Merge remote-tracking branch 'github/main' into polars_isin

Author: TrevorBergeron

CHANGELOG.md

@@ -4,6 +4,41 @@
 
 [1]: https://pypi.org/project/bigframes/#history
 
+## [2.16.0](https://github.com/googleapis/python-bigquery-dataframes/compare/v2.15.0...v2.16.0) (2025-08-20)
+
+
+### Features
+
+* Add `bigframes.pandas.options.display.precision` option ([#1979](https://github.com/googleapis/python-bigquery-dataframes/issues/1979)) ([15e6175](https://github.com/googleapis/python-bigquery-dataframes/commit/15e6175ec0aeb1b7b02d0bba9e8e1e018bd11c31))
+* Add level, inplace params to reset_index ([#1988](https://github.com/googleapis/python-bigquery-dataframes/issues/1988)) ([3446950](https://github.com/googleapis/python-bigquery-dataframes/commit/34469504b79a082d3380f9f25c597483aef2068a))
+* Add ML code samples from dbt blog post ([#1978](https://github.com/googleapis/python-bigquery-dataframes/issues/1978)) ([ebaa244](https://github.com/googleapis/python-bigquery-dataframes/commit/ebaa244a9eb7b87f7f9fd9c3bebe5c7db24cd013))
+* Add where, coalesce, fillna, casewhen, invert local impl ([#1976](https://github.com/googleapis/python-bigquery-dataframes/issues/1976)) ([f7f686c](https://github.com/googleapis/python-bigquery-dataframes/commit/f7f686cf85ab7e265d9c07ebc7f0cd59babc5357))
+* Adjust anywidget CSS to prevent overflow ([#1981](https://github.com/googleapis/python-bigquery-dataframes/issues/1981)) ([204f083](https://github.com/googleapis/python-bigquery-dataframes/commit/204f083a2f00fcc9fd1500dcd7a738eda3904d2f))
+* Format page number in table widget ([#1992](https://github.com/googleapis/python-bigquery-dataframes/issues/1992)) ([e83836e](https://github.com/googleapis/python-bigquery-dataframes/commit/e83836e8e1357f009f3f95666f1661bdbe0d3751))
+* Or, And, Xor can execute locally ([#1994](https://github.com/googleapis/python-bigquery-dataframes/issues/1994)) ([59c52a5](https://github.com/googleapis/python-bigquery-dataframes/commit/59c52a55ebea697855eb4c70529e226cc077141f))
+* Support callable bigframes function for dataframe where ([#1990](https://github.com/googleapis/python-bigquery-dataframes/issues/1990)) ([44c1ec4](https://github.com/googleapis/python-bigquery-dataframes/commit/44c1ec48cc4db1c4c9c15ec1fab43d4ef0758e56))
+* Support callable for series where method ([#2005](https://github.com/googleapis/python-bigquery-dataframes/issues/2005)) ([768b82a](https://github.com/googleapis/python-bigquery-dataframes/commit/768b82af96a5dd0c434edcb171036eb42cfb9b41))
+* When using `repr_mode = "anywidget"`, numeric values align right ([15e6175](https://github.com/googleapis/python-bigquery-dataframes/commit/15e6175ec0aeb1b7b02d0bba9e8e1e018bd11c31))
+
+
+### Bug Fixes
+
+* Address the packages issue for bigframes function ([#1991](https://github.com/googleapis/python-bigquery-dataframes/issues/1991)) ([68f1d22](https://github.com/googleapis/python-bigquery-dataframes/commit/68f1d22d5ed8457a5cabc7751ed1d178063dd63e))
+* Correct pypdf dependency specifier for remote PDF functions ([#1980](https://github.com/googleapis/python-bigquery-dataframes/issues/1980)) ([0bd5e1b](https://github.com/googleapis/python-bigquery-dataframes/commit/0bd5e1b3c004124d2100c3fbec2fbe1e965d1e96))
+* Enable default retries in calls to BQ Storage Read API ([#1985](https://github.com/googleapis/python-bigquery-dataframes/issues/1985)) ([f25d7bd](https://github.com/googleapis/python-bigquery-dataframes/commit/f25d7bd30800dffa65b6c31b0b7ac711a13d790f))
+* Fix the copyright year in dbt sample files ([#1996](https://github.com/googleapis/python-bigquery-dataframes/issues/1996)) ([fad5722](https://github.com/googleapis/python-bigquery-dataframes/commit/fad57223d129f0c95d0c6a066179bb66880edd06))
+
+
+### Performance Improvements
+
+* Faster session startup by defering anon dataset fetch ([#1982](https://github.com/googleapis/python-bigquery-dataframes/issues/1982)) ([2720c4c](https://github.com/googleapis/python-bigquery-dataframes/commit/2720c4cf070bf57a0930d7623bfc41d89cc053ee))
+
+
+### Documentation
+
+* Add examples of running bigframes in kaggle ([#2002](https://github.com/googleapis/python-bigquery-dataframes/issues/2002)) ([7d89d76](https://github.com/googleapis/python-bigquery-dataframes/commit/7d89d76976595b75cb0105fbe7b4f7ca2fdf49f2))
+* Remove preview warning from partial ordering mode sample notebook ([#1986](https://github.com/googleapis/python-bigquery-dataframes/issues/1986)) ([132e0ed](https://github.com/googleapis/python-bigquery-dataframes/commit/132e0edfe9f96c15753649d77fcb6edd0b0708a3))
+
 ## [2.15.0](https://github.com/googleapis/python-bigquery-dataframes/compare/v2.14.0...v2.15.0) (2025-08-11)
 
