SQLAlchemy core improvements

CHANGELOG.md

@@ -24,9 +24,14 @@ The supported method of passing ClickHouse server settings is to prefix such arg
 ## UNRELEASED
 
 ### New Features
+- SQLAlchemy: Comprehensive ClickHouse JOIN support via the new `ch_join()` helper. All strictness modifiers (`ALL`, `ANY`, `SEMI`, `ANTI`, `ASOF`), the `GLOBAL` distribution modifier, and explicit `CROSS JOIN` are now available. Use with `select_from()` to generate ClickHouse-specific join syntax like `GLOBAL ALL LEFT OUTER JOIN`. Closes [#635](https://github.com/ClickHouse/clickhouse-connect/issues/635)
+- SQLAlchemy: `array_join()` now supports multiple columns for parallel array expansion. Pass a list of columns and a matching list of aliases to generate `ARRAY JOIN col1 AS a, col2 AS b, col3 AS c`. Single-column usage is unchanged. Closes [#633](https://github.com/ClickHouse/clickhouse-connect/issues/633)
+- SQLAlchemy: `ch_join()` now supports `USING` syntax via the new `using` parameter. Pass a list of column name strings to generate `USING (col1, col2)` instead of `ON`. This is important for `FULL OUTER JOIN` where `USING` merges the join column correctly while `ON` produces default values (0, '') for unmatched sides. Closes [#636](https://github.com/ClickHouse/clickhouse-connect/issues/636)
 - SQLAlchemy: Add missing Replicated table engine variants: `ReplicatedReplacingMergeTree`, `ReplicatedCollapsingMergeTree`, `ReplicatedVersionedCollapsingMergeTree`, and `ReplicatedGraphiteMergeTree`. Closes [#687](https://github.com/ClickHouse/clickhouse-connect/issues/687)
 
 ### Bug Fixes
+- SQLAlchemy: Fix `.final()` and `.sample()` silently overwriting each other when chained. Both methods now store modifiers as custom attributes on the `Select` instance and render them during compilation, replacing the previous `with_hint()` approach that only allowed one hint per table. Chaining in either order (e.g. `select(t).final().sample(0.1)`) correctly produces `FROM t FINAL SAMPLE 0.1`. Also fixes rendering for aliased tables (`FROM t AS u FINAL`) and supports explicit table targeting in joins. Fixes [#658](https://github.com/ClickHouse/clickhouse-connect/issues/658)
+- SQLAlchemy: Fix `sqlalchemy.values()` to generate ClickHouse's `VALUES` table function syntax. The compiler now emits `VALUES('col1 Type1, col2 Type2', ...)` with the column structure as the first argument, instead of the standard SQL form that places column names after the alias. Generic SQLAlchemy types are mapped to ClickHouse equivalents (e.g. `Integer` to `Int32`, `String` to `String`). Also handles CTE usage by wrapping in `SELECT * FROM VALUES(...)`. Fixes [#681](https://github.com/ClickHouse/clickhouse-connect/issues/681)
 - SQLAlchemy: Fix `GraphiteMergeTree` and `ReplicatedGraphiteMergeTree` to properly single-quote the `config_section` argument as ClickHouse requires.
 
 ## 0.14.1, 2026-03-11

README.md

@@ -32,16 +32,23 @@ When creating a Superset Data Source, either use the provided connection dialog,
 ### SQLAlchemy Implementation
 
 ClickHouse Connect includes a lightweight SQLAlchemy dialect implementation focused on compatibility with **Superset**
-and **SQLAlchemy Core**.
+and **SQLAlchemy Core**. Both SQLAlchemy 1.4 and 2.x are supported. SQLAlchemy 1.4 compatibility is maintained
+because Apache Superset currently requires `sqlalchemy>=1.4,<2`.
 
