docs: clarify physical property reference syntax

Signed-off-by: mday-io <mdaytn@gmail.com>

Author: mday-io

docs/integrations/engines/starrocks.md

@@ -517,9 +517,9 @@ SELECT user_id, COUNT(*) AS event_count FROM user_events GROUP BY user_id;
 * `excluded_trigger_tables`: base tables whose data changes should **not** automatically trigger a refresh of this MV.
 * `excluded_refresh_tables`: base tables that should **not** be scanned when the MV refreshes.
 
-Both properties accept a single table reference, a comma-separated string of table references, or a tuple or array of table references.
+Both properties accept a single reference, a comma-separated string, a tuple, or an array. References can be bare identifiers or quoted strings.
 
-StarRocks requires the **physical** base table name for these properties, not the logical view name that SQLMesh normally exposes. SQLMesh handles this automatically: when a reference matches a managed SQLMesh model, the logical name is resolved to its physical table name before the `CREATE MATERIALIZED VIEW` statement is issued. References that do not match any managed model are passed through unchanged.
+StarRocks requires physical base table names in these properties. SQLMesh replaces references to managed models with physical table names and passes other references through unchanged. It sends the resulting names to StarRocks as a comma-separated string.
 
 ```sql
 MODEL (
@@ -529,7 +529,6 @@ MODEL (
   ),
   physical_properties (
     refresh_scheme = 'ASYNC',
-    -- SQLMesh resolves mydb.orders and mydb.order_items to their physical table names
     excluded_trigger_tables = 'mydb.orders,mydb.order_items',
     excluded_refresh_tables = mydb.orders
   )
@@ -538,18 +537,16 @@ MODEL (
 SELECT order_id, SUM(amount) AS total FROM mydb.orders GROUP BY order_id;
 ```
 
-A single reference can be written as a bare identifier (`mydb.orders`) or as a quoted string. Multiple references can be provided as a quoted, comma-separated string (`'mydb.orders,mydb.order_items'`), a tuple, or an array. Tuple and array elements can be bare identifiers, quoted strings, or a mix of both:
+Tuple and array elements can mix bare identifiers and quoted strings:
 
 ```sql
 physical_properties (
-  refresh_scheme = 'ASYNC',
-  excluded_trigger_tables = (mydb.orders, 'mydb.order_items'),
-  excluded_refresh_tables = ['mydb.orders', mydb.order_items]
+	refresh_scheme = 'ASYNC',
+	excluded_trigger_tables = (mydb.orders, 'mydb.order_items'),
+	excluded_refresh_tables = ['mydb.orders', mydb.order_items]
 )
 ```
 
-SQLMesh resolves each reference individually and passes the result to StarRocks as a comma-separated string of table names, regardless of the input form.
-
 **Other properties:**
 
 You can specify `partitioning`, `distribution`, `order by` and `properties` the same as normal table properties. But notice that only supported MV properties are useful, Refer to StarRocks' doc for MV creation.

sqlmesh/core/model/definition.py

@@ -2832,13 +2832,10 @@ def _split_sql_model_statements(
 def _resolve_model_refs_to_physical_tables(
     value: exp.Expr, table_mapping: t.Dict[str, str], dialect: DialectType
 ) -> exp.Literal:
-    """Resolve managed-model references in a property value to their physical table names.
+    """Resolve managed-model references in a property value to physical table names.
 
-    The value is a single table reference, a comma-separated string, or a tuple/array of
-    table references. Each reference that matches a managed model (via ``table_mapping``) is
-    swapped for its physical ``db.table`` name;
-    anything else (e.g. a raw source) is kept as written. Returns a single string literal so the
-    property renders just like a hand-written value.
+    Unmatched references keep their table names. The function returns the names as one
+    comma-separated string literal.
     """
     if isinstance(value, exp.Literal) and value.is_string:
         refs = value.this.split(",")