 

bigframes/core/compile/polars/compiler.py

@@ -198,6 +198,10 @@ def _(self, op: ops.ScalarOp, l_input: pl.Expr, r_input: pl.Expr) -> pl.Expr:
         def _(self, op: ops.ScalarOp, l_input: pl.Expr, r_input: pl.Expr) -> pl.Expr:
             return l_input | r_input
 
+        @compile_op.register(bool_ops.XorOp)
+        def _(self, op: ops.ScalarOp, l_input: pl.Expr, r_input: pl.Expr) -> pl.Expr:
+            return l_input ^ r_input
+
         @compile_op.register(num_ops.AddOp)
         def _(self, op: ops.ScalarOp, l_input: pl.Expr, r_input: pl.Expr) -> pl.Expr:
             return l_input + r_input

bigframes/core/compile/sqlglot/aggregate_compiler.py

@@ -15,7 +15,7 @@
 
 import sqlglot.expressions as sge
 
-from bigframes.core import expression
+from bigframes.core import expression, window_spec
 from bigframes.core.compile.sqlglot.aggregations import (
     binary_compiler,
     nullary_compiler,
@@ -56,3 +56,21 @@ def compile_aggregate(
         return binary_compiler.compile(aggregate.op, left, right)
     else:
         raise ValueError(f"Unexpected aggregation: {aggregate}")
+
+
+def compile_analytic(
+    aggregate: expression.Aggregation,
+    window: window_spec.WindowSpec,
+) -> sge.Expression:
+    if isinstance(aggregate, expression.NullaryAggregation):
+        return nullary_compiler.compile(aggregate.op)
+    if isinstance(aggregate, expression.UnaryAggregation):
+        column = typed_expr.TypedExpr(
+            scalar_compiler.compile_scalar_expression(aggregate.arg),
+            aggregate.arg.output_type,
+        )
+        return unary_compiler.compile(aggregate.op, column, window)
+    elif isinstance(aggregate, expression.BinaryAggregation):
+        raise NotImplementedError("binary analytic operations not yet supported")
+    else:
+        raise ValueError(f"Unexpected analytic operation: {aggregate}")

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

@@ -18,6 +18,7 @@
 
 import sqlglot.expressions as sge
 
+from bigframes import dtypes
 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
@@ -36,14 +37,26 @@ def compile(
     return UNARY_OP_REGISTRATION[op](op, column, window=window)
 