 Supported features include:
 - Basic query execution via SQLAlchemy Core
-- `SELECT` queries with `JOIN`s, `ARRAY JOIN`, and `FINAL` modifier
+- `SELECT` queries with `JOIN`s (including ClickHouse-specific strictness, `USING`, and `GLOBAL` modifiers),
+  `ARRAY JOIN` (single and multi-column), `FINAL`, and `SAMPLE`
+- `VALUES` table function syntax
 - Lightweight `DELETE` statements
 
-The implementation does not include ORM support and is not intended as a full SQLAlchemy dialect. While it can support
-a range of Core-based applications beyond Superset, it may not be suitable for more complex SQLAlchemy applications
-that rely on full ORM or advanced dialect functionality.
+A small number of features require SQLAlchemy 2.x: `Values.cte()` and certain literal-rendering behaviors.
+All other dialect features, including those used by Superset, work on both 1.4 and 2.x.
+
+Basic ORM usage works for insert-heavy, read-focused workloads: declarative model definitions, `CREATE TABLE`,
+`session.add()`, `bulk_save_objects()`, and read queries all function correctly. However, full ORM support is not
+provided. UPDATE compilation, foreign key/relationship reflection, autoincrement/RETURNING, and cascade operations
+are not implemented. The dialect is best suited for SQLAlchemy Core usage and Superset connectivity.
 
 ### Asyncio Support
 

clickhouse_connect/cc_sqlalchemy/__init__.py

@@ -1,10 +1,10 @@
 from clickhouse_connect import driver_name
 from clickhouse_connect.cc_sqlalchemy.datatypes.base import schema_types
-from clickhouse_connect.cc_sqlalchemy.sql import final
-from clickhouse_connect.cc_sqlalchemy.sql.clauses import array_join, ArrayJoin
+from clickhouse_connect.cc_sqlalchemy.sql import final, sample
+from clickhouse_connect.cc_sqlalchemy.sql.clauses import array_join, ArrayJoin, ch_join, ClickHouseJoin
 
 # pylint: disable=invalid-name
 dialect_name = driver_name
 ischema_names = schema_types
 
-__all__ = ['dialect_name', 'ischema_names', 'array_join', 'ArrayJoin', 'final']
+__all__ = ['dialect_name', 'ischema_names', 'array_join', 'ArrayJoin', 'ch_join', 'ClickHouseJoin', 'final', 'sample']

clickhouse_connect/cc_sqlalchemy/sql/__init__.py

@@ -5,6 +5,10 @@
 
 from clickhouse_connect.driver.binding import quote_identifier
 
+# Dialect name used for non-rendering statement hints that only serve to
+# differentiate cache keys when FINAL/SAMPLE modifiers are applied.
+_CH_MODIFIER_DIALECT = "_ch_modifier"
+
 
 def full_table(table_name: str, schema: Optional[str] = None) -> str:
     if table_name.startswith('(') or '.' in table_name or not schema:
@@ -16,38 +20,61 @@ def format_table(table: Table):
     return full_table(table.name, table.schema)
 
 
-def final(select_stmt: Select, table: Optional[FromClause] = None) -> Select:
-    """
-    Apply the ClickHouse FINAL modifier to a select statement.
-
-    Args:
-        select_stmt: The SQLAlchemy Select statement to modify.
-        table: Optional explicit table/alias to apply FINAL to. When omitted the
-            method will use the single FROM element present on the select. A
-            ValueError is raised if the statement has no FROMs or more than one
-            FROM element and table is not provided.
-
-    Returns:
-        A new Select that renders the FINAL modifier for the target table.
-    """
+def _resolve_target(select_stmt: Select, table: Optional[FromClause], method_name: str) -> FromClause:
+    """Resolve the target FROM clause for ClickHouse modifiers (FINAL/SAMPLE)."""
     if not isinstance(select_stmt, Select):
-        raise TypeError("final() expects a SQLAlchemy Select instance")
+        raise TypeError(f"{method_name}() expects a SQLAlchemy Select instance")
 
     target = table
     if target is None:
         froms = select_stmt.get_final_froms()
         if not froms:
-            raise ValueError("final() requires a table to apply the FINAL modifier.")
+            raise ValueError(f"{method_name}() requires a table to apply the {method_name.upper()} modifier.")
         if len(froms) > 1:
             raise ValueError(
-                "final() is ambiguous for statements with multiple FROM clauses. Specify the table explicitly."
+                f"{method_name}() is ambiguous for statements with multiple FROM clauses. "
+                "Specify the table explicitly."
             )
         target = froms[0]
 
     if not isinstance(target, FromClause):
         raise TypeError("table must be a SQLAlchemy FromClause when provided")
 
