [(6,)],
+            exp.DataType.Type.TIMESTAMP: [(6,)],
             # TIME defaults to 0 digits of fractional seconds
-            exp.DataType.build("TIME", dialect=DIALECT).this: [(0,)],
+            exp.DataType.Type.TIME: [(0,)],
         },
         "types_with_unlimited_length": {
             # CLOB can be used for unlimited text
-            exp.DataType.build("CLOB", dialect=DIALECT).this: {
-                exp.DataType.build("VARCHAR", dialect=DIALECT).this,
-                exp.DataType.build("CHAR", dialect=DIALECT).this,
+            exp.DataType.Type.TEXT: {
+                exp.DataType.Type.VARCHAR,
+                exp.DataType.Type.CHAR,
             },
         },
         "drop_cascade": False,

sqlmesh/utils/migration.py

@@ -4,7 +4,6 @@
 MAX_TEXT_INDEX_LENGTH = {
     "mysql": "250",  # 250 characters per column, <= 767 byte index size limit
     "tsql": "450",  # 450 bytes per column, <= 900 byte index size limit
-    "db2": "255",  # Db2 has strict primary key size limits, keep it conservative
 }
 
 
@@ -24,8 +23,4 @@ def index_text_type(dialect: DialectType) -> str:
 
 
 def blob_text_type(dialect: DialectType) -> str:
-    if dialect == "mysql":
-        return "LONGTEXT"
-    if dialect == "db2":
-        return "VARCHAR(32000)"
-    return "TEXT"
+    return "LONGTEXT" if dialect == "mysql" else "TEXT"

tests/core/engine_adapter/integration/config.yaml

@@ -200,20 +200,6 @@ gateways:
     state_connection:
       type: duckdb
 
-  # inttest_db2:
-  #   connection:
-  #     type: db2
-  #     host: {{ env_var('DB2_HOST') }}
-  #     port: {{ env_var('DB2_PORT', '50000') }}
-  #     database: {{ env_var('DB2_DATABASE') }}
-  #     username: {{ env_var('DB2_USERNAME') }}
-  #     password: {{ env_var('DB2_PASSWORD') }}
-  #     # db2_schema sets CURRENTSCHEMA on the connection — controls the default schema
-  #     # for unqualified references. The test framework always uses fully-qualified names
-  #     # so any valid schema the user has access to works here (e.g. the username itself,
-  #     # which is the Db2 default when no schema is specified).
-  #     db2_schema: {{ env_var('DB2_SCHEMA', env_var('DB2_USERNAME')) }}
-  #     check_import: false
   inttest_db2:
     connection:
       type: db2

tests/core/engine_adapter/integration/test_integration.py

@@ -93,8 +93,6 @@ def dev_table_name_for(self, snapshot: Snapshot) -> str:
 
 def test_connection(ctx: TestContext):
     cursor_from_connection = ctx.engine_adapter.connection.cursor()
-    # cursor_from_connection.execute("SELECT 1")  # fails on Db2 — bare SELECT 1 raises SQL0104N
-    # Fix: use dialect-aware SQL so Db2 generates SELECT 1 FROM SYSIBM.SYSDUMMY1
     cursor_from_connection.execute(exp.select("1").sql(dialect=ctx.dialect))
     assert cursor_from_connection.fetchone()[0] == 1
 

tests/core/test_dialect.py

@@ -1,3 +1,4 @@
+import importlib.util
 import sys
 import pytest
 from sqlglot import Dialect, ParseError, exp, parse_one
@@ -1051,9 +1052,11 @@ def test_parse_snowflake_create_schema_ddl():
 
 @pytest.mark.parametrize("dialect", sorted(set(DIALECT_TO_TYPE.values())))
 def test_sqlglot_extended_correctly(dialect: str) -> None:
-    # Skip Db2 on Python 3.9 since db2-sqlglot-dialect requires Python 3.10+
-    if dialect == "db2" and sys.version_info < (3, 10):
-        pytest.skip("DB2 dialect requires Python 3.10+ for db2-sqlglot-dialect")
+    # Skip Db2 if db2-sqlglot-dialect is not installed
+    if dialect == "db2" and (
+        sys.version_info < (3, 10) or importlib.util.find_spec("db2_sqlglot") is None
+    ):
+        pytest.skip("db2-sqlglot-dialect is not installed")
 
     # MODEL is a SQLMesh extension and not part of SQLGlot
     # If we can roundtrip an expression containing MODEL across every dialect, then the SQLMesh extensions have been registered correctly