[
https://issues.apache.org/jira/browse/GROOVY-11402?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eric Milles updated GROOVY-11402:
---------------------------------
Description:
Consider the following:
{code:groovy}
class C extends HashMap<String,String> {
private Object f = 'field'
private static Object g = 'field'
@groovy.transform.TypeChecked
void test() {
{->
def a = f
println a // entry
def b = g
println b // field
def c = this.f
println c // entry or field (if SC)
def d = this.g
println d // field
}()
}
}
new C().test()
{code}
Fixes in Groovy 3 [1] and 4 [2] have made it so that "g" and "this.g" resolve
to the static field (even from outer class). Groovy 5 has dropped this special
case and should go through the closure's {{getProperty}} protocol. This is at
least consistent with "g" and "this.g" appearing outside of a closure.
However, "f" and "this.f" have some inconsistencies. "f" resolves to entry for
STC and SC. "this.f" resolves to field at compile time for STC and SC. But to
entry for STC and field for SC at run time.
Should "this.f" be entry within closure as "f" is for all cases? Or should
only the STC inferred type be fixed to match the run time behavior, leaving SC
unchanged?
GROOVY-5001, GROOVY-5491, GROOVY-8555, GROOVY-11367, GROOVY-11387, GROOVY-11401
[1]
https://github.com/apache/groovy/commit/419ce2f91754f101d8e65977da4dc7e8e6b6265f
[2]
https://github.com/apache/groovy/commit/3c2266e053c4a9a9127be43921afb0c5135725f6
was:
Consider the following:
{code:groovy}
class C extends HashMap<String,String> {
private Object f = 'field'
private static Object g = 'field'
@groovy.transform.TypeChecked
void test() {
{->
def a = f
println a // entry
def b = g
println b // field
def c = this.f
println c // entry or field (if SC)
def d = this.g
println d // field
}()
}
}
new C().test()
{code}
Fixes in Groovy 3 [1] and 4 [2] have made it so that "g" and "this.g" resolve
to the static field (even from outer class). Groovy 5 has dropped this special
case and should go through the closure's {{getProperty}} protocol. This is at
least consistent with "g" and "this.g" appearing outside of a closure.
However, "f" and "this.f" have some inconsistencies. "f" resolves to entry for
STC and SC. "this.f" resolves to field at compile time for STC and SC. But to
entry for STC and field for SC at run time.
Should "this.f" be entry within closure as "f" is for all cases? Or should
only the STC inferred type be fixed to match the run time behavior, leaving SC
unchanged?
GROOVY-5001, GROOVY-5491, GROOVY-8555, GROOVY-11367, GROOVY-11401
[1]
https://github.com/apache/groovy/commit/419ce2f91754f101d8e65977da4dc7e8e6b6265f
[2]
https://github.com/apache/groovy/commit/3c2266e053c4a9a9127be43921afb0c5135725f6
> STC: inferred type of this.name within closure
> ----------------------------------------------
>
> Key: GROOVY-11402
> URL: https://issues.apache.org/jira/browse/GROOVY-11402
> Project: Groovy
> Issue Type: Bug
> Components: Static Type Checker
> Affects Versions: 3.0.21, 4.0.21
> Reporter: Eric Milles
> Assignee: Eric Milles
> Priority: Major
>
> Consider the following:
> {code:groovy}
> class C extends HashMap<String,String> {
> private Object f = 'field'
> private static Object g = 'field'
> @groovy.transform.TypeChecked
> void test() {
> {->
> def a = f
> println a // entry
> def b = g
> println b // field
> def c = this.f
> println c // entry or field (if SC)
> def d = this.g
> println d // field
> }()
> }
> }
> new C().test()
> {code}
> Fixes in Groovy 3 [1] and 4 [2] have made it so that "g" and "this.g" resolve
> to the static field (even from outer class). Groovy 5 has dropped this
> special case and should go through the closure's {{getProperty}} protocol.
> This is at least consistent with "g" and "this.g" appearing outside of a
> closure.
> However, "f" and "this.f" have some inconsistencies. "f" resolves to entry
> for STC and SC. "this.f" resolves to field at compile time for STC and SC.
> But to entry for STC and field for SC at run time.
> Should "this.f" be entry within closure as "f" is for all cases? Or should
> only the STC inferred type be fixed to match the run time behavior, leaving
> SC unchanged?
> GROOVY-5001, GROOVY-5491, GROOVY-8555, GROOVY-11367, GROOVY-11387,
> GROOVY-11401
> [1]
> https://github.com/apache/groovy/commit/419ce2f91754f101d8e65977da4dc7e8e6b6265f
> [2]
> https://github.com/apache/groovy/commit/3c2266e053c4a9a9127be43921afb0c5135725f6
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
