Combine the logic of rank, dense_rank and percent_rank udwf to reduce duplications

[jatin510]

Which issue does this PR close?

Closes #.

Rationale for this change

The functionality of the rank, percent_rank and dense_rank is same.
Combining them into one

What changes are included in this PR?

Combining rank, dense_rank and percent_rank logic

Are these changes tested?

Are there any user-facing changes?

[jatin510]

@jcsherin while defining the other udwf like this, it is not working, giving an error during compile time.
Am I going in the right direction ?

Thanks !

[jcsherin]

The first argument needs to be a distinct identifier. I guess something like this will work,

define_udwf_and_expr!(BasicRank, rank, "...", Rank::basic);
define_udwf_and_expr!(PercentRank, rank, "...", Rank::percent_rank);
define_udwf_and_expr!(DenseRank, rank, "...", Rank::dense_rank);

[jcsherin]

The Rank.data_type field can be removed.

Suggested change

	let nullable = false;
	Ok(Field::new(
	field_args.name(),
	self.data_type.clone(),
	nullable,
	))
	let return_type = match self.rank_type {
	RankType::Basic | RankType::Dense => DataType::UInt64,
	RankType::Percent => DataType::Float64,
	};
	Ok(Field::new(
	field_args.name(),
	return_type.clone(),
	false,
	))

[jcsherin]

Minor: &self.name

[jcsherin]

Please apply this patch to fix cargo doc failing in CI.

Download fix_macro_intra_doc_links.patch

$ git apply fix_macro_intra_doc_links.patch

It will resolve these errors:

error: unresolved link to `PercentRank`
   --> datafusion/functions-window/src/macros.rs:305:21
    |
305 |               #[doc = " Create a [`WindowFunction`](datafusion_expr::Expr::WindowFunction) expression for"]
    |  _____________________^
306 | |             #[doc = concat!(" [`", stringify!($UDWF), "`] user-defined window function.")]
307 | |             #[doc = ""]
308 | |             #[doc = concat!(" ", $DOC)]
    | |______________________________________^

This is because PercentRank here is just an identifier for creating a unique UDWF and not something we can link to. So in the macro generated doc comments [PercentRank] is a broken link. This patch temporarily fixes it by changing it to PercentRank by removing the square brackets used for intra-doc linking.

---

Patch contents

diff --git a/datafusion/functions-window/src/macros.rs b/datafusion/functions-window/src/macros.rs
index e934f883b101..2905ccf4c204 100644
--- a/datafusion/functions-window/src/macros.rs
+++ b/datafusion/functions-window/src/macros.rs
@@ -303,7 +303,7 @@ macro_rules! create_udwf_expr {
     ($UDWF:ident, $OUT_FN_NAME:ident, $DOC:expr) => {
         paste::paste! {
             #[doc = " Create a [`WindowFunction`](datafusion_expr::Expr::WindowFunction) expression for"]
-            #[doc = concat!(" [`", stringify!($UDWF), "`] user-defined window function.")]
+            #[doc = concat!(" `", stringify!($UDWF), "` user-defined window function.")]
             #[doc = ""]
             #[doc = concat!(" ", $DOC)]
             pub fn $OUT_FN_NAME() -> datafusion_expr::Expr {
@@ -316,7 +316,7 @@ macro_rules! create_udwf_expr {
     ($UDWF:ident, $OUT_FN_NAME:ident, [$($PARAM:ident),+], $DOC:expr) => {
         paste::paste! {
             #[doc = " Create a [`WindowFunction`](datafusion_expr::Expr::WindowFunction) expression for"]
-            #[doc = concat!(" [`", stringify!($UDWF), "`] user-defined window function.")]
+            #[doc = concat!(" `", stringify!($UDWF), "` user-defined window function.")]
             #[doc = ""]
             #[doc = concat!(" ", $DOC)]
             pub fn $OUT_FN_NAME(

[jcsherin]

It is ok to remove Rank::get_type() as it is not used anywhere.

[jatin510]

Thanks @jcsherin

[alamb]

Thank you @jatin510 and @jcsherin -- this looks like a very nice cleanup to me

cc @berkaysynnada @ozankabak

[berkaysynnada]

LGTM. Module doc's can be updated in rank.rs.

[jatin510]

LGTM. Module doc's can be updated in rank.rs.

done @berkaysynnada

[alamb]

🎉