Merge branch 'main' into main_chelsealin_bool

Author: chelsea-lin

CHANGELOG.md

@@ -4,6 +4,27 @@
 
 [1]: https://pypi.org/project/bigframes/#history
 
+## [2.22.0](https://github.com/googleapis/python-bigquery-dataframes/compare/v2.21.0...v2.22.0) (2025-09-25)
+
+
+### Features
+
+* Add `GroupBy.__iter__` ([#1394](https://github.com/googleapis/python-bigquery-dataframes/issues/1394)) ([c56a78c](https://github.com/googleapis/python-bigquery-dataframes/commit/c56a78cd509a535d4998d5b9a99ec3ecd334b883))
+* Add ai.generate_int to bigframes.bigquery package ([#2109](https://github.com/googleapis/python-bigquery-dataframes/issues/2109)) ([af6b862](https://github.com/googleapis/python-bigquery-dataframes/commit/af6b862de5c3921684210ec169338815f45b19dd))
+* Add Groupby.describe() ([#2088](https://github.com/googleapis/python-bigquery-dataframes/issues/2088)) ([328a765](https://github.com/googleapis/python-bigquery-dataframes/commit/328a765e746138806a021bea22475e8c03512aeb))
+* Implement `Index.to_list()` ([#2106](https://github.com/googleapis/python-bigquery-dataframes/issues/2106)) ([60056ca](https://github.com/googleapis/python-bigquery-dataframes/commit/60056ca06511f99092647fe55fc02eeab486b4ca))
+* Implement inplace parameter for `DataFrame.drop` ([#2105](https://github.com/googleapis/python-bigquery-dataframes/issues/2105)) ([3487f13](https://github.com/googleapis/python-bigquery-dataframes/commit/3487f13d12e34999b385c2e11551b5e27bfbf4ff))
+* Support callable for series map method ([#2100](https://github.com/googleapis/python-bigquery-dataframes/issues/2100)) ([ac25618](https://github.com/googleapis/python-bigquery-dataframes/commit/ac25618feed2da11fe4fb85058d498d262c085c0))
+* Support df.info() with null index ([#2094](https://github.com/googleapis/python-bigquery-dataframes/issues/2094)) ([fb81eea](https://github.com/googleapis/python-bigquery-dataframes/commit/fb81eeaf13af059f32cb38e7f117fb3504243d51))
+
+
+### Bug Fixes
+
+* Avoid ibis fillna warning in compiler ([#2113](https://github.com/googleapis/python-bigquery-dataframes/issues/2113)) ([7ef667b](https://github.com/googleapis/python-bigquery-dataframes/commit/7ef667b0f46f13bcc8ad4f2ed8f81278132b5aec))
+* Negative start and stop parameter values in Series.str.slice() ([#2104](https://github.com/googleapis/python-bigquery-dataframes/issues/2104)) ([f57a348](https://github.com/googleapis/python-bigquery-dataframes/commit/f57a348f1935a4e2bb14c501bb4c47cd552d102a))
+* Throw type error for incomparable join keys ([#2098](https://github.com/googleapis/python-bigquery-dataframes/issues/2098)) ([9dc9695](https://github.com/googleapis/python-bigquery-dataframes/commit/9dc96959a84b751d18b290129c2926df6e50b3f5))
+* Transformers with non-standard column names throw errors ([#2089](https://github.com/googleapis/python-bigquery-dataframes/issues/2089)) ([a2daa3f](https://github.com/googleapis/python-bigquery-dataframes/commit/a2daa3fffe6743327edb9f4c74db93198bd12f8e))
+
 ## [2.21.0](https://github.com/googleapis/python-bigquery-dataframes/compare/v2.20.0...v2.21.0) (2025-09-17)
 
 

bigframes/bigquery/_operations/ai.py

@@ -188,6 +188,81 @@ def generate_int(
     return series_list[0]._apply_nary_op(operator, series_list[1:])
 
