[ 
https://issues.apache.org/jira/browse/GROOVY-8992?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17782735#comment-17782735
 ] 

Eric Milles commented on GROOVY-8992:
-------------------------------------

There was discussion in another ticket to add the {{TYPE_USE}} target to 
{{@ClosureParams}} and {{@DelegatesTo}} so the return type of an annotation 
attribute method could be tagged.  I think the {{METHOD}} target would allow 
this as well -- but would be wider than the intended use case.  A type 
annotation would allow for a representation that Java can understand.

I still feel that this custom syntax is not preferred over enhancements to 
ClosureParams and DelegatesTo.  For example, there is no support for 
disambiguation in the syntax-based approach.  {{each(Map map, 
@ClosureParams(MapEntryOrKeyValue.class) Closure consumer)}} is a quick example 
of this added flexibility.

> [GEP]Polish the generics type syntax for closure
> ------------------------------------------------
>
>                 Key: GROOVY-8992
>                 URL: https://issues.apache.org/jira/browse/GROOVY-8992
>             Project: Groovy
>          Issue Type: Improvement
>            Reporter: Daniel Sun
>            Priority: Major
>
> h2. 1. Background
> Currently the syntax specifying the generics type for closure is quite 
> verbose and not type safe, e.g.
> {code:java}
> @ClosureParams(value=SimpleType.class, options="groovy.sql.Sql") Closure<V> 
> closure
> {code}
> h2. 2. Solutions
> ① ×  I propose to make the above code groovier, but the proposed "arrow 
> syntax making it hard to read, in particular when the argument types have 
> generics themselves" reminded by [~melix], e.g.
> {code:java}
> Closure<groovy.sql.Sql -> V>
> {code}
> ② √  Suggestions of [~blackdrag] are much groovier for all cases:
> {code:java}
> Closure<groovy.sql.Sql : V>
> {code}
> ③ √  In the meanwhile, [~blackdrag] proposed other variants of the generics 
> type syntax for closure to handle "polymorphic closures (aka closures which 
> accept different kind of arguments)" reminded by [~melix]
> {code:java}
> Closure<():R1; (X):R2; (Y, Z):R3>
> {code}
> ④ ? [~emge] proposed the simplified version of ③.
> {code:java}
> Closure<R1(); R2(X); R3(Y, Z)>
> {code}
> h2. 3. Benefits
> ① The new syntax of generics type for closure is much more concise and 
> readable:
> {code:java}
> Closure<groovy.sql.Sql:V>
> {code}
> {code:java}
> Closure<Sql:V> // qualified name is not necessary if using imports
> {code}
> *VS*
> {code:java}
> @ClosureParams(value=SimpleType.class, options="groovy.sql.Sql") Closure<V>
> {code}
> ② Type checking can be completed in the compilation time, so we can find 
> errors in time, e.g. {{@ClosureParams(... options="groovy.sql.SqlAbc")}} of 
> annotation specifies the type with string literal, but the type does not 
> exist, so we can not the error in the compilation time. On the contrast, 
> {{Closure<groovy.sql.SqlAbc:V>}} can make compiler help us find type errors 
> in the compilation time.
>  ③ Better IDE support because of using the types instead of string literals 
> for types
> h2. 4. Rationale
> In order to keep "consistency between using annotations and a type-checking 
> only feature" reminded by [~melix], I propose to transform the groovier code 
> to the original code when compiling, e.g.
>  {{Closure<groovy.sql.Sql:V>}}
>  will be transformed to
>  {{@ClosureParams(value=SimpleType.class, options="groovy.sql.Sql") 
> Closure<V>}}
> h2. 5. Discussions in the dev mailing list
> [http://groovy.329449.n5.nabble.com/About-polish-the-generics-type-syntax-for-closure-tt5756586.html]



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to