 
+@UNARY_OP_REGISTRATION.register(agg_ops.CountOp)
+def _(
+    op: agg_ops.CountOp,
+    column: typed_expr.TypedExpr,
+    window: typing.Optional[window_spec.WindowSpec] = None,
+) -> sge.Expression:
+    return apply_window_if_present(sge.func("COUNT", column.expr), window)
+
+
 @UNARY_OP_REGISTRATION.register(agg_ops.SumOp)
 def _(
     op: agg_ops.SumOp,
     column: typed_expr.TypedExpr,
     window: typing.Optional[window_spec.WindowSpec] = None,
 ) -> sge.Expression:
+    expr = column.expr
+    if column.dtype == dtypes.BOOL_DTYPE:
+        expr = sge.Cast(this=column.expr, to="INT64")
     # Will be null if all inputs are null. Pandas defaults to zero sum though.
-    expr = apply_window_if_present(sge.func("SUM", column.expr), window)
+    expr = apply_window_if_present(sge.func("SUM", expr), window)
     return sge.func("IFNULL", expr, ir._literal(0, column.dtype))
 
 

bigframes/core/compile/sqlglot/compiler.py

@@ -298,6 +298,75 @@ def compile_aggregate(
 
         return child.aggregate(aggregations, by_cols, tuple(dropna_cols))
 
+    @_compile_node.register
+    def compile_window(
+        self, node: nodes.WindowOpNode, child: ir.SQLGlotIR
+    ) -> ir.SQLGlotIR:
+        window_spec = node.window_spec
+        if node.expression.op.order_independent and window_spec.is_unbounded:
+            # notably percentile_cont does not support ordering clause
+            window_spec = window_spec.without_order()
+
+        window_op = aggregate_compiler.compile_analytic(node.expression, window_spec)
+
+        inputs: tuple[sge.Expression, ...] = tuple(
+            scalar_compiler.compile_scalar_expression(expression.DerefOp(column))
+            for column in node.expression.column_references
+        )
+
+        clauses: list[tuple[sge.Expression, sge.Expression]] = []
+        if node.expression.op.skips_nulls and not node.never_skip_nulls:
+            for column in inputs:
+                clauses.append((sge.Is(this=column, expression=sge.Null()), sge.Null()))
+
+        if window_spec.min_periods and len(inputs) > 0:
+            if node.expression.op.skips_nulls:
+                # Most operations do not count NULL values towards min_periods
+                not_null_columns = [
+                    sge.Not(this=sge.Is(this=column, expression=sge.Null()))
+                    for column in inputs
+                ]
+                # All inputs must be non-null for observation to count
+                if not not_null_columns:
+                    is_observation_expr: sge.Expression = sge.convert(True)
+                else:
+                    is_observation_expr = not_null_columns[0]
+                    for expr in not_null_columns[1:]:
+                        is_observation_expr = sge.And(
+                            this=is_observation_expr, expression=expr
+                        )
+                is_observation = ir._cast(is_observation_expr, "INT64")
+                observation_count = windows.apply_window_if_present(
+                    sge.func("SUM", is_observation), window_spec
+                )
+            else:
+                # Operations like count treat even NULLs as valid observations
+                # for the sake of min_periods notnull is just used to convert
+                # null values to non-null (FALSE) values to be counted.
+                is_observation = ir._cast(
+                    sge.Not(this=sge.Is(this=inputs[0], expression=sge.Null())),
+                    "INT64",
+                )
+                observation_count = windows.apply_window_if_present(
+                    sge.func("COUNT", is_observation), window_spec
+                )
+
+            clauses.append(
+                (
+                    observation_count < sge.convert(window_spec.min_periods),
+                    sge.Null(),
+                )
+            )
+        if clauses:
+            when_expressions = [sge.When(this=cond, true=res) for cond, res in clauses]
+            window_op = sge.Case(ifs=when_expressions, default=window_op)
+
+        # TODO: check if we can directly window the expression.
+        return child.window(
+            window_op=window_op,
+            output_column_id=node.output_name.sql,
+        )
+
 
 def _replace_unsupported_ops(node: nodes.BigFrameNode):
     node = nodes.bottom_up(node, rewrite.rewrite_slice)