 
+@log_adapter.method_logger(custom_base_name="bigquery_ai")
+def generate_double(
+    prompt: PROMPT_TYPE,
+    *,
+    connection_id: str | None = None,
+    endpoint: str | None = None,
+    request_type: Literal["dedicated", "shared", "unspecified"] = "unspecified",
+    model_params: Mapping[Any, Any] | None = None,
+) -> series.Series:
+    """
+    Returns the AI analysis based on the prompt, which can be any combination of text and unstructured data.
+
+    **Examples:**
+
+        >>> import bigframes.pandas as bpd
+        >>> import bigframes.bigquery as bbq
+        >>> bpd.options.display.progress_bar = None
+        >>> animal = bpd.Series(["Kangaroo", "Rabbit", "Spider"])
+        >>> bbq.ai.generate_double(("How many legs does a ", animal, " have?"))
+        0    {'result': 2.0, 'full_response': '{"candidates...
+        1    {'result': 4.0, 'full_response': '{"candidates...
+        2    {'result': 8.0, 'full_response': '{"candidates...
+        dtype: struct<result: double, full_response: extension<dbjson<JSONArrowType>>, status: string>[pyarrow]
+
+        >>> bbq.ai.generate_double(("How many legs does a ", animal, " have?")).struct.field("result")
+        0    2.0
+        1    4.0
+        2    8.0
+        Name: result, dtype: Float64
+
+    Args:
+        prompt (Series | List[str|Series] | Tuple[str|Series, ...]):
+            A mixture of Series and string literals that specifies the prompt to send to the model. The Series can be BigFrames Series
+            or pandas Series.
+        connection_id (str, optional):
+            Specifies the connection to use to communicate with the model. For example, `myproject.us.myconnection`.
+            If not provided, the connection from the current session will be used.
+        endpoint (str, optional):
+            Specifies the Vertex AI endpoint to use for the model. For example `"gemini-2.5-flash"`. You can specify any
+            generally available or preview Gemini model. If you specify the model name, BigQuery ML automatically identifies and
+            uses the full endpoint of the model. If you don't specify an ENDPOINT value, BigQuery ML selects a recent stable
+            version of Gemini to use.
+        request_type (Literal["dedicated", "shared", "unspecified"]):
+            Specifies the type of inference request to send to the Gemini model. The request type determines what quota the request uses.
+            * "dedicated": function only uses Provisioned Throughput quota. The function returns the error Provisioned throughput is not
+            purchased or is not active if Provisioned Throughput quota isn't available.
+            * "shared": the function only uses dynamic shared quota (DSQ), even if you have purchased Provisioned Throughput quota.
+            * "unspecified": If you haven't purchased Provisioned Throughput quota, the function uses DSQ quota.
+            If you have purchased Provisioned Throughput quota, the function uses the Provisioned Throughput quota first.
+            If requests exceed the Provisioned Throughput quota, the overflow traffic uses DSQ quota.
+        model_params (Mapping[Any, Any]):
+            Provides additional parameters to the model. The MODEL_PARAMS value must conform to the generateContent request body format.
+
+    Returns:
+        bigframes.series.Series: A new struct Series with the result data. The struct contains these fields:
+        * "result": an DOUBLE value containing the model's response to the prompt. The result is None if the request fails or is filtered by responsible AI.
+        * "full_response": a JSON value containing the response from the projects.locations.endpoints.generateContent call to the model.
+        The generated text is in the text element.
+        * "status": a STRING value that contains the API response status for the corresponding row. This value is empty if the operation was successful.
+    """
+
+    prompt_context, series_list = _separate_context_and_series(prompt)
+    assert len(series_list) > 0
+
+    operator = ai_ops.AIGenerateDouble(
+        prompt_context=tuple(prompt_context),
+        connection_id=_resolve_connection_id(series_list[0], connection_id),
+        endpoint=endpoint,
+        request_type=request_type,
+        model_params=json.dumps(model_params) if model_params else None,
+    )
+
+    return series_list[0]._apply_nary_op(operator, series_list[1:])
+
+
 def _separate_context_and_series(
     prompt: PROMPT_TYPE,
 ) -> Tuple[List[str | None], List[series.Series]]:

bigframes/core/compile/ibis_compiler/scalar_op_registry.py

@@ -1986,7 +1986,7 @@ def ai_generate_bool(
 
 @scalar_op_compiler.register_nary_op(ops.AIGenerateInt, pass_op=True)
 def ai_generate_int(
-    *values: ibis_types.Value, op: ops.AIGenerateBool
+    *values: ibis_types.Value, op: ops.AIGenerateInt
 ) -> ibis_types.StructValue:
 
     return ai_ops.AIGenerateInt(
@@ -1998,6 +1998,20 @@ def ai_generate_int(
     ).to_expr()
 
 
+@scalar_op_compiler.register_nary_op(ops.AIGenerateDouble, pass_op=True)
+def ai_generate_double(
+    *values: ibis_types.Value, op: ops.AIGenerateDouble
+) -> ibis_types.StructValue:
+
+    return ai_ops.AIGenerateDouble(
+        _construct_prompt(values, op.prompt_context),  # type: ignore
+        op.connection_id,  # type: ignore
+        op.endpoint,  # type: ignore
+        op.request_type.upper(),  # type: ignore
+        op.model_params,  # type: ignore
+    ).to_expr()
+
+
 def _construct_prompt(
     col_refs: tuple[ibis_types.Value], prompt_context: tuple[str | None]
 ) -> ibis_types.StructValue:

bigframes/core/compile/sqlglot/aggregate_compiler.py

@@ -63,7 +63,7 @@ def compile_analytic(
     window: window_spec.WindowSpec,
 ) -> sge.Expression:
     if isinstance(aggregate, agg_expressions.NullaryAggregation):
-        return nullary_compiler.compile(aggregate.op)
+        return nullary_compiler.compile(aggregate.op, window)
     if isinstance(aggregate, agg_expressions.UnaryAggregation):
         column = typed_expr.TypedExpr(
             scalar_compiler.scalar_op_compiler.compile_expression(aggregate.arg),

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

@@ -20,6 +20,7 @@
 
 from bigframes.core import window_spec
 import bigframes.core.compile.sqlglot.aggregations.op_registration as reg
+from bigframes.core.compile.sqlglot.aggregations.windows import apply_window_if_present
 import bigframes.core.compile.sqlglot.expressions.typed_expr as typed_expr
 from bigframes.operations import aggregations as agg_ops
 
@@ -33,3 +34,25 @@ def compile(
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
     return BINARY_OP_REGISTRATION[op](op, left, right, window=window)
+
+
+@BINARY_OP_REGISTRATION.register(agg_ops.CorrOp)
+def _(
+    op: agg_ops.CorrOp,
+    left: typed_expr.TypedExpr,
+    right: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    result = sge.func("CORR", left.expr, right.expr)
+    return apply_window_if_present(result, window)
+
+
+@BINARY_OP_REGISTRATION.register(agg_ops.CovOp)
+def _(
+    op: agg_ops.CovOp,
+    left: typed_expr.TypedExpr,
+    right: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    result = sge.func("COVAR_SAMP", left.expr, right.expr)
+    return apply_window_if_present(result, window)

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

@@ -39,3 +39,15 @@ def _(
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
     return apply_window_if_present(sge.func("COUNT", sge.convert(1)), window)
+
+
+@NULLARY_OP_REGISTRATION.register(agg_ops.RowNumberOp)
+def _(
+    op: agg_ops.RowNumberOp,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    result: sge.Expression = sge.func("ROW_NUMBER")
+    if window is None:
+        # ROW_NUMBER always needs an OVER clause.
+        return sge.Window(this=result)
+    return apply_window_if_present(result, window)

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

@@ -41,22 +41,16 @@ def arg_checker(*args, **kwargs):
                     )
                 return item(*args, **kwargs)
 
-            if hasattr(op, "name"):
-                key = typing.cast(str, op.name)
-                if key in self._registered_ops:
-                    raise ValueError(f"{key} is already registered")
-            else:
-                raise ValueError(f"The operator must have a 'name' attribute. Got {op}")
+            key = str(op)
+            if key in self._registered_ops:
+                raise ValueError(f"{key} is already registered")
             self._registered_ops[key] = item
             return arg_checker
 
         return decorator
 
     def __getitem__(self, op: str | agg_ops.WindowOp) -> CompilationFunc:
-        if isinstance(op, agg_ops.WindowOp):
-            if not hasattr(op, "name"):
-                raise ValueError(f"The operator must have a 'name' attribute. Got {op}")
-            else:
-                key = typing.cast(str, op.name)
-                return self._registered_ops[key]
-        return self._registered_ops[op]
+        key = op if isinstance(op, type) else type(op)
+        if str(key) not in self._registered_ops:
+            raise ValueError(f"{key} is already not registered")
+        return self._registered_ops[str(key)]

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

@@ -38,6 +38,37 @@ def compile(
     return UNARY_OP_REGISTRATION[op](op, column, window=window)
 
