refactor: add agg_ops.QcutOp to the sqlglot compiler

Author: chelsea-lin

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

@@ -400,6 +400,43 @@ def _(
     return apply_window_if_present(expr, window)
 
 
+@UNARY_OP_REGISTRATION.register(agg_ops.QcutOp)
+def _(
+    op: agg_ops.QcutOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    percent_ranks_order_by = sge.Ordered(this=column.expr, desc=False)
+    percent_ranks = apply_window_if_present(
+        sge.func("PERCENT_RANK"),
+        window,
+        include_framing_clauses=False,
+        order_by_override=[percent_ranks_order_by],
+    )
+    if isinstance(op.quantiles, int):
+        scaled_rank = percent_ranks * sge.convert(op.quantiles)
+        # Calculate the 0-based bucket index.
+        bucket_index = sge.func("CEIL", scaled_rank) - sge.convert(1)
+        safe_bucket_index = sge.func("GREATEST", bucket_index, 0)
+
+        return sge.If(
+            this=sge.Is(this=column.expr, expression=sge.Null()),
+            true=sge.Null(),
+            false=sge.Cast(this=safe_bucket_index, to="INT64"),
+        )
+    else:
+        case = sge.Case()
+        first_quantile = sge.convert(op.quantiles[0])
+        case = case.when(
+            sge.LT(this=percent_ranks, expression=first_quantile), sge.Null()
+        )
+        for bucket_n in range(len(op.quantiles) - 1):
+            quantile = sge.convert(op.quantiles[bucket_n + 1])
+            bucket = sge.convert(bucket_n)
+            case = case.when(sge.LTE(this=percent_ranks, expression=quantile), bucket)
+        return case.else_(sge.Null())
+
+
 @UNARY_OP_REGISTRATION.register(agg_ops.QuantileOp)
 def _(
     op: agg_ops.QuantileOp,

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

@@ -26,6 +26,7 @@ def apply_window_if_present(
     value: sge.Expression,
     window: typing.Optional[window_spec.WindowSpec] = None,
     include_framing_clauses: bool = True,
+    order_by_override: typing.Optional[typing.List[sge.Ordered]] = None,
 ) -> sge.Expression:
     if window is None:
         return value
@@ -44,7 +45,11 @@ def apply_window_if_present(
     else:
         order_by = get_window_order_by(window.ordering)
 
-    order = sge.Order(expressions=order_by) if order_by else None
+    order = None
+    if order_by_override is not None and len(order_by_override) > 0:
+        order = sge.Order(expressions=order_by_override)
+    elif order_by:
+        order = sge.Order(expressions=order_by)
 
     group_by = (
         [

bigframes/core/compile/sqlglot/sqlglot_ir.py

@@ -637,7 +637,12 @@ def _select_to_cte(expr: sge.Select, cte_name: sge.Identifier) -> sge.Select:
 
 
 def _literal(value: typing.Any, dtype: dtypes.Dtype) -> sge.Expression:
-    sqlglot_type = sgt.from_bigframes_dtype(dtype)
+    sqlglot_type = sgt.from_bigframes_dtype(dtype) if dtype else None
+    if sqlglot_type is None:
+        if value is not None:
+            raise ValueError("Cannot infer SQLGlot type from None dtype.")
+        return sge.Null()
+
     if value is None:
         return _cast(sge.Null(), sqlglot_type)
     elif dtype == dtypes.BYTES_DTYPE:

bigframes/core/reshape/tile.py

@@ -22,8 +22,6 @@
 import pandas as pd
 
 import bigframes
-import bigframes.constants
-import bigframes.core.expression as ex
 import bigframes.core.ordering as order
 import bigframes.core.utils as utils
 import bigframes.core.window_spec as window_specs
@@ -165,7 +163,6 @@ def qcut(
             f"Only duplicates='drop' is supported in BigQuery DataFrames so far. {constants.FEEDBACK_LINK}"
         )
     block = x._block
-    label = block.col_id_to_label[x._value_column]
     block, nullity_id = block.apply_unary_op(x._value_column, ops.notnull_op)
     block, result = block.apply_window_op(
         x._value_column,
@@ -175,9 +172,6 @@ def qcut(
             ordering=(order.ascending_over(x._value_column),),
         ),
     )
-    block, result = block.project_expr(
-        ops.where_op.as_expr(result, nullity_id, ex.const(None)), label=label
-    )
     return bigframes.series.Series(block.select_column(result))
 
 

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

@@ -435,6 +435,22 @@ def test_pop_var(scalar_types_df: bpd.DataFrame, snapshot):
     snapshot.assert_match(sql_window, "window_out.sql")
 
 
+def test_qcut(scalar_types_df: bpd.DataFrame, snapshot):
+    col_name = "int64_col"
+    bf = scalar_types_df[[col_name]]
+    bf["qcut_w_int"] = bpd.qcut(bf[col_name], q=4, labels=False, duplicates="drop")
+
+    q_list = tuple([0, 0.25, 0.5, 0.75, 1])
+    bf["qcut_w_list"] = bpd.qcut(
+        scalar_types_df[col_name],
+        q=q_list,
+        labels=False,
+        duplicates="drop",
+    )
+
+    snapshot.assert_match(bf.sql, "out.sql")
+
+
 def test_quantile(scalar_types_df: bpd.DataFrame, snapshot):
     col_name = "int64_col"
     bf_df = scalar_types_df[[col_name]]