SNOW-638845 ValueError: unsupported format character with version 1.4.0 (#355)

Author: sfc-gh-aling

DESCRIPTION.md

@@ -9,6 +9,10 @@ Source code is also available at:
 
 # Release Notes
 
+- v1.4.4(Unreleased)
+
+  - Fixed a bug that percent signs in a non-compiled statement should not be interpolated with emtpy sequence when executed.
+
 - v1.4.3(Oct 17, 2022)
 
   - Fixed a bug that `SnowflakeDialect.normalize_name` and `SnowflakeDialect.denormalize_name` could not handle empty string.

setup.cfg

@@ -3,7 +3,7 @@ universal = 1
 
 [metadata]
 name = snowflake-sqlalchemy
-version = 1.4.3
+version = 1.4.4
 description = Snowflake SQLAlchemy Dialect
 long_description = file: DESCRIPTION.md
 long_description_content_type = text/markdown

src/snowflake/sqlalchemy/base.py

@@ -13,6 +13,7 @@
 from sqlalchemy.util.compat import string_types
 
 from .custom_commands import AWSBucket, AzureContainer, ExternalStage
+from .util import _set_connection_interpolate_empty_sequences
 
 RESERVED_WORDS = frozenset(
     [
@@ -416,34 +417,23 @@ def should_autocommit(self):
     def pre_exec(self):
         if self.compiled and self.identifier_preparer._double_percents:
             # for compiled statements, percent is doubled for escape, we turn on _interpolate_empty_sequences
-            if hasattr(self._dbapi_connection, "driver_connection"):
-                # _dbapi_connection is a _ConnectionFairy which proxies raw SnowflakeConnection
-                self._dbapi_connection.driver_connection._interpolate_empty_sequences = (
-                    True
-                )
-            else:
-                # _dbapi_connection is a raw SnowflakeConnection
-                self._dbapi_connection._interpolate_empty_sequences = True
+            _set_connection_interpolate_empty_sequences(self._dbapi_connection, True)
 
             # if the statement is executemany insert, setting _interpolate_empty_sequences to True is not enough,
             # because executemany pre-processes the param binding and then pass None params to execute so
             # _interpolate_empty_sequences condition not getting met for the command.
             # Therefore, we manually revert the escape percent in the command here
             if self.executemany and self.INSERT_SQL_RE.match(self.statement):
                 self.statement = self.statement.replace("%%", "%")
+        else:
+            # for other cases, do no interpolate empty sequences as "%" is not double escaped
+            _set_connection_interpolate_empty_sequences(self._dbapi_connection, False)
 
     def post_exec(self):
         if self.compiled and self.identifier_preparer._double_percents:
             # for compiled statements, percent is doubled for escapeafter execution
             # we reset _interpolate_empty_sequences to false which is turned on in pre_exec
-            if hasattr(self._dbapi_connection, "driver_connection"):
-                # _dbapi_connection is a _ConnectionFairy which proxies raw SnowflakeConnection
-                self._dbapi_connection.driver_connection._interpolate_empty_sequences = (
-                    False
-                )
-            else:
-                # _dbapi_connection is a raw SnowflakeConnection
-                self._dbapi_connection._interpolate_empty_sequences = False
+            _set_connection_interpolate_empty_sequences(self._dbapi_connection, False)
 
     @property
     def rowcount(self):

src/snowflake/sqlalchemy/util.py

@@ -8,6 +8,7 @@
 from sqlalchemy import exc
 
 from snowflake.connector.compat import IS_STR
+from snowflake.connector.connection import SnowflakeConnection
 
 
 def _rfc_1738_quote(text):
@@ -72,3 +73,15 @@ def sep(is_first_parameter):
             ret += sep(is_first_parameter) + p + "=" + encoded_value
             is_first_parameter = False
     return ret
+
+
+def _set_connection_interpolate_empty_sequences(
+    dbapi_connection: SnowflakeConnection, flag: bool
+) -> None:
+    """set the _interpolate_empty_sequences config of the underlying connection"""
+    if hasattr(dbapi_connection, "driver_connection"):
+        # _dbapi_connection is a _ConnectionFairy which proxies raw SnowflakeConnection
+        dbapi_connection.driver_connection._interpolate_empty_sequences = flag
+    else:
+        # _dbapi_connection is a raw SnowflakeConnection
+        dbapi_connection._interpolate_empty_sequences = flag

tests/test_pandas.py

@@ -411,3 +411,38 @@ def test_pandas_invalid_make_pd_writer(engine_testaccount):
             index=False,
             method=make_pd_writer(df=test_df),
         )
+
+
+def test_percent_signs(engine_testaccount, run_v20_sqlalchemy):
+    if run_v20_sqlalchemy and sys.version_info < (3, 8):
+        pytest.skip(
+            "In Python 3.7, this test depends on pandas features of which the implementation is incompatible with sqlachemy 2.0, and pandas does not support Python 3.7 anymore."
+        )
+
+    table_name = f"test_table_{uuid.uuid4().hex}".upper()
+    with engine_testaccount.connect() as conn:
+        with conn.begin():
+            conn.exec_driver_sql(
+                f"CREATE OR REPLACE TEMP TABLE {table_name}(c1 int, c2 string)"
+            )
+            conn.exec_driver_sql(
+                f"""
+                INSERT INTO {table_name}(c1, c2) values
+                (1, 'abc'),
+                (2, 'def'),
+                (3, 'ghi')
+                """
+            )
+
+            df = pd.read_sql(
+                f"select * from {table_name} where c2 not like '%b%'", conn
+            )
+            assert list(df.itertuples(index=False, name=None)) == [
+                (2, "def"),
+                (3, "ghi"),
+            ]
+
+            df = pd.read_sql(f"select * from {table_name} where c2 like '%b%'", conn)
+            assert list(df.itertuples(index=False, name=None)) == [
+                (1, "abc"),
+            ]