-    return select_stmt.with_hint(target, "FINAL")
+    return target
+
+
+def _target_cache_key(target: FromClause) -> str:
+    """Stable string identifying a FROM target for cache key differentiation."""
+    if hasattr(target, "fullname"):
+        return target.fullname
+    return target.name
+
+
+# pylint: disable=protected-access
+def final(select_stmt: Select, table: Optional[FromClause] = None) -> Select:
+    """Apply the ClickHouse FINAL modifier to a select statement.
+
+    FINAL forces ClickHouse to merge data parts before returning results,
+    guaranteeing fully collapsed rows for ReplacingMergeTree, CollapsingMergeTree,
+    and similar engines.
+
+    Args:
+        select_stmt: The SELECT statement to modify.
+        table: The target table to apply FINAL to. Required when the query
+            joins multiple tables, optional when there is a single FROM target.
+    """
+    target = _resolve_target(select_stmt, table, "final")
+    ch_final = getattr(select_stmt, "_ch_final", set())
+
+    if target in ch_final:
+        return select_stmt
+
+    # with_statement_hint creates a generative copy and adds a non-rendering
+    # hint that participates in the statement cache key.
+    hint_key = _target_cache_key(target)
+    new_stmt = select_stmt.with_statement_hint(f"FINAL:{hint_key}", dialect_name=_CH_MODIFIER_DIALECT)
+    new_stmt._ch_final = ch_final | {target}
+    return new_stmt
 
 
 def _select_final(self: Select, table: Optional[FromClause] = None) -> Select:
@@ -58,39 +85,27 @@ def _select_final(self: Select, table: Optional[FromClause] = None) -> Select:
 
 
 def sample(select_stmt: Select, sample_value: Union[str, int, float], table: Optional[FromClause] = None) -> Select:
-    """
-    Apply ClickHouse SAMPLE clause to a select statement.
-    Reference: https://clickhouse.com/docs/sql-reference/statements/select/sample
+    """Apply the ClickHouse SAMPLE modifier to a select statement.
+
     Args:
-        select_stmt: The SQLAlchemy Select statement to modify.
-        sample_value: Controls the sampling behavior. Accepts three forms:
-            - A float in (0, 1) for proportional sampling (e.g., 0.1 for ~10% of data).
-            - A positive integer for row-count sampling (e.g., 10000000 for ~10M rows).
-            - A string for fraction or offset notation (e.g., "1/10" or "1/10 OFFSET 1/2").
-        table: Optional explicit table to apply SAMPLE to. When omitted the
-            method will use the single FROM element present on the select. A
-            ValueError is raised if the statement has no FROMs or more than one
-            FROM element and table is not provided.
-
-    Returns:
-        A new Select that renders the SAMPLE clause for the target table.
+        select_stmt: The SELECT statement to modify.
+        sample_value: The sample expression. Can be a float between 0 and 1
+            for a fractional sample (e.g. 0.1 for 10%), an integer for an
+            approximate row count, or a string for SAMPLE expressions like
+            '1/10 OFFSET 1/2'.
+        table: The target table to sample. Required when the query joins
+            multiple tables, optional when there is a single FROM target.
     """
-    if not isinstance(select_stmt, Select):
-        raise TypeError("sample() expects a SQLAlchemy Select instance")
-
-    target_table = table
-    if target_table is None:
-        froms = select_stmt.get_final_froms()
-        if not froms:
-            raise ValueError("sample() requires a FROM clause to apply the SAMPLE modifier.")
-        if len(froms) > 1:
-            raise ValueError("sample() is ambiguous for statements with multiple FROM clauses. Specify the table explicitly.")
-        target_table = froms[0]
-
-    if not isinstance(target_table, FromClause):
-        raise TypeError("table must be a SQLAlchemy FromClause when provided")
-
-    return select_stmt.with_hint(target_table, f"SAMPLE {sample_value}")
+    target = _resolve_target(select_stmt, table, "sample")
+
+    hint_key = _target_cache_key(target)
+    new_stmt = select_stmt.with_statement_hint(
+        f"SAMPLE:{hint_key}:{sample_value}", dialect_name=_CH_MODIFIER_DIALECT
+    )
+    ch_sample = dict(getattr(select_stmt, "_ch_sample", {}))
+    ch_sample[target] = sample_value
+    new_stmt._ch_sample = ch_sample
+    return new_stmt
 
 
 def _select_sample(self: Select, sample_value: Union[str, int, float], table: Optional[FromClause] = None) -> Select: