NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI
PyPI, #749 most downloaded on PyPI
Snowflake SQLAlchemy Dialect
Latest stable
2.1.0
Last release 27 days ago
26 Oct 2026
Release timing varies
gaps range from 8 days to 3 months
Some releases are documented
notes for 27 of 57 stable releases
4 versions withdrawn
withdrawn after publishing
10 years old
65 releases · first in 2017
Drag across the ruler to compare, or tap a release to open it.
Snowflake custom types declare SQLAlchemy 2.1 operator_classes, so expressions on them no longer emit SADeprecationWarning ("does not refer to an Oper…
SQLAlchemy>=2.0.44,<2.2).
"..." are part of the name, not a delimiter. For hand-written URLs, a + in the path is no longer read as a space, and an encoded %2F outside quotes now separates the database from the schema (previously it stayed in the database name). See the migration guide.snowflake.sqlalchemy.URL round-trip unchanged on SQLAlchemy 2.0 and 2.1, including a database-only quoted name with special characters (previously left percent-encoded). On SQLAlchemy 2.1, URL.create(...) values are no longer percent-decoded; on 2.0, a database or schema name containing % and no raw " is still decoded. See Special characters in database and schema names.NUMBER(..., 0) / INTEGER columns report SQLAlchemy's generic Integer type affinity.ROWS/RANGE bounds are compiled as integer literals (no bind-parameter conflicts).operator_classes, so expressions on them no longer emit SADeprecationWarning ("does not refer to an OperatorClass").timedelta literals render as Snowflake INTERVAL '...' constants, so SQLAlchemy 2.1 FrameClause RANGE bounds such as timedelta(days=7) compile to RANGE BETWEEN INTERVAL '7 DAYS' PRECEDING; literal_binds date arithmetic no longer raises CompileError. Negative intervals raise a CompileError (Snowflake interval constants are unsigned). Bound (non-literal) timedelta parameters are unchanged.Delete.using() joins that contain the DELETE target table raise a CompileError with guidance (Snowflake rejects them as "duplicate alias"); join an alias of the target or use a subquery instead.snowflake-sqlalchemy[async] (or sqlalchemy[asyncio]) for create_async_engine(). SQLAlchemy 2.1 no longer bundles greenlet; without the extra the install succeeds but async engine creation raises an ImportError with the install command.snowflake-connector-python >=5.1.0,<6.0Your coding agent can read these notes before it upgrades. Set up the MCP server →
…to False restores raw JSON strings and emits a DeprecationWarning (SNOW-3942921).
SQLAlchemy>=2.0.44,<2.1. SQLAlchemy 2.1.0 removed sqlalchemy.orm.context.ORMSelectCompileState and changed other APIs the dialect and async shim import at module load time, so 2.1 is excluded. The internal snowflake.sqlalchemy.compat module (SQLAlchemy 1.4/2.0 shim) was removed. Users still on SQLAlchemy 1.4 should pin snowflake-sqlalchemy<2.0.0.snowflake-connector-python>=5.0.0 (connector 5.x), which provides the async driver. Connector 4.x is no longer supported.requires-python, Python classifiers, and mypy/ruff targets accordingly.enable_structured_type_json is on by default (see Features); explicitly setting it to False restores raw JSON strings and emits a DeprecationWarning (SNOW-3942921).force_div_is_floordiv dialect flag. The / operator always performs true division (left / right) and // always performs floor division (FLOOR(left / right)). Passing the flag to create_engine() now raises TypeError. Coming from a release older than 1.11.1, integer // previously compiled to plain / (true division in Snowflake) under the default flag value; it now floors (GH #756).legacy_url_params opt-out and its SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS environment variable. Passing it to create_engine() now raises TypeError and the environment variable has no effect; sensitive connection parameters must be passed via connect_args= (SNOW-3942921). unsafe_skip_config_file_permissions_check and crl_unsafe_skip_file_permissions_check are now also blocked from the URL query string.asyncio extension. Use create_async_engine() with the same snowflake:// URL as the sync dialect — there is no separate +driver suffix — to obtain AsyncEngine / AsyncSession for non-blocking database access. The async dialect checks for greenlet when the engine is created and raises an ImportError with install instructions.enable_structured_type_json dialect flag (URL parameter or create_engine(..., enable_structured_type_json=...), on by default). For semi-structured VARIANT, OBJECT and ARRAY columns, reading deserializes the JSON text Snowflake returns into native Python (dict/list/...), and writing native dict/list values serializes them and wraps them in PARSE_JSON — rendered as INSERT ... SELECT (single row and multi-row UNION ALL) since Snowflake rejects functions in a VALUES clause, and as SET col = PARSE_JSON(...) for UPDATE. Typed structured columns (MAP(...), ARRAY(<type>), OBJECT(...) with fields) are covered too: writes render CAST(PARSE_JSON(...) AS <declared type>) and reads return dict/list. Element values are not converted to the declared types: they come back as JSON decodes them (for example MAP(NUMBER, ...) keys stay strings and DATE elements stay ISO strings). The engine's json_serializer/json_deserializer are used when provided (SNOW-1820647 / GH #107).VARIANT and structured types. Structured type compilers now compile their inner types against the Snowflake dialect, so ARRAY/MAP/OBJECT element, key and field types render consistently with standalone columns. The semi-structured JSON write path serializes decimal.Decimal as a JSON string (preserving all 38 DECFLOAT digits, which a JSON number would truncate); reading one back requires an intermediate ::VARCHAR cast, since Snowflake rejects a direct TEXT -> DECFLOAT cast on PARSE_JSON output (SNOW-4081644).enable_native_uuid dialect flag (URL parameter or create_engine(..., enable_native_uuid=True)), which renders SQLAlchemy's generic Uuid and Mapped[uuid.UUID] as native Snowflake UUID instead of CHAR(32) and allows uuid.UUID inside semi-structured payloads. uuid.UUID values are bound as strings and reads honour as_uuid (True returns uuid.UUID, False returns str). Off by default, since enabling it changes the DDL of existing Uuid columns (SNOW-4081644).strict_host_names / SNOWFLAKE_SQLALCHEMY_STRICT_HOST_NAMES): the dialect no longer infers or appends domains to host names and forwards connection parameters, including connection_name profiles, to the connector unvalidated. Explicit stage endpoints are supported (SNOW-4248737).COLLATE support: expression-level Column.collate('<spec>') and column-type collation (String(collation='<spec>')) now render the collation as a single-quoted Snowflake string literal (e.g. COLLATE 'en-ci') instead of an invalid double-quoted identifier. Enables the SQLAlchemy suite CollateTest (SNOW-629086).col["key"] and col[index] on VARIANT, OBJECT, ARRAY and MAP columns now compile to Snowflake's native bracket accessor (e.g. col['key'], col[0]), including nested chains such as col["a"]["b"]. See Reading keys and elements with subscript access (SNOW-1820647 / GH #546).INSERT ALL / INSERT FIRST statements via the new InsertMulti custom command, including conditional (when/else_), OVERWRITE, and per-target column/value mapping (SNOW-782235 / GH #403).SnowflakeTable via the row_access_policy parameter; the policy name accepts a string (optionally fully qualified, e.g. 'db.schema.policy') or a snowflake.sqlalchemy.FQN. See Row Access Policy Support (SNOW-884382 / GH #432).snowflake_rely=True to render RELY on PRIMARY KEY / FOREIGN KEY / UNIQUE constraints so the optimizer can trust them for query rewrites such as join elimination (SNOW-1023317 / GH #463).json_serializer and json_deserializer parameters in create_engine, matching the built-in SQLAlchemy dialects (SNOW-889293 / GH #433).asdecimal parameter on _CUSTOM_DECIMAL type to allow opt-in conversion of numeric results to float instead of Decimal; default remains Decimal for backward compatibility (SNOW-728151 / GH #373).python_type property to Snowflake custom types (VARIANT, OBJECT, MAP, ARRAY, VECTOR, TIMESTAMP_*, GEOGRAPHY, GEOMETRY, DECFLOAT) for SQLAlchemy compatibility (SNOW-1866493 / GH #562).JSONFormatter with the full Snowflake TYPE=JSON option set (date_format, time_format, timestamp_format, binary_format, trim_space, null_if, enable_octal, allow_duplicate, strip_outer_array, strip_null_values, replace_invalid_characters, ignore_utf8_errors, skip_byte_order_mark) (SNOW-589946).if_not_exists and comment options to CreateFileFormat for closer parity with Snowflake's CREATE FILE FORMAT (SNOW-589962 / GH #291).enable_server_session_keep_alive_auto_detection=True explicitly on new connections. This pins the connector's present default so behaviour stays stable when that default changes, and silences the connector's FutureWarning. A value supplied via the URL or connect_args still takes precedence.COPY INTO and staged-file SQL (CWE-89): ExternalStage paths, including sub-paths created with from_parent_stage, are rendered safely with escaping when quoting is required. Ordinary bare-path values retain their existing rendering (SNOW-4134196).repr() of CopyIntoStorage unloading into an ExternalStage or cloud location (AWSBucket, AzureContainer, GCSBucket), which raised UnsupportedCompilationError; the target now uses its own secret-masking repr().None) in get_multi_indexes result keys so indexes attach when reflecting under the default schema, and return [] instead of [''] for empty INCLUDE columns (GH #755, fixes GH #754).get_multi_indexes() raising TypeError when filter_names is None (e.g. Inspector.get_multi_indexes(schema=...)) and the schema contains Hybrid Tables; it now returns indexes for all Hybrid Tables in the schema.SHOW commands (get_table_names, get_view_names, get_temp_table_names, get_schema_names, get_sequence_names) with LIMIT ... FROM ... (SNOW-796954 / GH #406).gnamsrm-vi65876.privatelink) were incorrectly truncated at the dash, breaking JWT/key-pair authentication. Account derivation now delegates to the Snowflake Connector's own parse_account, so all host notations (plain, regional, regional/regionless PrivateLink, and .global) match the driver (SNOW-730644, also fixes SNOW-1209099).URL failing to encode [, ], ? and # in passwords, which produced connect strings that could not be parsed (SNOW-828206 / GH #415).LIMIT -1 SQL generated for a SELECT with OFFSET but no LIMIT; the SnowflakeCompiler now emits LIMIT NULL OFFSET ... as Snowflake requires (NO-SNOW / GH #745).390195 (authentication token expired, variant), as disconnects so connection_invalidated is set and pooled connections are recycled by pool_pre_ping (SNOW-669163 / GH #348, SNOW-3882904 / GH #702).authenticator connect_args parameter (SNOW-715550).executemany inserts via qmark paramstyle (SNOW-710474).VALUES source with MergeInto (no staging table needed) (SNOW-889678 / GH #435).PUT with an in-memory file_stream (BytesIO) (SNOW-645168 / #337).CLUSTER BY usage in ORM/declarative models via SnowflakeTable (SNOW-638838 / #313).IS TRUE/IS FALSE; use col == true()/col == false() (#680).Pin SQLAlchemy<2.1 (floor remains >=2.0.44). SQLAlchemy 2.1.0 removed sqlalchemy.orm.context.ORMSelectCompileState and changed other APIs the dialect
SQLAlchemy<2.1 (floor remains >=2.0.44). SQLAlchemy 2.1.0 removed sqlalchemy.orm.context.ORMSelectCompileState and changed other APIs the dialect and async shim import at module load time, which broke import snowflake.sqlalchemy outright under the previously unbounded dependency range.COPY INTO and staged-file SQL (CWE-89): ExternalStage paths, including sub-paths created with from_parent_stage, are rendered safely with escaping when quoting is required. Ordinary bare-path values retain their existing rendering (SNOW-4134196).VARIANT and structured types. Structured type compilers now compile their inner types against the Snowflake dialect, so ARRAY/MAP/OBJECT element, key and field types render consistently with standalone columns. The semi-structured JSON write path serializes decimal.Decimal as a JSON string (preserving all 38 DECFLOAT digits, which a JSON number would truncate); reading one back requires an intermediate ::VARCHAR cast, since Snowflake rejects a direct TEXT -> DECFLOAT cast on PARSE_JSON output. New opt-in enable_native_uuid dialect flag (URL parameter or create_engine(..., enable_native_uuid=True)) renders SQLAlchemy's generic Uuid and Mapped[uuid.UUID] as native Snowflake UUID instead of CHAR(32), and allows uuid.UUID inside semi-structured payloads. Off by default, since enabling it changes the DDL of existing Uuid columns (SNOW-4081644).…the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing sp…
v2.0.0rc1 (Sep 14, 2026)
Async support: the dialect now implements SQLAlchemy's asyncio extension. Use create_async_engine() with the same snowflake:// URL as the sync dialect — there is no separate +driver suffix — to obtain AsyncEngine / AsyncSession for non-blocking database access.
Breaking change: require snowflake-connector-python>=5.0 (connector 5.x), which is what provides the async driver. Connector 4.x is no longer supported.
Breaking change: raise the minimum supported Python to 3.11 (drop Python 3.10), since snowflake-connector-python>=5.0.0rc3 itself requires Python 3.11+ (dropped 3.10). Updates requires-python, Python classifiers, and mypy/ruff targets accordingly.
The dialect now sets enable_server_session_keep_alive_auto_detection=True explicitly on new connections. This pins the connector's present default so behaviour stays stable when that default changes, and silences the connector's FutureWarning. A value supplied via the URL or connect_args still takes precedence.
Remove the force_div_is_floordiv dialect flag entirely. The / operator always performs true division (left / right) and // always performs floor division (FLOOR(left / right)); the flag had no SQL effect. Passing it to create_engine() now raises ArgumentError (GH #756).
v2.0.0a2 (Aug 20, 2026)
requires-python, the CI/build matrices, and mypy/ruff targets accordingly.enable_structured_type_json now defaults to True (was False); explicitly setting it to False emits a DeprecationWarning (SNOW-3942921).force_div_is_floordiv now defaults to False (was True); explicitly setting it to True emits a DeprecationWarning (SNOW-3942921).legacy_url_params (and its SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS environment variable) removed; passing it now raises ArgumentError (SNOW-3942921).None) in get_multi_indexes result keys so indexes attach when reflecting under the default schema, and return [] instead of [''] for empty INCLUDE columns (GH #755, fixes GH #754).COLLATE support: expression-level Column.collate('<spec>') and column-type collation (String(collation='<spec>')) now render the collation as a single-quoted Snowflake string literal (e.g. COLLATE 'en-ci') instead of an invalid double-quoted identifier. Enables the SQLAlchemy suite CollateTest (SNOW-629086).col["key"] and col[index] on VARIANT, OBJECT, ARRAY and MAP columns now compile to Snowflake's native bracket accessor (e.g. col['key'], col[0]), including nested chains such as col["a"]["b"]. See Reading keys and elements with subscript access (SNOW-1820647 / GH #546).enable_structured_type_json dialect flag (URL parameter or create_engine(..., enable_structured_type_json=True)). When enabled, for semi-structured (untyped) VARIANT, OBJECT and ARRAY columns: reading deserializes the JSON text Snowflake returns into native Python (dict/list/...), and writing native dict/list values serializes them and wraps them in PARSE_JSON — rendered as INSERT ... SELECT (single row and multi-row UNION ALL) since Snowflake rejects functions in a VALUES clause, and as SET col = PARSE_JSON(...) for UPDATE. Typed/structured columns (OBJECT(...) with fields, ARRAY(<type>), MAP) keep their native connector handling. The engine's json_serializer/json_deserializer are used when provided. Off by default, so existing code that reads raw JSON strings or writes pre-serialized values is unchanged (SNOW-1820647 / GH #107).390195 (authentication token expired, variant) as a session disconnect in is_disconnect, so pool_pre_ping / connection_invalidated recover from it like the other token/session-loss codes (SNOW-3882904 / GH #702).asdecimal parameter on _CUSTOM_DECIMAL type to allow opt-in conversion of numeric results to float instead of Decimal; default remains Decimal for backward compatibility (SNOW-728151 / GH #373).gnamsrm-vi65876.privatelink) were incorrectly truncated at the dash, breaking JWT/key-pair authentication. Account derivation now delegates to the Snowflake Connector's own parse_account, so all host notations (plain, regional, regional/regionless PrivateLink, and .global) match the driver (SNOW-730644, also fixes SNOW-1209099).INSERT ALL / INSERT FIRST statements via the new InsertMulti custom command, including conditional (when/else_), OVERWRITE, and per-target column/value mapping (SNOW-782235 / GH #403).SnowflakeTable via the row_access_policy parameter; the policy name accepts a string (optionally fully qualified, e.g. 'db.schema.policy') or a snowflake.sqlalchemy.FQN. See Row Access Policy Support (SNOW-884382 / GH #432).SHOW commands (get_table_names, get_view_names, get_temp_table_names, get_schema_names, get_sequence_names) with LIMIT ... FROM ... (SNOW-796954 / GH #406).snowflake_rely=True to render RELY on PRIMARY KEY / FOREIGN KEY / UNIQUE constraints so the optimizer can trust them for query rewrites such as join elimination (SNOW-1023317 / GH #463).python_type property to Snowflake custom types (VARIANT, OBJECT, MAP, ARRAY, VECTOR, TIMESTAMP_*, GEOGRAPHY, GEOMETRY, DECFLOAT) for SQLAlchemy compatibility (SNOW-1866493 / GH #562).URL failing to encode [, ], ? and # in passwords, which produced connect strings that could not be parsed (SNOW-828206 / GH #415).JSONFormatter with the full Snowflake TYPE=JSON option set (date_format, time_format, timestamp_format, binary_format, trim_space, null_if, enable_octal, allow_duplicate, strip_outer_array, strip_null_values, replace_invalid_characters, ignore_utf8_errors, skip_byte_order_mark) (SNOW-589946).json_serializer and json_deserializer parameters in create_engine, matching the built-in SQLAlchemy dialects (SNOW-889293 / GH #433).connection_invalidated is set and pooled connections are recycled (SNOW-669163 / GH #348).if_not_exists and comment options to CreateFileFormat for closer parity with Snowflake's CREATE FILE FORMAT (SNOW-589962 / GH #291).authenticator connect_args parameter (SNOW-715550).executemany inserts via qmark paramstyle (SNOW-710474).VALUES source with MergeInto (no staging table needed) (SNOW-889678 / GH #435).INSERT ... SELECT with PARSE_JSON (SNOW-801402 / GH #411).PUT with an in-memory file_stream (BytesIO) (SNOW-645168 / #337).CLUSTER BY usage in ORM/declarative models via SnowflakeTable (SNOW-638838 / #313).IS TRUE/IS FALSE; use col == true()/col == false() (#680).LIMIT -1 SQL generated for a SELECT with OFFSET but no LIMIT; the SnowflakeCompiler now emits LIMIT NULL OFFSET ... as Snowflake requires (NO-SNOW / GH #745).v2.0.0a2 (Unreleased)
v2.0.0a0 (July 16, 2026)
SQLAlchemy>=2.0.0. Users still on SQLAlchemy 1.4 should pin to snowflake-sqlalchemy<2.0.0.>=3.9, <=3.14.regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regexp_match(..., flags=...) and ColumnElement.regexp_replace(..., flags=...) were processed through the standard parameter pipeline, producing incorrect SQL. Flags are now rendered as inline string literals, matching Snowflake's expected REGEXP_LIKE(col, pattern, 'i') / REGEXP_REPLACE(col, pattern, replacement, 'i') syntax.v1.11.0 (July 7, 2026)
Sensitive connection parameters: We curated a set of connector kwargs (host, protocol, token_file_path, private_key_file, ocsp_response_cache_filename, connection_diag_log_path, crl_cache_dir, unsafe_file_write, unsafe_skip_file_permissions_check) that can no longer be supplied via the URL query string — pass them via connect_args= in create_engine() instead (applications already doing so are unaffected). If you encounter a possible behavioral change, set SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS=1 and follow the instructions at Sensitive connection parameters.
Improve _url() helper and create_connect_args connection-parameter handling:
account and region values are now validated against an allowlist of DNS-safe characters (alphanumeric, -, ., _) before being interpolated into the connection URL, preventing URL-authority corruption from unexpected characters.user values are percent-encoded before being placed in the URL userinfo component, preserving the original value delivered to the connector while preventing @, ?, and # from being misinterpreted as URL delimiters.ClusterByOption now raises TypeError at DDL compile time when an expression element is neither a str nor a sqlalchemy.sql.expression.TextClause. Previously, such values were silently coerced via str(), which produced malformed DDL (e.g. bind-parameter placeholders like :id_1) for any expression beyond a bare column name. The accepted types match the documented constructor signature; code using only str or text(...) is unaffected.
Improve identifier quoting and string-literal escaping for caller-supplied values across the DDL compiler so they are rendered consistently with the rest of the dialect. Column keys in MERGE INTO (both the WHEN NOT MATCHED … INSERT list and WHEN MATCHED … SET targets), stage namespaces/names, format_name, and file_format are now routed through the identifier preparer, and a quote=False schema or column label is quoted when it contains characters that would otherwise require quoting instead of being emitted verbatim. Cloud-storage URIs, CREDENTIALS, and ENCRYPTION clauses (shared by COPY INTO and CREATE STAGE) and FILES=(…) entries now apply the dialect's standard literal escaping ('→'', \→\\); CREATE STAGE builds its SQL from the container's fields rather than repr(container). Legal identifiers — including upper-case names that Snowflake folds — still render bare, so existing DDL is unchanged.
Improve escaping in CopyIntoStorage / CreateStage / CreateFileFormat and add secret redaction (SNOW-3656048):
FILE_FORMAT option string values (e.g. CSVFormatter().date_format(...), file_extension, timestamp_format) are now escaped before being embedded in the compiled SQL, improving single-quote escaping (SNOW-3649888).repr() of AWSBucket, AzureContainer, GCSBucket (and, transitively, CopyIntoStorage) now masks cloud secrets — AWS_SECRET_KEY, AWS_KEY_ID, AWS_TOKEN, AZURE_SAS_TOKEN, MASTER_KEY are rendered as '***' (SNOW-3649782). Compiled SQL is unchanged.SnowflakeSecretRedactionFilter, add_secret_redaction_filter(), and redact_secrets(). Prefer STORAGE_INTEGRATION to avoid putting secrets in SQL at all (SNOW-3649850).Fix get_view_definition silently truncating or failing for view names that contain a single quote or backslash (e.g. o'brien). The name is now SQL-escaped (' → '', \ → \\) before being embedded in the SHOW VIEWS LIKE '...' literal, so such views are found correctly and no query manipulation is possible via a crafted view name.
Optimise single-table reflection (SNOW-3720548): reflecting one table no longer scans the entire schema, reducing latency and Snowflake credit usage for targeted Inspector calls.
v1.10.2 (June 18, 2026)
Fix double-escaped identifier quoting for database-qualified schemas in _StructuredTypeInfoManager.get_table_columns. Since v1.10.1 the schema was quoted as a single identifier, so a qualified schema such as "MYDB"."MYSCHEMA" became """MYDB"".""MYSCHEMA""" and the DESC TABLE fallback failed (emitting a Failed to reflect table ... sqlalchemy:_get_schema_columns warning) for every structured-typed table when reflecting a non-default database.schema. The schema is now split on the dot and each component is double-quoted individually, preserving the SNOW-3480955 injection guard while correctly handling qualified schemas.
v1.10.1 (June 15, 2026)
regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regexp_match(..., flags=...) and ColumnElement.regexp_replace(..., flags=...) were processed through the standard parameter pipeline, producing incorrect SQL. Flags are now rendered as inline string literals, matching Snowflake's expected REGEXP_LIKE(col, pattern, 'i') / REGEXP_REPLACE(col, pattern, replacement, 'i') syntax._StructuredTypeInfoManager.get_table_columns. The DESC TABLE fallback path used raw denormalised names in an f-string while all other reflection paths apply ip.quote(denormalize_name(...)) via _always_quote_join. Schema and table components are now consistently double-quoted before the statement is constructed, and the method delegates to get_table_columns_by_full_name to collapse the two previously divergent code paths.v1.10.0 (May 20, 2026)
with_loader_criteria silently dropping filters on non-Snowflake dialects (#676). Importing snowflake-sqlalchemy previously altered SQLAlchemy's ORM compilation for every dialect in the process, causing loader-criteria filters to be omitted inside sealed subqueries when using PostgreSQL, MySQL, SQLite, etc. Snowflake dialect behavior is unchanged; the BCR-1057 lateral-join workaround is now scoped to Snowflake connections only.UUID column type to sqlalchemy.sql.sqltypes.UUID for reflection on SQLAlchemy 2.x (#681). Previously reflected as NullType. Values are returned as plain strings (as_uuid=False) rather than uuid.UUID instances. No change on SQLAlchemy 1.4 where the generic UUID type does not exist.CopyIntoStorage (SNOW-721174, #368).referred_schema=None normalization in foreign key reflection to the default schema only (#610, SNOW-2313675):
referred_schema=None, preserving compatibility with the upstream reflection test suite and with applications that do not qualify default-schema FK targets.referred_schema, which prevents SQLAlchemy's _reflect_fk from autoloading a non-default-schema target from the wrong place (the bug behind #610) and avoids the Alembic autogenerate mismatch that previously occurred when user metadata explicitly qualified a cross-schema FK that happened to target the default schema.SnowflakeBase, snowflake_declarative_base(), and SnowflakeSession to enable efficient bulk inserts for ORM models with nullable optional columns (SNOW-893080, #441). When session.bulk_save_objects() is used with models that have randomly populated nullable columns, SQLAlchemy normally groups objects by their set of non-None column keys, producing O(N) separate INSERT statements. SnowflakeBase / snowflake_declarative_base() pre-populate all plain-nullable columns at construction time, and SnowflakeSession passes render_nulls=True so all objects share the same parameter-key set and are batched into a single executemany INSERT._split_schema_by_dot now correctly parses SQL-escaped double-quotes ("") inside quoted schema/database identifiers (e.g. "my""schema" → my"schema), preventing silent truncation of identifiers containing literal quote characters.denormalize_column_name now correctly double-quotes quoted_name("mycol", True) columns in CLUSTER BY clauses instead of silently dropping the case-sensitivity signal. The caller has already opted into case-sensitive semantics by constructing a quoted_name(..., True), so this is honoured independently of the dialect flag._has_object (used by has_table / has_sequence) now applies denormalize_name to both the schema and object name before building the DESC SQL, making it consistent with all other reflection methods._NameUtils swap in create_connect_args — when the URL's case_sensitive_identifiers value differs from the current dialect state, the name_utils instance is replaced rather than mutated in place, so concurrent readers on other threads observe either the old or the new instance but never a torn update.case_sensitive_identifiers opt-in engine flag (kwarg or ?case_sensitive_identifiers=True URL param) governing three related behaviours. The default is False; existing applications are unaffected unless they explicitly opt in:
TABLE) are normalised to quoted_name("table", True) instead of returning unchanged, preventing key-lookup mismatches between creation and reflection.MyCol from a quoted Snowflake column) are returned as quoted_name("MyCol", True) instead of a plain str. Emitted SQL is identical in both modes (_requires_quotes force-quotes any name containing uppercase chars); the difference is only observable via isinstance(..., quoted_name) and .quote.'"myschema"' or '"mydb"."myschema"' — have their extracted parts marked quote=True by _split_schema_by_dot, preserving case-sensitivity in emitted SQL. Without the flag, the extracted parts keep quote=None and the preparer's _requires_quotes heuristic decides per-part (stripping inner quotes for all-lowercase parts, which matches pre-PR behaviour). Use quoted_name("myschema", True) or MetaData(schema=quoted_name(..., True)) to opt into case-sensitivity on a per-value basis without enabling the flag.create_snowflake_engine(url, schema=..., case_sensitive_schema=True) helper that URL-encodes case-sensitive schema names using %22 so the Snowflake connector receives the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing special characters (?, #, /) from being misinterpreted as URL delimiters by SQLAlchemy's URL parser.snowflake.sqlalchemy.alembic_util.render_item — a drop-in Alembic render_item hook for env.py that serialises quoted_name columns with quote=True correctly in generated migration files, preventing Alembic autogenerate from silently converting case-sensitive column names to uppercase.SnowflakeWarning at DDL compile time when Identity() is used on a primary key column, alerting users that ORM flush operations will raise a FlushError. The warning is emitted once per unique (table, column) pair per Python process. Use Sequence() instead.schema='database.schema' notation. This allows reflecting and joining tables from different databases in a single session without raw SQL. (#456)/) when div_is_floordiv=True: the Snowflake compiler now correctly delegates to the SQLAlchemy base implementation, emitting CAST(col AS NUMERIC) for integer operands as it did before #545 introduced the override (#618).get_multi_columns, get_multi_pk_constraint, get_multi_unique_constraints, get_multi_foreign_keys for SQLAlchemy 2.x bulk reflection — each issues one schema-wide query per reflection pass instead of one query per table.get_columns now uses DESC TABLE directly (per-table, live) since get_multi_columns handles all bulk reflection; temporary tables and dynamic tables are reflected correctly without schema-wide queries.SHOW INDEXES IN TABLE replacing the previous SHOW TABLES LIKE approach for single-table index reflection, eliminating SQL LIKE wildcard false-positives and case-sensitivity bugs._always_quote_join helper that always quotes denormalised identifiers — ensures correct SQL for case-sensitive table and schema names in per-table reflection paths.referred_schema resolution so reflected FKs always keep their actual schema unless the target lives in the connection's default schema. Previously FKs whose target shared the reflected non-default schema were reported with referred_schema=None, which caused SQLAlchemy's _reflect_fk to autoload from the wrong schema and raise NoReferencedColumnError during Alembic autogenerate._parse_pk_rows, _parse_uk_rows, _parse_fk_rows) so correctness fixes propagate to both per-table and schema-wide reflection paths.cache_column_metadata=True opt-in enables per-table SHOW … IN TABLE queries for get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes on SQLAlchemy 1.4.get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes now automatically use per-table SHOW … IN TABLE queries without any opt-in flag. Previously these methods always issued SHOW … IN SCHEMA even for single-table Inspector calls (e.g. pandas.read_sql_table()), causing ~20-second delays on schemas with thousands of tables (SNOW-689531).v1.9.0 (March 4, 2026)
DECFLOAT and VECTOR data typesILIKE in queriesSYSDATE() renderingNone arguments in internal schema resolution (#623).SHOW TABLES returns empty string table names, causing IndexError during reflection (#296).always, cycle, order, etc.).pytest-xdist parallel test support via per-worker schema provisioning hooks.pandas lower bound in sa14 test environment from <2.1 to >=2.1.1,<2.2 to ensure pre-built wheels are available for Python 3.12v1.8.2 (December 9, 2025)
snowflake-connector-pythonv1.8.1 (December 9, 2025)
v1.8.0(December 5, 2025)
snowflake-connector-python<5.0.0v1.7.7(September 3, 2025)
v1.7.6(July 10, 2025)
v1.7.5(June 20, 2025)
v1.7.4(June 10, 2025)
v1.7.3(January 15, 2025)
_ as initial character.force_div_is_floordiv flag to override div_is_floordiv new default value False in SnowflakeDialect.
False, the / division operator will be treated as a float division and // as a floor division.div_is_floor_div as False.v1.7.2(December 18, 2024)
_ as column namev1.7.1(December 02, 2024)
v1.7.0(November 21, 2024)
v1.6.1(July 9, 2024)
v1.6.0(July 8, 2024)
hatch & uv for managing project virtual environmentsv1.5.4
v1.5.3(April 16, 2024)
v1.5.2(April 11, 2024)
v1.5.1(November 03, 2023)
externalbrowser authentication not caching due to incorrect parsing of boolean query parameters.
v1.5.0(Aug 23, 2023)
v1.4.7(Mar 22, 2023)
SnowflakeConnection to SnowflakeSQLAlchemy.SnowflakeDialect.get_columns now throws a NoSuchTableError exception when the specified table doesn't exist, instead of the more vague KeyError.sqlalchemy.func.now is not rendered correctly.v1.4.6(Feb 8, 2023)
v1.4.5(Dec 7, 2022)
SnowflakeConnection to SnowflakeSQLAlchemy.v1.4.4(Nov 16, 2022)
v1.4.3(Oct 17, 2022)
SnowflakeDialect.normalize_name and SnowflakeDialect.denormalize_name could not handle empty string.sqlalchemy.engine.url._rfc_1738_quote as it is removed from SQLAlchemy v1.4.42.v1.4.2(Sep 19, 2022)
v1.4.1(Aug 18, 2022)
DATE should not be removed from SnowflakeDialect.ischema_names.inspect.get_foreign_keys and inspect.get_pk_constraint.v1.4.0(July 20, 2022)
regexp_match, regexp_replace in sqlalchemy.sql.expression.ColumnOperators.Date, DateTime, Time, Float and Numeric, and converting the values into corresponding Python objects.get_sequence_names in SnowflakeDialect.v1.3.4(April 27,2022)
v1.3.3(December 19,2021)
v1.3.2 (September 14,2021)
v1.3.1 (July 23,2021)
v1.2.5 (July 20,2021)
v1.2.4 (October 05,2020)
v1.2.3 (March 30, 2020)
v1.2.2 (March 9, 2020)
v1.2.1 (February 18,2020)
v1.2.0 (January 27, 2020)
v1.1.18 (January 6,2020)
v1.1.17 (December 2,2019)
v1.1.16 (October 21,2019)
v1.1.15 (September 30, 2019)
v1.1.14 (August 12, 2019)
FIELD_DELIMETER spelled incorrectlyv1.1.13 (May 20,2019)
v1.1.12 (April 8,2019)
v1.1.11 (March 25, 2019)
v1.1.10 (February 22, 2019)
v1.1.9 (February 11, 2019)
v1.1.8 (February 8, 2019)
v1.1.7 (February 8, 2019)
v1.1.6 (January 3, 2019)
v1.1.5 (December 19, 2018)
v1.1.4 (November 13, 2018)
v1.1.3 (October 30, 2018)
v1.1.2 (June 7, 2018)
v1.1.1 (May 17, 2018)
region parameter support to URL utility method.v1.1.0 (February 1, 2018)
role example.get_pk_constraint and get_primary_keys. Those applications that depend on the old behaviors must update codes. Issue #38 (@nrth)open and close connections.v1.0.9 (January 4, 2018)
v1.0.8 (December 21, 2017)
get_schema_names method to Snowflake SQLAlchemy dialect. PR #20(andrewsali)v1.0.7 (May 18, 2017)
v1.0.6 (April 20, 2017)
v1.0.5 (April 13, 2017)
snowflake_clusterby option support to Table object so that the user can create a table with clustering keysv1.0.4 (March 9, 2017)
v1.0.3 (October 20, 2016)
VARIANT, OBJECT and ARRAY data type supports for fetchv1.0.2 (July 5, 2016)
v1.0.1 (July 4, 2016)
v1.0.0 (June 28, 2016)
Breaking change: raise the minimum supported Python to 3.10 (drop Python 3.9). Updates requires-python, the CI/build matrices, and mypy/ruff targets a…
v2.0.0a2 (Aug 20, 2026)
requires-python, the CI/build matrices, and mypy/ruff targets accordingly.enable_structured_type_json now defaults to True (was False); explicitly setting it to False emits a DeprecationWarning (SNOW-3942921).force_div_is_floordiv now defaults to False (was True); explicitly setting it to True emits a DeprecationWarning (SNOW-3942921).legacy_url_params (and its SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS environment variable) removed; passing it now raises ArgumentError (SNOW-3942921).None) in get_multi_indexes result keys so indexes attach when reflecting under the default schema, and return [] instead of [''] for empty INCLUDE columns (GH #755, fixes GH #754).COLLATE support: expression-level Column.collate('<spec>') and column-type collation (String(collation='<spec>')) now render the collation as a single-quoted Snowflake string literal (e.g. COLLATE 'en-ci') instead of an invalid double-quoted identifier. Enables the SQLAlchemy suite CollateTest (SNOW-629086).col["key"] and col[index] on VARIANT, OBJECT, ARRAY and MAP columns now compile to Snowflake's native bracket accessor (e.g. col['key'], col[0]), including nested chains such as col["a"]["b"]. See Reading keys and elements with subscript access (SNOW-1820647 / GH #546).enable_structured_type_json dialect flag (URL parameter or create_engine(..., enable_structured_type_json=True)). When enabled, for semi-structured (untyped) VARIANT, OBJECT and ARRAY columns: reading deserializes the JSON text Snowflake returns into native Python (dict/list/...), and writing native dict/list values serializes them and wraps them in PARSE_JSON — rendered as INSERT ... SELECT (single row and multi-row UNION ALL) since Snowflake rejects functions in a VALUES clause, and as SET col = PARSE_JSON(...) for UPDATE. Typed/structured columns (OBJECT(...) with fields, ARRAY(<type>), MAP) keep their native connector handling. The engine's json_serializer/json_deserializer are used when provided. Off by default, so existing code that reads raw JSON strings or writes pre-serialized values is unchanged (SNOW-1820647 / GH #107).390195 (authentication token expired, variant) as a session disconnect in is_disconnect, so pool_pre_ping / connection_invalidated recover from it like the other token/session-loss codes (SNOW-3882904 / GH #702).asdecimal parameter on _CUSTOM_DECIMAL type to allow opt-in conversion of numeric results to float instead of Decimal; default remains Decimal for backward compatibility (SNOW-728151 / GH #373).gnamsrm-vi65876.privatelink) were incorrectly truncated at the dash, breaking JWT/key-pair authentication. Account derivation now delegates to the Snowflake Connector's own parse_account, so all host notations (plain, regional, regional/regionless PrivateLink, and .global) match the driver (SNOW-730644, also fixes SNOW-1209099).INSERT ALL / INSERT FIRST statements via the new InsertMulti custom command, including conditional (when/else_), OVERWRITE, and per-target column/value mapping (SNOW-782235 / GH #403).SnowflakeTable via the row_access_policy parameter; the policy name accepts a string (optionally fully qualified, e.g. 'db.schema.policy') or a snowflake.sqlalchemy.FQN. See Row Access Policy Support (SNOW-884382 / GH #432).SHOW commands (get_table_names, get_view_names, get_temp_table_names, get_schema_names, get_sequence_names) with LIMIT ... FROM ... (SNOW-796954 / GH #406).snowflake_rely=True to render RELY on PRIMARY KEY / FOREIGN KEY / UNIQUE constraints so the optimizer can trust them for query rewrites such as join elimination (SNOW-1023317 / GH #463).python_type property to Snowflake custom types (VARIANT, OBJECT, MAP, ARRAY, VECTOR, TIMESTAMP_*, GEOGRAPHY, GEOMETRY, DECFLOAT) for SQLAlchemy compatibility (SNOW-1866493 / GH #562).URL failing to encode [, ], ? and # in passwords, which produced connect strings that could not be parsed (SNOW-828206 / GH #415).JSONFormatter with the full Snowflake TYPE=JSON option set (date_format, time_format, timestamp_format, binary_format, trim_space, null_if, enable_octal, allow_duplicate, strip_outer_array, strip_null_values, replace_invalid_characters, ignore_utf8_errors, skip_byte_order_mark) (SNOW-589946).json_serializer and json_deserializer parameters in create_engine, matching the built-in SQLAlchemy dialects (SNOW-889293 / GH #433).connection_invalidated is set and pooled connections are recycled (SNOW-669163 / GH #348).if_not_exists and comment options to CreateFileFormat for closer parity with Snowflake's CREATE FILE FORMAT (SNOW-589962 / GH #291).authenticator connect_args parameter (SNOW-715550).executemany inserts via qmark paramstyle (SNOW-710474).VALUES source with MergeInto (no staging table needed) (SNOW-889678 / GH #435).INSERT ... SELECT with PARSE_JSON (SNOW-801402 / GH #411).PUT with an in-memory file_stream (BytesIO) (SNOW-645168 / #337).CLUSTER BY usage in ORM/declarative models via SnowflakeTable (SNOW-638838 / #313).IS TRUE/IS FALSE; use col == true()/col == false() (#680).LIMIT -1 SQL generated for a SELECT with OFFSET but no LIMIT; the SnowflakeCompiler now emits LIMIT NULL OFFSET ... as Snowflake requires (NO-SNOW / GH #745).v2.0.0a2 (Unreleased)
Breaking change: drop SQLAlchemy 1.4 support. The dialect now requires SQLAlchemy>=2.0.0. Users still on SQLAlchemy 1.4 should pin to snowflake-sqlalc…
SQLAlchemy>=2.0.0. Users still on SQLAlchemy 1.4 should pin to snowflake-sqlalchemy<2.0.0.>=3.9, <=3.14.regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regexp_match(..., flags=...) and ColumnElement.regexp_replace(..., flags=...) were processed through the standard parameter pipeline, producing incorrect SQL. Flags are now rendered as inline string literals, matching Snowflake's expected REGEXP_LIKE(col, pattern, 'i') / REGEXP_REPLACE(col, pattern, replacement, 'i') syntax.…the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing sp…
v1.11.1 (Sep 3, 2026)
Fix / operator to always emit true division (left / right). The flag gates a deprecation warning. The // operator continues to always emit FLOOR(left / right) (GH #756).
v1.11.0 (July 7, 2026)
Sensitive connection parameters: We curated a set of connector kwargs (host, protocol, token_file_path, private_key_file, ocsp_response_cache_filename, connection_diag_log_path, crl_cache_dir, unsafe_file_write, unsafe_skip_file_permissions_check) that can no longer be supplied via the URL query string — pass them via connect_args= in create_engine() instead (applications already doing so are unaffected). If you encounter a possible behavioral change, set SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS=1 and follow the instructions at Sensitive connection parameters.
Improve _url() helper and create_connect_args connection-parameter handling:
account and region values are now validated against an allowlist of DNS-safe characters (alphanumeric, -, ., _) before being interpolated into the connection URL, preventing URL-authority corruption from unexpected characters.user values are percent-encoded before being placed in the URL userinfo component, preserving the original value delivered to the connector while preventing @, ?, and # from being misinterpreted as URL delimiters.ClusterByOption now raises TypeError at DDL compile time when an expression element is neither a str nor a sqlalchemy.sql.expression.TextClause. Previously, such values were silently coerced via str(), which produced malformed DDL (e.g. bind-parameter placeholders like :id_1) for any expression beyond a bare column name. The accepted types match the documented constructor signature; code using only str or text(...) is unaffected.
Improve identifier quoting and string-literal escaping for caller-supplied values across the DDL compiler so they are rendered consistently with the rest of the dialect. Column keys in MERGE INTO (both the WHEN NOT MATCHED … INSERT list and WHEN MATCHED … SET targets), stage namespaces/names, format_name, and file_format are now routed through the identifier preparer, and a quote=False schema or column label is quoted when it contains characters that would otherwise require quoting instead of being emitted verbatim. Cloud-storage URIs, CREDENTIALS, and ENCRYPTION clauses (shared by COPY INTO and CREATE STAGE) and FILES=(…) entries now apply the dialect's standard literal escaping ('→'', \→\\); CREATE STAGE builds its SQL from the container's fields rather than repr(container). Legal identifiers — including upper-case names that Snowflake folds — still render bare, so existing DDL is unchanged.
Improve escaping in CopyIntoStorage / CreateStage / CreateFileFormat and add secret redaction (SNOW-3656048):
FILE_FORMAT option string values (e.g. CSVFormatter().date_format(...), file_extension, timestamp_format) are now escaped before being embedded in the compiled SQL, improving single-quote escaping (SNOW-3649888).repr() of AWSBucket, AzureContainer, GCSBucket (and, transitively, CopyIntoStorage) now masks cloud secrets — AWS_SECRET_KEY, AWS_KEY_ID, AWS_TOKEN, AZURE_SAS_TOKEN, MASTER_KEY are rendered as '***' (SNOW-3649782). Compiled SQL is unchanged.SnowflakeSecretRedactionFilter, add_secret_redaction_filter(), and redact_secrets(). Prefer STORAGE_INTEGRATION to avoid putting secrets in SQL at all (SNOW-3649850).Fix get_view_definition silently truncating or failing for view names that contain a single quote or backslash (e.g. o'brien). The name is now SQL-escaped (' → '', \ → \\) before being embedded in the SHOW VIEWS LIKE '...' literal, so such views are found correctly and no query manipulation is possible via a crafted view name.
Optimise single-table reflection (SNOW-3720548): reflecting one table no longer scans the entire schema, reducing latency and Snowflake credit usage for targeted Inspector calls.
v1.10.2 (June 18, 2026)
Fix double-escaped identifier quoting for database-qualified schemas in _StructuredTypeInfoManager.get_table_columns. Since v1.10.1 the schema was quoted as a single identifier, so a qualified schema such as "MYDB"."MYSCHEMA" became """MYDB"".""MYSCHEMA""" and the DESC TABLE fallback failed (emitting a Failed to reflect table ... sqlalchemy:_get_schema_columns warning) for every structured-typed table when reflecting a non-default database.schema. The schema is now split on the dot and each component is double-quoted individually, preserving the SNOW-3480955 injection guard while correctly handling qualified schemas.
v1.10.1 (June 15, 2026)
regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regexp_match(..., flags=...) and ColumnElement.regexp_replace(..., flags=...) were processed through the standard parameter pipeline, producing incorrect SQL. Flags are now rendered as inline string literals, matching Snowflake's expected REGEXP_LIKE(col, pattern, 'i') / REGEXP_REPLACE(col, pattern, replacement, 'i') syntax._StructuredTypeInfoManager.get_table_columns. The DESC TABLE fallback path used raw denormalised names in an f-string while all other reflection paths apply ip.quote(denormalize_name(...)) via _always_quote_join. Schema and table components are now consistently double-quoted before the statement is constructed, and the method delegates to get_table_columns_by_full_name to collapse the two previously divergent code paths.v1.10.0 (May 20, 2026)
with_loader_criteria silently dropping filters on non-Snowflake dialects (#676). Importing snowflake-sqlalchemy previously altered SQLAlchemy's ORM compilation for every dialect in the process, causing loader-criteria filters to be omitted inside sealed subqueries when using PostgreSQL, MySQL, SQLite, etc. Snowflake dialect behavior is unchanged; the BCR-1057 lateral-join workaround is now scoped to Snowflake connections only.UUID column type to sqlalchemy.sql.sqltypes.UUID for reflection on SQLAlchemy 2.x (#681). Previously reflected as NullType. Values are returned as plain strings (as_uuid=False) rather than uuid.UUID instances. No change on SQLAlchemy 1.4 where the generic UUID type does not exist.CopyIntoStorage (SNOW-721174, #368).referred_schema=None normalization in foreign key reflection to the default schema only (#610, SNOW-2313675):
referred_schema=None, preserving compatibility with the upstream reflection test suite and with applications that do not qualify default-schema FK targets.referred_schema, which prevents SQLAlchemy's _reflect_fk from autoloading a non-default-schema target from the wrong place (the bug behind #610) and avoids the Alembic autogenerate mismatch that previously occurred when user metadata explicitly qualified a cross-schema FK that happened to target the default schema.SnowflakeBase, snowflake_declarative_base(), and SnowflakeSession to enable efficient bulk inserts for ORM models with nullable optional columns (SNOW-893080, #441). When session.bulk_save_objects() is used with models that have randomly populated nullable columns, SQLAlchemy normally groups objects by their set of non-None column keys, producing O(N) separate INSERT statements. SnowflakeBase / snowflake_declarative_base() pre-populate all plain-nullable columns at construction time, and SnowflakeSession passes render_nulls=True so all objects share the same parameter-key set and are batched into a single executemany INSERT._split_schema_by_dot now correctly parses SQL-escaped double-quotes ("") inside quoted schema/database identifiers (e.g. "my""schema" → my"schema), preventing silent truncation of identifiers containing literal quote characters.denormalize_column_name now correctly double-quotes quoted_name("mycol", True) columns in CLUSTER BY clauses instead of silently dropping the case-sensitivity signal. The caller has already opted into case-sensitive semantics by constructing a quoted_name(..., True), so this is honoured independently of the dialect flag._has_object (used by has_table / has_sequence) now applies denormalize_name to both the schema and object name before building the DESC SQL, making it consistent with all other reflection methods._NameUtils swap in create_connect_args — when the URL's case_sensitive_identifiers value differs from the current dialect state, the name_utils instance is replaced rather than mutated in place, so concurrent readers on other threads observe either the old or the new instance but never a torn update.case_sensitive_identifiers opt-in engine flag (kwarg or ?case_sensitive_identifiers=True URL param) governing three related behaviours. The default is False; existing applications are unaffected unless they explicitly opt in:
TABLE) are normalised to quoted_name("table", True) instead of returning unchanged, preventing key-lookup mismatches between creation and reflection.MyCol from a quoted Snowflake column) are returned as quoted_name("MyCol", True) instead of a plain str. Emitted SQL is identical in both modes (_requires_quotes force-quotes any name containing uppercase chars); the difference is only observable via isinstance(..., quoted_name) and .quote.'"myschema"' or '"mydb"."myschema"' — have their extracted parts marked quote=True by _split_schema_by_dot, preserving case-sensitivity in emitted SQL. Without the flag, the extracted parts keep quote=None and the preparer's _requires_quotes heuristic decides per-part (stripping inner quotes for all-lowercase parts, which matches pre-PR behaviour). Use quoted_name("myschema", True) or MetaData(schema=quoted_name(..., True)) to opt into case-sensitivity on a per-value basis without enabling the flag.create_snowflake_engine(url, schema=..., case_sensitive_schema=True) helper that URL-encodes case-sensitive schema names using %22 so the Snowflake connector receives the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing special characters (?, #, /) from being misinterpreted as URL delimiters by SQLAlchemy's URL parser.snowflake.sqlalchemy.alembic_util.render_item — a drop-in Alembic render_item hook for env.py that serialises quoted_name columns with quote=True correctly in generated migration files, preventing Alembic autogenerate from silently converting case-sensitive column names to uppercase.SnowflakeWarning at DDL compile time when Identity() is used on a primary key column, alerting users that ORM flush operations will raise a FlushError. The warning is emitted once per unique (table, column) pair per Python process. Use Sequence() instead.schema='database.schema' notation. This allows reflecting and joining tables from different databases in a single session without raw SQL. (#456)/) when div_is_floordiv=True: the Snowflake compiler now correctly delegates to the SQLAlchemy base implementation, emitting CAST(col AS NUMERIC) for integer operands as it did before #545 introduced the override (#618).get_multi_columns, get_multi_pk_constraint, get_multi_unique_constraints, get_multi_foreign_keys for SQLAlchemy 2.x bulk reflection — each issues one schema-wide query per reflection pass instead of one query per table.get_columns now uses DESC TABLE directly (per-table, live) since get_multi_columns handles all bulk reflection; temporary tables and dynamic tables are reflected correctly without schema-wide queries.SHOW INDEXES IN TABLE replacing the previous SHOW TABLES LIKE approach for single-table index reflection, eliminating SQL LIKE wildcard false-positives and case-sensitivity bugs._always_quote_join helper that always quotes denormalised identifiers — ensures correct SQL for case-sensitive table and schema names in per-table reflection paths.referred_schema resolution so reflected FKs always keep their actual schema unless the target lives in the connection's default schema. Previously FKs whose target shared the reflected non-default schema were reported with referred_schema=None, which caused SQLAlchemy's _reflect_fk to autoload from the wrong schema and raise NoReferencedColumnError during Alembic autogenerate._parse_pk_rows, _parse_uk_rows, _parse_fk_rows) so correctness fixes propagate to both per-table and schema-wide reflection paths.cache_column_metadata=True opt-in enables per-table SHOW … IN TABLE queries for get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes on SQLAlchemy 1.4.get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes now automatically use per-table SHOW … IN TABLE queries without any opt-in flag. Previously these methods always issued SHOW … IN SCHEMA even for single-table Inspector calls (e.g. pandas.read_sql_table()), causing ~20-second delays on schemas with thousands of tables (SNOW-689531).v1.9.0 (March 4, 2026)
DECFLOAT and VECTOR data typesILIKE in queriesSYSDATE() renderingNone arguments in internal schema resolution (#623).SHOW TABLES returns empty string table names, causing IndexError during reflection (#296).always, cycle, order, etc.).pytest-xdist parallel test support via per-worker schema provisioning hooks.pandas lower bound in sa14 test environment from <2.1 to >=2.1.1,<2.2 to ensure pre-built wheels are available for Python 3.12v1.8.2 (December 9, 2025)
snowflake-connector-pythonv1.8.1 (December 9, 2025)
v1.8.0(December 5, 2025)
snowflake-connector-python<5.0.0v1.7.7(September 3, 2025)
v1.7.6(July 10, 2025)
v1.7.5(June 20, 2025)
v1.7.4(June 10, 2025)
v1.7.3(January 15, 2025)
_ as initial character.force_div_is_floordiv flag to override div_is_floordiv new default value False in SnowflakeDialect.
