UDAF sum workaround

Cargo.toml

@@ -56,6 +56,7 @@ parking_lot = "0.12"
 regex-syntax = "0.8.1"
 syn = "2.0.67"
 url = "2.2"
+sqlparser = "0.47.0"
 
 [build-dependencies]
 pyo3-build-config = "0.21"

examples/tpch/_tests.py

@@ -74,7 +74,6 @@ def check_q17(df):
         ("q10_returned_item_reporting", "q10"),
         pytest.param(
             "q11_important_stock_identification", "q11", 
-            marks=pytest.mark.xfail # https://github.com/apache/datafusion-python/issues/730
         ),
         ("q12_ship_mode_order_priority", "q12"),
         ("q13_customer_distribution", "q13"),

src/common.rs

@@ -29,6 +29,7 @@ pub(crate) fn init_module(m: &Bound<'_, PyModule>) -> PyResult<()> {
     m.add_class::<data_type::DataTypeMap>()?;
     m.add_class::<data_type::PythonType>()?;
     m.add_class::<data_type::SqlType>()?;
+    m.add_class::<data_type::NullTreatment>()?;
     m.add_class::<schema::SqlTable>()?;
     m.add_class::<schema::SqlSchema>()?;
     m.add_class::<schema::SqlView>()?;

src/common/data_type.rs

@@ -757,3 +757,33 @@ pub enum SqlType {
     VARBINARY,
     VARCHAR,
 }
+
+/// Specifies Ignore / Respect NULL within window functions.
+/// For example
+/// `FIRST_VALUE(column2) IGNORE NULLS OVER (PARTITION BY column1)`
+#[allow(non_camel_case_types)]
+#[allow(clippy::upper_case_acronyms)]
+#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
+#[pyclass(name = "PythonType", module = "datafusion.common")]
+pub enum NullTreatment {
+    IGNORE_NULLS,
+    RESPECT_NULLS,
+}
+
+impl From<NullTreatment> for sqlparser::ast::NullTreatment {
+    fn from(null_treatment: NullTreatment) -> sqlparser::ast::NullTreatment {
+        match null_treatment {
+            NullTreatment::IGNORE_NULLS => sqlparser::ast::NullTreatment::IgnoreNulls,
+            NullTreatment::RESPECT_NULLS => sqlparser::ast::NullTreatment::RespectNulls,
+        }
+    }
+}
+
+impl From<sqlparser::ast::NullTreatment> for NullTreatment {
+    fn from(null_treatment: sqlparser::ast::NullTreatment) -> NullTreatment {
+        match null_treatment {
+            sqlparser::ast::NullTreatment::IgnoreNulls => NullTreatment::IGNORE_NULLS,
+            sqlparser::ast::NullTreatment::RespectNulls => NullTreatment::RESPECT_NULLS,
+        }
+    }
+}

src/functions.rs

@@ -17,6 +17,7 @@
 
 use pyo3::{prelude::*, wrap_pyfunction};
 
+use crate::common::data_type::NullTreatment;
 use crate::context::PySessionContext;
 use crate::errors::DataFusionError;
 use crate::expr::conditional_expr::PyCaseBuilder;
@@ -73,15 +74,15 @@ pub fn var(y: PyExpr) -> PyExpr {
 }
 
 #[pyfunction]
-#[pyo3(signature = (*args, distinct = false, filter = None, order_by = None))]
+#[pyo3(signature = (*args, distinct = false, filter = None, order_by = None, null_treatment = None))]
 pub fn first_value(
     args: Vec<PyExpr>,
     distinct: bool,
     filter: Option<PyExpr>,
     order_by: Option<Vec<PyExpr>>,
+    null_treatment: Option<NullTreatment>,
 ) -> PyExpr {
-    // TODO: allow user to select null_treatment
-    let null_treatment = None;
+    let null_treatment = null_treatment.map(Into::into);
     let args = args.into_iter().map(|x| x.expr).collect::<Vec<_>>();
     let order_by = order_by.map(|x| x.into_iter().map(|x| x.expr).collect::<Vec<_>>());
     functions_aggregate::expr_fn::first_value(
@@ -95,15 +96,15 @@ pub fn first_value(
 }
 
 #[pyfunction]
-#[pyo3(signature = (*args, distinct = false, filter = None, order_by = None))]
+#[pyo3(signature = (*args, distinct = false, filter = None, order_by = None, null_treatment = None))]
 pub fn last_value(
     args: Vec<PyExpr>,
     distinct: bool,
     filter: Option<PyExpr>,
     order_by: Option<Vec<PyExpr>>,
+    null_treatment: Option<NullTreatment>,
 ) -> PyExpr {
-    // TODO: allow user to select null_treatment
-    let null_treatment = None;
+    let null_treatment = null_treatment.map(Into::into);
     let args = args.into_iter().map(|x| x.expr).collect::<Vec<_>>();
     let order_by = order_by.map(|x| x.into_iter().map(|x| x.expr).collect::<Vec<_>>());
     functions_aggregate::expr_fn::last_value(
@@ -320,14 +321,20 @@ fn window(
     window_frame: Option<PyWindowFrame>,
     ctx: Option<PySessionContext>,
 ) -> PyResult<PyExpr> {
-    let fun = find_df_window_func(name).or_else(|| {
-        ctx.and_then(|ctx| {
-            ctx.ctx
-                .udaf(name)
-                .map(WindowFunctionDefinition::AggregateUDF)
-                .ok()
+    // workaround for https://github.com/apache/datafusion-python/issues/730
+    let fun = if name == "sum" {
+        let sum_udf = functions_aggregate::sum::sum_udaf();
+        Some(WindowFunctionDefinition::AggregateUDF(sum_udf))
+    } else {
+        find_df_window_func(name).or_else(|| {
+            ctx.and_then(|ctx| {
+                ctx.ctx
+                    .udaf(name)
+                    .map(WindowFunctionDefinition::AggregateUDF)
+                    .ok()
+            })
         })
-    });
+    };
     if fun.is_none() {
         return Err(DataFusionError::Common("window function not found".to_string()).into());
     }