Quote reserved-word identifiers via sqlglot.to_identifier

_quote_identifier was only quoting names with dots/special chars but
passing reserved words like 'order' through unquoted, producing
invalid SQL. Now delegates to sqlglot.to_identifier for all simple
names so reserved words get quoted automatically. Uses lru_cache to
avoid perf regression from repeated sqlglot calls.

Author: nicosuave

sidemantic/sql/generator.py

@@ -2,6 +2,7 @@
 
 import logging
 import threading
+from functools import lru_cache
 
 import sqlglot
 from sqlglot import exp, select
@@ -12,6 +13,15 @@
 from sidemantic.core.symmetric_aggregate import build_symmetric_aggregate_sql
 from sidemantic.sql.aggregation_detection import sql_has_aggregate
 
+
+@lru_cache(maxsize=4096)
+def _quote_identifier_cached(name: str, dialect: str, is_simple: bool) -> str:
+    """Cached identifier quoting, shared across all SQLGenerator instances."""
+    if is_simple:
+        return sqlglot.to_identifier(name).sql(dialect=dialect)
+    return sqlglot.to_identifier(name, quoted=True).sql(dialect=dialect)
+
+
 _dialect_cache: dict[str, Dialect] = {}
 _tls = threading.local()
 
@@ -232,11 +242,10 @@ def _quote_identifier(self, name: str) -> str:
         """Quote a SQL identifier for the current dialect.
 
         Delegates to sqlglot which handles reserved words (e.g., 'order')
-        and special characters automatically.
+        and special characters automatically. Results are cached since the
+        same identifiers are used many times during query generation.
         """
-        if self._is_simple_identifier(name):
-            return name
-        return sqlglot.to_identifier(name, quoted=True).sql(dialect=self.dialect)
+        return _quote_identifier_cached(name, self.dialect, self._is_simple_identifier(name))
 
     def _cte_name(self, model_name: str) -> str:
         """Get the CTE identifier name for a model."""