 
+@UNARY_OP_REGISTRATION.register(agg_ops.ApproxQuartilesOp)
+def _(
+    op: agg_ops.ApproxQuartilesOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    if window is not None:
+        raise NotImplementedError("Approx Quartiles with windowing is not supported.")
+    # APPROX_QUANTILES returns an array of the quartiles, so we need to index it.
+    # The op.quartile is 1-based for the quartile, but array is 0-indexed.
+    # The quartiles are Q0, Q1, Q2, Q3, Q4. op.quartile is 1, 2, or 3.
+    # The array has 5 elements (for N=4 intervals).
+    # So we want the element at index `op.quartile`.
+    approx_quantiles_expr = sge.func("APPROX_QUANTILES", column.expr, sge.convert(4))
+    return sge.Bracket(
+        this=approx_quantiles_expr,
+        expressions=[sge.func("OFFSET", sge.convert(op.quartile))],
+    )
+
+
+@UNARY_OP_REGISTRATION.register(agg_ops.ApproxTopCountOp)
+def _(
+    op: agg_ops.ApproxTopCountOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    if window is not None:
+        raise NotImplementedError("Approx top count with windowing is not supported.")
+    return sge.func("APPROX_TOP_COUNT", column.expr, sge.convert(op.number))
+
+
 @UNARY_OP_REGISTRATION.register(agg_ops.CountOp)
 def _(
     op: agg_ops.CountOp,
@@ -53,10 +84,7 @@ def _(
     column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
-    # Ranking functions do not support window framing clauses.
-    return apply_window_if_present(
-        sge.func("DENSE_RANK"), window, include_framing_clauses=False
-    )
+    return apply_window_if_present(sge.func("DENSE_RANK"), window)
 
 
 @UNARY_OP_REGISTRATION.register(agg_ops.MaxOp)
@@ -109,13 +137,23 @@ def _(
     return apply_window_if_present(sge.func("MIN", column.expr), window)
 
 
-@UNARY_OP_REGISTRATION.register(agg_ops.SizeUnaryOp)
+@UNARY_OP_REGISTRATION.register(agg_ops.QuantileOp)
 def _(
-    op: agg_ops.SizeUnaryOp,
-    _,
+    op: agg_ops.QuantileOp,
+    column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
-    return apply_window_if_present(sge.func("COUNT", sge.convert(1)), window)
+    # TODO: Support interpolation argument
+    # TODO: Support percentile_disc
+    result: sge.Expression = sge.func("PERCENTILE_CONT", column.expr, sge.convert(op.q))
+    if window is None:
+        # PERCENTILE_CONT is a navigation function, not an aggregate function, so it always needs an OVER clause.
+        result = sge.Window(this=result)
+    else:
+        result = apply_window_if_present(result, window)
+    if op.should_floor_result:
+        result = sge.Cast(this=sge.func("FLOOR", result), to="INT64")
+    return result
 
 
 @UNARY_OP_REGISTRATION.register(agg_ops.RankOp)
@@ -124,10 +162,16 @@ def _(
     column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
-    # Ranking functions do not support window framing clauses.
-    return apply_window_if_present(
-        sge.func("RANK"), window, include_framing_clauses=False
-    )
+    return apply_window_if_present(sge.func("RANK"), window)
+
+
+@UNARY_OP_REGISTRATION.register(agg_ops.SizeUnaryOp)
+def _(
+    op: agg_ops.SizeUnaryOp,
+    _,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    return apply_window_if_present(sge.func("COUNT", sge.convert(1)), window)
 
 
 @UNARY_OP_REGISTRATION.register(agg_ops.SumOp)

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

@@ -25,7 +25,6 @@
 def apply_window_if_present(
     value: sge.Expression,
     window: typing.Optional[window_spec.WindowSpec] = None,
-    include_framing_clauses: bool = True,
 ) -> sge.Expression:
     if window is None:
         return value
@@ -65,7 +64,7 @@ def apply_window_if_present(
     if not window.bounds and not order:
         return sge.Window(this=value, partition_by=group_by)
 
-    if not window.bounds and not include_framing_clauses:
+    if not window.bounds:
         return sge.Window(this=value, partition_by=group_by, order=order)
 
     kind = (

bigframes/core/compile/sqlglot/expressions/ai_ops.py

@@ -40,6 +40,13 @@ def _(*exprs: TypedExpr, op: ops.AIGenerateInt) -> sge.Expression:
     return sge.func("AI.GENERATE_INT", *args)
 
 
+@register_nary_op(ops.AIGenerateDouble, pass_op=True)
+def _(*exprs: TypedExpr, op: ops.AIGenerateDouble) -> sge.Expression:
+    args = [_construct_prompt(exprs, op.prompt_context)] + _construct_named_args(op)
+
+    return sge.func("AI.GENERATE_DOUBLE", *args)
+
+
 def _construct_prompt(
     exprs: tuple[TypedExpr, ...], prompt_context: tuple[str | None, ...]
 ) -> sge.Kwarg: