Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Convert built-in row_number to user-defined window function #12030

Merged
merged 48 commits into from
Aug 17, 2024

Conversation

jcsherin
Copy link
Contributor

Which issue does this PR close?

Closes #12029.

Rationale for this change

Same as described in #8709.

What changes are included in this PR?

  1. Adds new crate datafusion-functions-window.
  2. Converts BuiltinWindowFunction::RowNumber to user-defined window function.
  3. Extends WindowUDFImpl with two additional methods.
  4. Export a fluent-style API for creating row_number expressions.
  5. Implements BuiltinWindowExpr::get_result_ordering for WindowUDFExpr which is necessary for updating ordering equivalences.

Are these changes tested?

  1. Adds logical plan round trip test.
  2. Updates existing unit tests in row_number to use the new API.
  3. Updates explain output in sqllogictests to match lowercase row_number.

Are there any user-facing changes?

Yes.

@github-actions github-actions bot added sqllogictest SQL Logic Tests (.slt) proto Related to proto crate labels Aug 16, 2024
[package]
name = "datafusion-functions-window"
description = "Window function packages for the DataFusion query engine"
keywords = ["datafusion", "logical", "plan", "expressions"]
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Any suggestions on what keywords to place here?

Ok(Field::new(
&self.name,
self.data_type.clone(),
self.fun.nullable(),
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is why nullable method was added to WindowUDFImpl so that the user-defined window function can customize nullability of column values.


fn get_result_ordering(&self, schema: &SchemaRef) -> Option<PhysicalSortExpr> {
self.fun
.sort_options()
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The user-defined window function can specify its result ordering which will be used for updating ordering equivalences.

Comment on lines +484 to +485
UNSPECIFIED = 0; // https://protobuf.dev/programming-guides/dos-donts/#unspecified-enum
// ROW_NUMBER = 0;
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The proto enum requires the first value to be zero. See: https://protobuf.dev/programming-guides/dos-donts/#unspecified-enum

Leaving the older value commented out instead of removing it so we know that it was ROW_NUMBER. This will anyways be removed once all the other built-ins are moved.

@@ -903,6 +904,7 @@ async fn roundtrip_expr_api() -> Result<()> {
vec![lit(1), lit(2), lit(3)],
vec![lit(10), lit(20), lit(30)],
),
row_number(vec![col("a")]),
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Logical round trip test.

@@ -2660,14 +2660,14 @@ EXPLAIN SELECT
logical_plan
01)Limit: skip=0, fetch=5
02)--Sort: annotated_data_finite.ts DESC NULLS FIRST, fetch=5
03)----Projection: annotated_data_finite.ts, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fv1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fv2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lv1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lv2, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS nv1, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS nv2, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS rn1, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS rn2, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS rank1, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS rank2, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS dense_rank1, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS dense_rank2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lag1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lag2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lead1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lead2, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fvr1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fvr2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lvr1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lvr2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lagr1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lagr2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS leadr1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS leadr2
04)------WindowAggr: windowExpr=[[first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, NTH_VALUE(annotated_data_finite.inc_col, Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING, NTH_VALUE(annotated_data_finite.inc_col, Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, LAG(annotated_data_finite.inc_col, Int64(1), Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, LAG(annotated_data_finite.inc_col, Int64(2), Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, LEAD(annotated_data_finite.inc_col, Int64(-1), Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, LEAD(annotated_data_finite.inc_col, Int64(4), Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING]]
03)----Projection: annotated_data_finite.ts, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fv1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fv2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lv1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lv2, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS nv1, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS nv2, row_number() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS rn1, row_number() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS rn2, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS rank1, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS rank2, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS dense_rank1, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS dense_rank2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lag1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lag2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lead1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lead2, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fvr1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS fvr2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lvr1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lvr2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS lagr1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS lagr2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING AS leadr1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING AS leadr2
Copy link
Contributor Author

@jcsherin jcsherin Aug 16, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This diff is large and therefore I think the usual inline highlighting the words/phrases is disabled. The change here is from uppercase ROW_NUMBER to lowercase row_number.

