refactor: support agg_ops.LastOp, LastNonNullOp in the sqlglot compiler

Author: chelsea-lin

bigframes/core/compile/sqlglot/aggregations/unary_compiler.py

@@ -104,7 +104,9 @@ def _(
     column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
-    return apply_window_if_present(sge.func("DENSE_RANK"), window, include_framing_clauses=False)
+    return apply_window_if_present(
+        sge.func("DENSE_RANK"), window, include_framing_clauses=False
+    )
 
 
 @UNARY_OP_REGISTRATION.register(agg_ops.FirstOp)
@@ -128,6 +130,27 @@ def _(
     )
 
 
+@UNARY_OP_REGISTRATION.register(agg_ops.LastOp)
+def _(
+    op: agg_ops.LastOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    # LAST_VALUE in BQ respects nulls by default.
+    return apply_window_if_present(sge.LastValue(this=column.expr), window)
+
+
+@UNARY_OP_REGISTRATION.register(agg_ops.LastNonNullOp)
+def _(
+    op: agg_ops.LastNonNullOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    return apply_window_if_present(
+        sge.IgnoreNulls(this=sge.LastValue(this=column.expr)), window
+    )
+
+
 @UNARY_OP_REGISTRATION.register(agg_ops.MaxOp)
 def _(
     op: agg_ops.MaxOp,
@@ -203,7 +226,9 @@ def _(
     column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
-    return apply_window_if_present(sge.func("RANK"), window, include_framing_clauses=False)
+    return apply_window_if_present(
+        sge.func("RANK"), window, include_framing_clauses=False
+    )
 
 
 @UNARY_OP_REGISTRATION.register(agg_ops.SizeUnaryOp)

tests/unit/core/compile/sqlglot/aggregations/snapshots/test_unary_compiler/test_last/out.sql

@@ -0,0 +1,20 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `int64_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    CASE
+      WHEN `bfcol_0` IS NULL
+      THEN NULL
+      ELSE LAST_VALUE(`bfcol_0`) OVER (
+        ORDER BY `bfcol_0` IS NULL ASC NULLS LAST, `bfcol_0` ASC NULLS LAST
+        ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING
+      )
+    END AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `agg_int64`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/aggregations/snapshots/test_unary_compiler/test_last_non_null/out.sql

@@ -0,0 +1,16 @@
+WITH `bfcte_0` AS (
+  SELECT
+    `int64_col` AS `bfcol_0`
+  FROM `bigframes-dev`.`sqlglot_test`.`scalar_types`
+), `bfcte_1` AS (
+  SELECT
+    *,
+    LAST_VALUE(`bfcol_0` IGNORE NULLS) OVER (
+      ORDER BY `bfcol_0` IS NULL ASC NULLS LAST, `bfcol_0` ASC NULLS LAST
+      ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING
+    ) AS `bfcol_1`
+  FROM `bfcte_0`
+)
+SELECT
+  `bfcol_1` AS `agg_int64`
+FROM `bfcte_1`

tests/unit/core/compile/sqlglot/aggregations/test_unary_compiler.py

@@ -148,6 +148,28 @@ def test_first_non_null(scalar_types_df: bpd.DataFrame, snapshot):
     snapshot.assert_match(sql, "out.sql")
 
 
+def test_last(scalar_types_df: bpd.DataFrame, snapshot):
+    col_name = "int64_col"
+    bf_df = scalar_types_df[[col_name]]
+    agg_expr = agg_exprs.UnaryAggregation(agg_ops.LastOp(), expression.deref(col_name))
+    window = window_spec.WindowSpec(ordering=(ordering.ascending_over(col_name),))
+    sql = _apply_unary_window_op(bf_df, agg_expr, window, "agg_int64")
+
+    snapshot.assert_match(sql, "out.sql")
+
+
+def test_last_non_null(scalar_types_df: bpd.DataFrame, snapshot):
+    col_name = "int64_col"
+    bf_df = scalar_types_df[[col_name]]
+    agg_expr = agg_exprs.UnaryAggregation(
+        agg_ops.LastNonNullOp(), expression.deref(col_name)
+    )
+    window = window_spec.WindowSpec(ordering=(ordering.ascending_over(col_name),))
+    sql = _apply_unary_window_op(bf_df, agg_expr, window, "agg_int64")
+
+    snapshot.assert_match(sql, "out.sql")
+
+
 def test_max(scalar_types_df: bpd.DataFrame, snapshot):
     col_name = "int64_col"
     bf_df = scalar_types_df[[col_name]]