False, the / division operator will be treated as a float division and // as a floor division.div_is_floor_div as False.v1.7.2(December 18, 2024)
_ as column namev1.7.1(December 02, 2024)
v1.7.0(November 21, 2024)
v1.6.1(July 9, 2024)
v1.6.0(July 8, 2024)
hatch & uv for managing project virtual environmentsv1.5.4
v1.5.3(April 16, 2024)
v1.5.2(April 11, 2024)
v1.5.1(November 03, 2023)
externalbrowser authentication not caching due to incorrect parsing of boolean query parameters.
v1.5.0(Aug 23, 2023)
v1.4.7(Mar 22, 2023)
SnowflakeConnection to SnowflakeSQLAlchemy.SnowflakeDialect.get_columns now throws a NoSuchTableError exception when the specified table doesn't exist, instead of the more vague KeyError.sqlalchemy.func.now is not rendered correctly.v1.4.6(Feb 8, 2023)
v1.4.5(Dec 7, 2022)
SnowflakeConnection to SnowflakeSQLAlchemy.v1.4.4(Nov 16, 2022)
v1.4.3(Oct 17, 2022)
SnowflakeDialect.normalize_name and SnowflakeDialect.denormalize_name could not handle empty string.sqlalchemy.engine.url._rfc_1738_quote as it is removed from SQLAlchemy v1.4.42.v1.4.2(Sep 19, 2022)
v1.4.1(Aug 18, 2022)
DATE should not be removed from SnowflakeDialect.ischema_names.inspect.get_foreign_keys and inspect.get_pk_constraint.v1.4.0(July 20, 2022)
regexp_match, regexp_replace in sqlalchemy.sql.expression.ColumnOperators.Date, DateTime, Time, Float and Numeric, and converting the values into corresponding Python objects.get_sequence_names in SnowflakeDialect.v1.3.4(April 27,2022)
v1.3.3(December 19,2021)
v1.3.2 (September 14,2021)
v1.3.1 (July 23,2021)
v1.2.5 (July 20,2021)
v1.2.4 (October 05,2020)
v1.2.3 (March 30, 2020)
v1.2.2 (March 9, 2020)
v1.2.1 (February 18,2020)
v1.2.0 (January 27, 2020)
v1.1.18 (January 6,2020)
v1.1.17 (December 2,2019)
v1.1.16 (October 21,2019)
v1.1.15 (September 30, 2019)
v1.1.14 (August 12, 2019)
FIELD_DELIMETER spelled incorrectlyv1.1.13 (May 20,2019)
v1.1.12 (April 8,2019)
v1.1.11 (March 25, 2019)
v1.1.10 (February 22, 2019)
v1.1.9 (February 11, 2019)
v1.1.8 (February 8, 2019)
v1.1.7 (February 8, 2019)
v1.1.6 (January 3, 2019)
v1.1.5 (December 19, 2018)
v1.1.4 (November 13, 2018)
v1.1.3 (October 30, 2018)
v1.1.2 (June 7, 2018)
v1.1.1 (May 17, 2018)
region parameter support to URL utility method.v1.1.0 (February 1, 2018)
role example.get_pk_constraint and get_primary_keys. Those applications that depend on the old behaviors must update codes. Issue #38 (@nrth)open and close connections.v1.0.9 (January 4, 2018)
v1.0.8 (December 21, 2017)
get_schema_names method to Snowflake SQLAlchemy dialect. PR #20(andrewsali)v1.0.7 (May 18, 2017)
v1.0.6 (April 20, 2017)
v1.0.5 (April 13, 2017)
snowflake_clusterby option support to Table object so that the user can create a table with clustering keysv1.0.4 (March 9, 2017)
v1.0.3 (October 20, 2016)
VARIANT, OBJECT and ARRAY data type supports for fetchv1.0.2 (July 5, 2016)
v1.0.1 (July 4, 2016)
v1.0.0 (June 28, 2016)
[Sensitive connection parameters](https://github.com/snowflakedb/snowflake-sqlalchemy#sensitive-connection-parameters): We curated a set of connector
v1.11.0 (July 7, 2026)
Sensitive connection parameters: We curated a set of connector kwargs (host, protocol, token_file_path, private_key_file, ocsp_response_cache_filename, connection_diag_log_path, crl_cache_dir, unsafe_file_write, unsafe_skip_file_permissions_check) that can no longer be supplied via the URL query string — pass them via connect_args= in create_engine() instead (applications already doing so are unaffected). If you encounter a possible behavioral change, set SNOWFLAKE_SQLALCHEMY_LEGACY_URL_PARAMS=1 and follow the instructions at Sensitive connection parameters.
Improve _url() helper and create_connect_args connection-parameter handling:
account and region values are now validated against an allowlist of DNS-safe characters (alphanumeric, -, ., _) before being interpolated into the connection URL, preventing URL-authority corruption from unexpected characters.user values are percent-encoded before being placed in the URL userinfo component, preserving the original value delivered to the connector while preventing @, ?, and # from being misinterpreted as URL delimiters.ClusterByOption now raises TypeError at DDL compile time when an expression element is neither a str nor a sqlalchemy.sql.expression.TextClause. Previously, such values were silently coerced via str(), which produced malformed DDL (e.g. bind-parameter placeholders like :id_1) for any expression beyond a bare column name. The accepted types match the documented constructor signature; code using only str or text(...) is unaffected.
Improve identifier quoting and string-literal escaping for caller-supplied values across the DDL compiler so they are rendered consistently with the rest of the dialect. Column keys in MERGE INTO (both the WHEN NOT MATCHED … INSERT list and WHEN MATCHED … SET targets), stage namespaces/names, format_name, and file_format are now routed through the identifier preparer, and a quote=False schema or column label is quoted when it contains characters that would otherwise require quoting instead of being emitted verbatim. Cloud-storage URIs, CREDENTIALS, and ENCRYPTION clauses (shared by COPY INTO and CREATE STAGE) and FILES=(…) entries now apply the dialect's standard literal escaping ('→'', \→\\); CREATE STAGE builds its SQL from the container's fields rather than repr(container). Legal identifiers — including upper-case names that Snowflake folds — still render bare, so existing DDL is unchanged.
Improve escaping in CopyIntoStorage / CreateStage / CreateFileFormat and add secret redaction (SNOW-3656048):
FILE_FORMAT option string values (e.g. CSVFormatter().date_format(...), file_extension, timestamp_format) are now escaped before being embedded in the compiled SQL, improving single-quote escaping (SNOW-3649888).repr() of AWSBucket, AzureContainer, GCSBucket (and, transitively, CopyIntoStorage) now masks cloud secrets — AWS_SECRET_KEY, AWS_KEY_ID, AWS_TOKEN, AZURE_SAS_TOKEN, MASTER_KEY are rendered as '***' (SNOW-3649782). Compiled SQL is unchanged.SnowflakeSecretRedactionFilter, add_secret_redaction_filter(), and redact_secrets(). Prefer STORAGE_INTEGRATION to avoid putting secrets in SQL at all (SNOW-3649850).Fix get_view_definition silently truncating or failing for view names that contain a single quote or backslash (e.g. o'brien). The name is now SQL-escaped (' → '', \ → \\) before being embedded in the SHOW VIEWS LIKE '...' literal, so such views are found correctly and no query manipulation is possible via a crafted view name.
Optimise single-table reflection (SNOW-3720548): reflecting one table no longer scans the entire schema, reducing latency and Snowflake credit usage for targeted Inspector calls.
Nothing published for this version
Fix regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regex
regexp_match and regexp_replace flags rendered as bound parameters instead of literal strings (#SNOW-3573046). Flags passed to ColumnElement.regexp_match(..., flags=...) and ColumnElement.regexp_replace(..., flags=...) were processed through the standard parameter pipeline, producing incorrect SQL. Flags are now rendered as inline string literals, matching Snowflake's expected REGEXP_LIKE(col, pattern, 'i') / REGEXP_REPLACE(col, pattern, replacement, 'i') syntax._StructuredTypeInfoManager.get_table_columns. The DESC TABLE fallback path used raw denormalised names in an f-string while all other reflection paths apply ip.quote(denormalize_name(...)) via _always_quote_join. Schema and table components are now consistently double-quoted before the statement is constructed, and the method delegates to get_table_columns_by_full_name to collapse the two previously divergent code paths.…the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing sp…
with_loader_criteria silently dropping filters on non-Snowflake dialects (#676). Importing snowflake-sqlalchemy previously altered SQLAlchemy's ORM compilation for every dialect in the process, causing loader-criteria filters to be omitted inside sealed subqueries when using PostgreSQL, MySQL, SQLite, etc. Snowflake dialect behavior is unchanged; the BCR-1057 lateral-join workaround is now scoped to Snowflake connections only.UUID column type to sqlalchemy.sql.sqltypes.UUID for reflection on SQLAlchemy 2.x (#681). Previously reflected as NullType. Values are returned as plain strings (as_uuid=False) rather than uuid.UUID instances. No change on SQLAlchemy 1.4 where the generic UUID type does not exist.CopyIntoStorage (SNOW-721174, #368).referred_schema=None normalization in foreign key reflection to the default schema only (#610, SNOW-2313675):
referred_schema=None, preserving compatibility with the upstream reflection test suite and with applications that do not qualify default-schema FK targets.referred_schema, which prevents SQLAlchemy's _reflect_fk from autoloading a non-default-schema target from the wrong place (the bug behind #610) and avoids the Alembic autogenerate mismatch that previously occurred when user metadata explicitly qualified a cross-schema FK that happened to target the default schema.SnowflakeBase, snowflake_declarative_base(), and SnowflakeSession to enable efficient bulk inserts for ORM models with nullable optional columns (SNOW-893080, #441). When session.bulk_save_objects() is used with models that have randomly populated nullable columns, SQLAlchemy normally groups objects by their set of non-None column keys, producing O(N) separate INSERT statements. SnowflakeBase / snowflake_declarative_base() pre-populate all plain-nullable columns at construction time, and SnowflakeSession passes render_nulls=True so all objects share the same parameter-key set and are batched into a single executemany INSERT._split_schema_by_dot now correctly parses SQL-escaped double-quotes ("") inside quoted schema/database identifiers (e.g. "my""schema" → my"schema), preventing silent truncation of identifiers containing literal quote characters.denormalize_column_name now correctly double-quotes quoted_name("mycol", True) columns in CLUSTER BY clauses instead of silently dropping the case-sensitivity signal. The caller has already opted into case-sensitive semantics by constructing a quoted_name(..., True), so this is honoured independently of the dialect flag._has_object (used by has_table / has_sequence) now applies denormalize_name to both the schema and object name before building the DESC SQL, making it consistent with all other reflection methods._NameUtils swap in create_connect_args — when the URL's case_sensitive_identifiers value differs from the current dialect state, the name_utils instance is replaced rather than mutated in place, so concurrent readers on other threads observe either the old or the new instance but never a torn update.case_sensitive_identifiers opt-in engine flag (kwarg or ?case_sensitive_identifiers=True URL param) governing three related behaviours. The default is False; existing applications are unaffected unless they explicitly opt in:
TABLE) are normalised to quoted_name("table", True) instead of returning unchanged, preventing key-lookup mismatches between creation and reflection.MyCol from a quoted Snowflake column) are returned as quoted_name("MyCol", True) instead of a plain str. Emitted SQL is identical in both modes (_requires_quotes force-quotes any name containing uppercase chars); the difference is only observable via isinstance(..., quoted_name) and .quote.'"myschema"' or '"mydb"."myschema"' — have their extracted parts marked quote=True by _split_schema_by_dot, preserving case-sensitivity in emitted SQL. Without the flag, the extracted parts keep quote=None and the preparer's _requires_quotes heuristic decides per-part (stripping inner quotes for all-lowercase parts, which matches pre-PR behaviour). Use quoted_name("myschema", True) or MetaData(schema=quoted_name(..., True)) to opt into case-sensitivity on a per-value basis without enabling the flag.create_snowflake_engine(url, schema=..., case_sensitive_schema=True) helper that URL-encodes case-sensitive schema names using %22 so the Snowflake connector receives the literal double-quoted form. Fix security vulnerability: schema names are now always URL-encoded regardless of case_sensitive_schema, preventing special characters (?, #, /) from being misinterpreted as URL delimiters by SQLAlchemy's URL parser.snowflake.sqlalchemy.alembic_util.render_item — a drop-in Alembic render_item hook for env.py that serialises quoted_name columns with quote=True correctly in generated migration files, preventing Alembic autogenerate from silently converting case-sensitive column names to uppercase.SnowflakeWarning at DDL compile time when Identity() is used on a primary key column, alerting users that ORM flush operations will raise a FlushError. The warning is emitted once per unique (table, column) pair per Python process. Use Sequence() instead.schema='database.schema' notation. This allows reflecting and joining tables from different databases in a single session without raw SQL. (#456)/) when div_is_floordiv=True: the Snowflake compiler now correctly delegates to the SQLAlchemy base implementation, emitting CAST(col AS NUMERIC) for integer operands as it did before #545 introduced the override (#618).get_multi_columns, get_multi_pk_constraint, get_multi_unique_constraints, get_multi_foreign_keys for SQLAlchemy 2.x bulk reflection — each issues one schema-wide query per reflection pass instead of one query per table.get_columns now uses DESC TABLE directly (per-table, live) since get_multi_columns handles all bulk reflection; temporary tables and dynamic tables are reflected correctly without schema-wide queries.SHOW INDEXES IN TABLE replacing the previous SHOW TABLES LIKE approach for single-table index reflection, eliminating SQL LIKE wildcard false-positives and case-sensitivity bugs._always_quote_join helper that always quotes denormalised identifiers — ensures correct SQL for case-sensitive table and schema names in per-table reflection paths.referred_schema resolution so reflected FKs always keep their actual schema unless the target lives in the connection's default schema. Previously FKs whose target shared the reflected non-default schema were reported with referred_schema=None, which caused SQLAlchemy's _reflect_fk to autoload from the wrong schema and raise NoReferencedColumnError during Alembic autogenerate._parse_pk_rows, _parse_uk_rows, _parse_fk_rows) so correctness fixes propagate to both per-table and schema-wide reflection paths.cache_column_metadata=True opt-in enables per-table SHOW … IN TABLE queries for get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes on SQLAlchemy 1.4.get_pk_constraint, get_unique_constraints, get_foreign_keys, and get_indexes now automatically use per-table SHOW … IN TABLE queries without any opt-in flag. Previously these methods always issued SHOW … IN SCHEMA even for single-table Inspector calls (e.g. pandas.read_sql_table()), causing ~20-second delays on schemas with thousands of tables (SNOW-689531).Add support for DECFLOAT and VECTOR data types
DECFLOAT and VECTOR data typesILIKE in queriesSYSDATE() renderingNone arguments in internal schema resolution (#623).SHOW TABLES returns empty string table names, causing IndexError during reflection (#296).always, cycle, order, etc.).pytest-xdist parallel test support via per-worker schema provisioning hooks.pandas lower bound in sa14 test environment from <2.1 to >=2.1.1,<2.2 to ensure pre-built wheels are available for Python 3.12Updated supported max python version to 3.13
v1.8.2 (December 9, 2025)
snowflake-connector-pythonv1.8.1 (December 9, 2025)
Add logging of SQLAlchemy version
snowflake-connector-python<5.0.0Fix exception for structured type columns dropped while collecting meetadata
Fix get_multi_indexes issue, wrong assign of returned indexes when processing multiple indexes in a table
Fix compilation of Merge and Copy Into was not working
Fix dependency on DESCRIBE TABLE columns quantity (differences in columns caused by Snowflake parameters).
Fix support for SqlAlchemy ARRAY.
_ as initial character.force_div_is_floordiv flag to override div_is_floordiv new default value False in SnowflakeDialect.
False, the / division operator will be treated as a float division and // as a floor division.div_is_floor_div as False.Fix quoting of _ as column name
_ as column nameAdd support for partition by to copy into
Add support for dynamic tables and required options
v1.7.0(November 22, 2024)
Update internal project workflow with pypi publishing
v1.6.1(July 9, 2024)
v1.6.0(July 8, 2024)
hatch & uv for managing project virtual environmentsv1.5.4
Limit SQLAlchemy to < 2.0.0 before releasing version compatible with 2.0
v1.5.3(April 16, 2024)
Bump min SQLAlchemy to 1.4.19 for outer lateral join
v1.5.2(April 11, 2024)
Fixed a compatibility issue with Snowflake Behavioral Change 1057 on outer lateral join, for more details check https://docs.snowflake.com/en/release-
externalbrowser authentication not caching due to incorrect parsing of boolean query parameters.
Added option to create a temporary stage command. Added support for geometry type. Fixed a compatibility issue of regex expression with SQLAlchemy 1.4
Added option to create a temporary stage command. Added support for geometry type. Fixed a compatibility issue of regex expression with SQLAlchemy 1.4.49.
Re-applied the application name of driver connection SnowflakeConnection to SnowflakeSQLAlchemy.
SnowflakeConnection to SnowflakeSQLAlchemy.SnowflakeDialect.get_columns now throws a NoSuchTableError exception when the specified table doesn't exist, instead of the more vague KeyError.sqlalchemy.func.now is not rendered correctly.Bumped snowflake-connector-python dependency to newest version which supports Python 3.11.
Updated the application name of driver connection SnowflakeConnection to SnowflakeSQLAlchemy.
SnowflakeConnection to SnowflakeSQLAlchemy.Fixed a bug that percent signs in a non-compiled statement should not be interpolated with emtpy sequence when executed.
Fixed a bug that SnowflakeDialect.normalize_name and SnowflakeDialect.denormalize_name could not handle empty string.
SnowflakeDialect.normalize_name and SnowflakeDialect.denormalize_name could not handle empty string.sqlalchemy.engine.url._rfc_1738_quote as it is removed from SQLAlchemy v1.4.42.Improved performance by standardizing string interpolations to f-strings.
Fixed breaking changes introduced in release 1.4.0 that:
Added support for regexp_match, regexp_replace in sqlalchemy.sql.expression.ColumnOperators.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per quarter.