05)--------WindowAggr: windowExpr=[[first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, LAG(annotated_data_finite.inc_col, Int64(1), Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, LAG(annotated_data_finite.inc_col, Int64(2), Int64(1002)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING, LEAD(annotated_data_finite.inc_col, Int64(-1), Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING, LEAD(annotated_data_finite.inc_col, Int64(4), Int64(1004)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING]]
06)----------TableScan: annotated_data_finite projection=[ts, inc_col]
physical_plan
01)SortExec: TopK(fetch=5), expr=[ts@0 DESC], preserve_partitioning=[false]
02)--ProjectionExec: expr=[ts@0 as ts, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@10 as fv1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@11 as fv2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@12 as lv1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@13 as lv2, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@14 as nv1, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@15 as nv2, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@16 as rn1, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@17 as rn2, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@18 as rank1, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@19 as rank2, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@20 as dense_rank1, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@21 as dense_rank2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@22 as lag1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@23 as lag2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@24 as lead1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@25 as lead2, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@2 as fvr1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@3 as fvr2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@4 as lvr1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@5 as lvr2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@6 as lagr1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@7 as lagr2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@8 as leadr1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@9 as leadr2]
03)----BoundedWindowAggExec: wdw=[first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(10)), end_bound: Following(Int32(1)), is_causal: false }, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(10)), end_bound: Following(Int32(1)), is_causal: false }, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(10)), end_bound: Following(Int32(1)), is_causal: false }, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING: Ok(Field { name: "ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(1)), end_bound: Following(Int32(10)), is_causal: false }, ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "ROW_NUMBER() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING: Ok(Field { name: "RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(1)), end_bound: Following(Int32(10)), is_causal: false }, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING: Ok(Field { name: "DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(1)), end_bound: Following(Int32(10)), is_causal: false }, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING: Ok(Field { name: "LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(1)), end_bound: Following(Int32(10)), is_causal: false }, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING: Ok(Field { name: "LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Range, start_bound: Preceding(Int32(1)), end_bound: Following(Int32(10)), is_causal: false }, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING: Ok(Field { name: "LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {} }), frame: WindowFrame { units: Rows, start_bound: Preceding(UInt64(10)), end_bound: Following(UInt64(1)), is_causal: false }], mode=[Sorted]
02)--ProjectionExec: expr=[ts@0 as ts, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@10 as fv1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@11 as fv2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@12 as lv1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@13 as lv2, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@14 as nv1, NTH_VALUE(annotated_data_finite.inc_col,Int64(5)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@15 as nv2, row_number() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@16 as rn1, row_number() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@17 as rn2, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@18 as rank1, RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@19 as rank2, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@20 as dense_rank1, DENSE_RANK() ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@21 as dense_rank2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@22 as lag1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@23 as lag2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@24 as lead1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts ASC NULLS LAST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@25 as lead2, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@2 as fvr1, first_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@3 as fvr2, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 10 PRECEDING AND 1 FOLLOWING@4 as lvr1, last_value(annotated_data_finite.inc_col) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@5 as lvr2, LAG(annotated_data_finite.inc_col,Int64(1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@6 as lagr1, LAG(annotated_data_finite.inc_col,Int64(2),Int64(1002)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@7 as lagr2, LEAD(annotated_data_finite.inc_col,Int64(-1),Int64(1001)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] RANGE BETWEEN 1 PRECEDING AND 10 FOLLOWING@8 as leadr1, LEAD(annotated_data_finite.inc_col,Int64(4),Int64(1004)) ORDER BY [annotated_data_finite.ts DESC NULLS FIRST] ROWS BETWEEN 10 PRECEDING AND 1 FOLLOWING@9 as leadr2]
Copy link
Contributor Author

@jcsherin jcsherin Aug 16, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same here as well. Change is from uppercase ROW_NUMBER to lowercase row_number.

@alamb
Copy link
Contributor

alamb commented Aug 16, 2024

It starts 😍

@jcsherin jcsherin changed the title Convert built-in row_number to user-defined window function Convert built-in row_number to user-defined window function Aug 16, 2024
Copy link
Contributor

@jayzhan211 jayzhan211 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great to see the progress on #8709 🚀


/// Create a [`WindowFunction`](Expr::WindowFunction) expression for
/// `row_number` user-defined window function.
pub fn row_number(args: Vec<Expr>) -> Expr {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We will have a macro create_func similar to aggregate later on, right?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes definitely.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

BTW I don't think row_number() takes arguments -- I removed the argument in #12043

This is the previous row_number implementation here https://github.com/apache/datafusion/pull/12030/files#diff-790fda7313189b10bd29f89102416b93f5708ab88fdd5c89733758e1ed2aafadL22

/// Create an expression to represent the `row_number` window function
pub fn row_number() -> Expr {
    Expr::WindowFunction(WindowFunction::new(
        BuiltInWindowFunction::RowNumber,
        vec![],
    ))
}

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@alamb Thank you for taking care of this in #12043.

&self.signature
}

fn return_type(&self, _arg_types: &[DataType]) -> Result<DataType> {
Copy link
Contributor

@jayzhan211 jayzhan211 Aug 16, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Instead of return_type + nullable. I think field is a better choice. Given windowudf is pretty new so not widely used yet (?). It would be nice to have field(&self, args: FieldArgs) -> Result<Field>

FieldArgs {
 arg_type: &[DataType]
 schema: Schema // for nullable
}

It is too late for AggregateUDF and ScalarUDF :(

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this makes sense.

All the existing built-in window functions define nullable within the BuiltInWindowFunctionExpr::field implementation.

Therefore we can define field like this (without FieldArgs struct):

fn field(&self, arg_types: &[DataType]) -> Result<Field>`

This will also eliminate data_type field from WindowUDFExpr :

/// Implements [`BuiltInWindowFunctionExpr`] for [`WindowUDF`]
#[derive(Clone, Debug)]
struct WindowUDFExpr {
fun: Arc<WindowUDF>,
args: Vec<Arc<dyn PhysicalExpr>>,
/// Display name
name: String,
/// result type
data_type: DataType,
}

Then the field method implementation:

impl BuiltInWindowFunctionExpr for WindowUDFExpr {
fn as_any(&self) -> &dyn std::any::Any {
self
}
fn field(&self) -> Result<Field> {
let nullable = true;
Ok(Field::new(&self.name, self.data_type.clone(), nullable))
}

Will instead be this:

impl BuiltInWindowFunctionExpr for WindowUDFExpr {

    fn field(&self) -> Result<Field> { self.fun.field() }

}

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd like to try unifying return type + nullable in a follow-on PR.

Copy link
Contributor

@jayzhan211 jayzhan211 Aug 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

With FieldArgs the benefit is that breaking change is not required if we find out we need other information in the future, such as Schema.

I'd like to try unifying return type + nullable in a follow-on PR.

Fine to me. I would be nice to have it included in the same release to avoid breaking change

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Forgot what I said about FieldArgs, if there are only two args at the end, it might be overkill.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fine to me. I would be nice to have it included in the same release to avoid breaking change

Will do.

Copy link
Contributor

@jayzhan211 jayzhan211 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍

@jcsherin
Copy link
Contributor Author

Thanks @jayzhan211, @alamb

@jayzhan211 jayzhan211 merged commit bd2d4ee into apache:main Aug 17, 2024
27 checks passed
@jayzhan211
Copy link
Contributor

Thanks @jcsherin @alamb

@jcsherin jcsherin deleted the window-udf-row-number branch August 21, 2024 07:24
emgeee added a commit to probably-nothing-labs/datafusion-python that referenced this pull request Sep 10, 2024
row_number was converted to a UDF in datafusion v42 apache/datafusion#12030
This specific functionality needs to be added back in.
timsaucer added a commit to apache/datafusion-python that referenced this pull request Sep 17, 2024
* update dependencies

* update get_logical_plan signature

* remove row_number() function

row_number was converted to a UDF in datafusion v42 apache/datafusion#12030
This specific functionality needs to be added back in.

* remove unneeded dependency

* fix pyo3 warnings

Implicit defaults for trailing optional arguments have been deprecated
in pyo3 v0.22.0 PyO3/pyo3#4078

* update object_store dependency

* change PyExpr -> PySortExpr

* comment out key.extract::<&PyTuple>() condition statement

* change more instances of PyExpr > PySortExpr

* update function signatures to use _bound versions

* remove clone

* Working through some of the sort requirement changes

* remove unused import

* expr.display_name is deprecated, used format!() + schema_name() instead

* expr.canonical_name() is deprecated, use format!() expr instead

* remove comment

* fix tuple extraction in dataframe.__getitem__()

* remove unneeded import

* Add docstring comments to SortExpr python class

* change extract() to downcast()

Co-authored-by: Michael J Ward <Michael-J-Ward@users.noreply.github.com>

* deprecate Expr::display_name

Ref: apache/datafusion#11797

* fix lint errors

* update datafusion commit hash

* fix type in cargo file for arrow features

* upgrade to datafusion 42

* cleanup

---------

Co-authored-by: Tim Saucer <timsaucer@gmail.com>
Co-authored-by: Michael J Ward <Michael-J-Ward@users.noreply.github.com>
Co-authored-by: Michael-J-Ward <ward.michael.j@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
core Core DataFusion crate logical-expr Logical plan and expressions physical-expr Physical Expressions proto Related to proto crate sql SQL Planner sqllogictest SQL Logic Tests (.slt)
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Convert built-in row_number to user-defined window function
3 participants