Snowflake dialect issues

[sineline]

Bug: Snowflake Iceberg tables fail on CLONE and ALTER — missing ICEBERG keyword in DDL

Description

When using table_format = iceberg with Snowflake, SQLMesh generates standard DDL statements (CREATE TABLE ... CLONE, ALTER TABLE ...) that Snowflake rejects for Iceberg tables. Snowflake requires dedicated Iceberg-prefixed commands:

• CREATE ICEBERG TABLE ... CLONE instead of CREATE TABLE ... CLONE
• ALTER ICEBERG TABLE ... instead of ALTER TABLE ...

This causes ProgrammingError failures during both the virtual layer update (clone phase) and schema migration (alter phase).

How to Reproduce

1. Configure a SQLMesh project targeting Snowflake with Iceberg table format
2. Define any materialized model (e.g., INCREMENTAL_BY_TIME_RANGE) with table_format iceberg
3. Run sqlmesh plan to create a dev environment that triggers cloning or schema migration

Error Messages

Clone error (during virtual layer update):

ProgrammingError: 002008 (55000): SQL compilation error:
Invalid operation The table STAGING__STG_CLOUDFRONT_LOGS__788519766 is an Iceberg table.
Clone Iceberg table should use CREATE ICEBERG TABLE CLONE command.
Please ensure the appropriate command format is used.

Alter error (during schema migration):

ProgrammingError: 091367 (42601): SQL compilation error:
The table STAGING__STG_TIVO_CONSENT_PREFERENCES__1774833790__dev is an Iceberg table.
Iceberg tables should use ALTER ICEBERG TABLE commands.
Please ensure the appropriate command format is used.

Root Cause

The table_format attribute is available on the model (model.table_format) and is correctly used during table creation (_create_table / ctas generate CREATE ICEBERG TABLE), but it is never propagated to the clone and alter code paths:

1. Clone: EngineAdapter.clone_table() in base.py hardcodes kind="TABLE" in exp.Create. The Snowflake override in snowflake.py handles TransientProperty but not the Iceberg table_format. The evaluator's _clone_snapshot_in_dev() does not pass table_format to clone_table().

2. Alter: TableAlterOperation.expression in schema_diff.py hardcodes kind="TABLE" in exp.Alter. The base alter_table() method does not accept additional context, and the evaluator's migrate() does not pass table_format through.

Affected Code

File	Method	Issue
sqlmesh/core/engine_adapter/base.py	clone_table()	Hardcodes kind="TABLE"
sqlmesh/core/engine_adapter/base.py	alter_table()	No mechanism to receive table_format
sqlmesh/core/engine_adapter/snowflake.py	clone_table()	Handles TransientProperty but not Iceberg
sqlmesh/core/engine_adapter/snowflake.py	—	No alter_table() override for Iceberg
sqlmesh/core/snapshot/evaluator.py	_clone_snapshot_in_dev()	Doesn't pass table_format to clone_table()
sqlmesh/core/snapshot/evaluator.py	migrate()	Doesn't pass table_format to alter_table()

Environment

• SQLMesh version: 0.231.2
• Engine: Snowflake with Iceberg table format
• Trigger: sqlmesh plan with dev environments (clone + migration phases)

[sineline]

I have fixed the issue in my local branch, can I submit a PR?