[
https://issues.apache.org/jira/browse/GROOVY-12673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18137448#comment-18137448
]
ASF GitHub Bot commented on GROOVY-12673:
-----------------------------------------
codeconsole opened a new pull request, #3194:
URL: https://github.com/apache/groovy/pull/3194
The closures returned by `trampoline()`, `memoize*()`, `curry()`, `>>`/`<<`
and `asWritable()` do not report or forward the delegation settings of the
closure they wrap:
```groovy
class D { def m() { 'delegate' } }
class O {
def m() { 'owner' }
def run() {
def c = { x -> m() }
assert c.curry(1).thisObject.is(this) // null
def t = c.trampoline()
t.resolveStrategy = Closure.DELEGATE_FIRST
t.delegate = new D()
assert t(1) == 'delegate' // 'owner'
}
}
new O().run()
```
Seen in 4.0.33, 5.0.8 and 6.0.0 too.
- `TrampolineClosure` passed the wrapped closure's delegate as its
`thisObject`, so its `delegate` reported the owner. It now takes the wrapped
closure's `thisObject` and forwards `getDelegate`, `setDelegate`,
`getResolveStrategy` and `setResolveStrategy` to the wrapped closure, as
`ComposedClosure` does. Its `clone()` copies the wrapped closure, so
`with`/`tap`, which clone and then set the delegate, leave the original
trampoline unchanged. Its parameter count and types now come from the wrapped
closure's getters; before, a trampolined writable or composed closure reported
a single `Object[]` parameter.
- `Memoize.MemoizeFunction` gets the same forwarding, `thisObject` and
`clone()`. The copy shares the cache.
- `CurriedClosure`, `ComposedClosure` and `Closure.WritableClosure` forward
`getThisObject` to the closure they call. `getOwner()` of these three still
returns the wrapped closure, which they call through.
One commit per wrapper.
Merging: with #3186 (GROOVY-12665, wrapper rehydrate/dehydrate) this merges
cleanly. With #3190 (GROOVY-12669, memoize rehydrate/dehydrate) one hunk in
`MemoizeFunction` conflicts: keep #3190's `cacheFactory` field and constructor
parameter, make `closure` non-final (needed by `clone()` here), and call
`super(closure.getOwner(), closure.getThisObject())`.
## Tests
These fail without the change:
- `ClosureCurryTest.testThisObject`
- `ClosureComposeTest.testThisObject`
- `ClosureTest.testWritableClosureThisObject` (includes a `@CompileStatic`
case)
- `TrampolineTest`: `testDelegateAndThisObject`,
`testSetDelegateAndResolveStrategy`, `testWith`,
`testParametersOfWrappedClosure`
- `AbstractMemoizeTestCase`: `testDelegateAndThisObject`,
`testSetDelegateAndResolveStrategy`, `testWith`, run by `MemoizeTest`,
`MemoizeAtMostTest`, `MemoizeAtLeastTest` and `MemoizeBetweenTest`
`groovy.lang.*Closure*`, `groovy.*Closure*`,
`org.codehaus.groovy.runtime.*Closure*`,
`org.codehaus.groovy.runtime.trampoline.*`,
`org.codehaus.groovy.runtime.memoize.*`, `MemoizedASTTransformationTest`,
`gls.closures.*`, `groovy.transform.stc.*` and
`org.codehaus.groovy.classgen.asm.sc.*` pass. The Checkstyle and CodeNarc gates
pass.
> Closure wrappers (trampoline, memoize, curry, compose, asWritable) report the
> wrong delegate or thisObject and ignore delegate and resolveStrategy changes
> ----------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: GROOVY-12673
> URL: https://issues.apache.org/jira/browse/GROOVY-12673
> Project: Groovy
> Issue Type: Bug
> Affects Versions: 4.0.33, 5.0.8, 7.0.0-alpha-1, 6.0.0
> Reporter: Scott Murphy Heiberg
> Priority: Major
>
> {code:groovy}
> class D { def m() { 'delegate' } }
> class O {
> def m() { 'owner' }
> def run() {
> def c = { x -> m() }
> assert c.curry(1).thisObject.is(this) // fails: null (also for
> c >> {...}, and c.asWritable().getThisObject())
> def t = c.trampoline()
> assert t.thisObject.is(this) // fails: the delegate
> of c
> t.resolveStrategy = Closure.DELEGATE_FIRST
> t.delegate = new D()
> assert t(1) == 'delegate' // fails: 'owner'
> def m = { x -> m() }.memoize()
> m.resolveStrategy = Closure.DELEGATE_FIRST
> m.delegate = new D()
> assert m(1) == 'delegate' // fails: 'owner'
> }
> }
> new O().run()
> {code}
> *Actual:* curried and composed closures report a null thisObject. A writable
> closure's getThisObject() returns null, so the property read under
> @CompileStatic gives null too. A trampolined closure reports the wrapped
> closure's delegate as its thisObject and its owner as its delegate. A
> memoized closure reports the owner as its delegate and null as its
> thisObject. Setting delegate or resolveStrategy on a trampolined or memoized
> closure has no effect on the call. {{c.asWritable().trampoline()}} also
> reports one Object[] parameter instead of the parameters of c.
> *Expected:* a wrapper reports the delegate, resolve strategy and thisObject
> of the closure it wraps, and setting delegate or resolveStrategy on it
> applies to that closure, as ComposedClosure and CurriedClosure already do for
> delegate and resolveStrategy.
> Cause: TrampolineClosure passes original.getDelegate() as thisObject, and
> neither TrampolineClosure nor Memoize.MemoizeFunction forwards the delegate
> and resolve strategy. CurriedClosure, ComposedClosure and WritableClosure do
> not forward getThisObject. TrampolineClosure reads the wrapped closure's
> parameter fields instead of its getters.
> Verified on 4.0.33, 5.0.8, 6.0.0 and master.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
