| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
I screwed around with type printing for a long time and now I have
to be done, so I suggest we accept the imperfection in here for now
because it's still way ahead of "..." as types go.
|
| |\
| |
| | |
IDE improvements, with particular focus on making worksheets work.
|
| | |
| |
| |
| | |
Adds a raw string interpolator raw"..." which does not do any escape sequence processing.
|
| | |
| |
| |
| |
| |
| | |
Added all the known cases of failing self-types. Added tests for symbol
access (SI-4601) as well which should now be fixed for all of the same cases
as well.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Following this commit are reversions of three prior commits which
introduced difficulties of their own, plus extra tests for this
issue and SI-4601, all of which I pinched from jrudolph.
Maybe there was a good reason for some of the complicated code
related to this ticket. I took the naive position that if we avoided
generating a method call to a particular receiver if the receiver
has no such method, we would encounter fewer NoSuchMethodErrors.
I would believe one can construct a scenario which this doesn't
handle correctly, but it's hard to believe this isn't a big
improvement. Review by @jrudolph, @odersky.
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An extractor pattern `X(p)` should type check for any `X.unapply`/`X.unapplySeq`
that returns an `Option[_]` -- previously we were confused about the case where
it was an `Option[(T1, ... , Tn)]`. In this case, the expected type for the
pattern `p` is simply `(T1, ... , Tn)`.
While I was at it, tried to clean up unapplyTypeList and friends (by replacing them by extractorFormalTypes).
From the spec:
8.1.8 ExtractorPatterns
An extractor pattern x(p1, ..., pn) where n ≥ 0 is of the same syntactic form as a constructor pattern.
However, instead of a case class, the stable identifier x denotes an object which has a member method named unapply or unapplySeq that matches the pattern.
An unapply method in an object x matches the pattern x(p1, ..., pn) if it takes exactly one argument and one of the following applies:
n = 0 and unapply’s result type is Boolean.
n = 1 and unapply’s result type is Option[T], for some type T.
the (only) argument pattern p1 is typed in turn with expected type T
n > 1 and unapply’s result type is Option[(T1, ..., Tn)], for some types T1, ..., Tn.
the argument patterns p1, ..., pn are typed in turn with expected types T1, ..., Tn
|
| |
|
|
| |
Sharpens the test so that only Object_== and Object_!= methods on valueclasses may be rewritten in posterasure, whereas user-defined methods are not rewritten.
|
| |\
| |
| | |
Removes redundant outers
|
| | |
| |
| |
| |
| |
| | |
Widens the criterion when outer fields can be omitted. It used to be that sub- and superclass had to be enclosed by the same outer class. Only in that case was the outer field of the class omitted. We now omit if subclass is contained in an outer class that is itself a subclass of the superclasses outer class.
See test case "outertest.scala" for an example.
|
| |\ \
| | |
| | | |
Fix for SI-5385.
|
| | | |
| | |
| | |
| | |
| | | |
Nodes which hit EOF with no whitespace afterward had
wrong position.
|
| |\ \ \
| | | |
| | | | |
Implement @static annotation on singleton object fields.
|
| | | | | |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This commit introduces the `@static` annotation on fields of static singleton objects.
A static singleton object is a top-level singleton object or an object nested within
a static singleton object.
The field annotated with `@static` is generated as a true static field in the companion
class of the object. The initialization of the field is done in the static initializer of the
companion class, instead of the object itself.
Here's an example. This:
object Foo {
@static val FIELD = 123
}
class Foo
generates :
object Foo {
def FIELD(): Int = Foo.FIELD
}
class Foo {
<static> val FIELD = 123
}
The accessor in `object Foo` is changed to return the static
field in `class Foo` (the companion class is generated if it is
not already present, and the same is done for getters if `FIELD`
is a `var`).
Furthermore, all the callsites to accessor `Foo.FIELD` are rewritten
to access the static field directly.
It is illegal to annotate a field in the singleton object as `@static` if there is
already a member of the same name in `class Foo`.
This allows better Java interoperability with frameworks which rely on static fields being
present in the class.
The `AtomicReferenceFieldUpdater`s are one such example. Instead of having
a Java base class holding the `volatile` mutable field `f` and a static field containing
a reference to the `AtomicReferenceFieldUpdater` that mutates `f` atomically,
and extending this Java base class from Scala, we can now simply do:
object Foo {
@static val arfu = AtomicReferenceUpdater.newUpdater(
classOf[Foo], classOf[String], "f"
)
}
class Foo {
@volatile var f = null
import Foo._
def CAS(ov: String, nv: String) = arfu.compareAndSet(this, null, "")
}
In summary, this commit introduces the following:
- adds the `@static` annotation
- for objects without a companion class and `@static` members, creates a companion class
(this is done in `CleanUp`)
- checks for conflicting names and if `@static` is used on static singleton object member
- creates the `static` field in the companion class for `@static` members
- moves the field initialization from the companion object to the static initializer of
the class (the initialization of `@static` members is bypassed in the `Constructors` phase,
and added to static ctors in `CleanUp`)
- all callsites to the accessors of `@static` are rewritten to access the static fields directly
(this is done in `GenICode`)
- getters and setters to `@static` fields access the static field directly, and the field in the
singleton object is not emitted (this is done in `GenICode`)
- changes in `GenJVM`, `GenASM` - now computing local var indices in static initializers
as well - this was an oversight before, now is necessary
Future work: allow the `@static` annotation on methods as well - this will
be added in a future commit.
Review by @odersky, @dragos, @paulp, @heathermiller.
|
| |\ \ \ \
| | | | |
| | | | | |
SI-6104 support This pattern
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This(name) is treated just like Ident(name)
apparently this pattern was used in 2.9 code,
though I'm not sure it's spec'ed
|
| |\ \ \ \ \
| | | | | |
| | | | | | |
SI-5999 consistent behavior for c.reify and c.universe.reify
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
After a discussion on a reflection meeting on Jul 17
we concluded that we should split staticModule into
staticModule and staticPackage to remove the ambiguity
between packageless objects and packageless packages
(more in the comments in the body of the commit).
The motivation is verbosely outlined in the comments,
but the bottom line is that Scala allows packages and
packageless objects to have the same name within the same program.
Therefore at times we need to disambiguate, hence the introduction
of the staticPackage method.
As of such staticModule no longer works for packages.
In the same fashion staticPackage doesn't work for modules.
This is done to ensure robustness of reification.
I would like to do the same for getModule in Definitions,
but we have to maintain backward compatibility. That's why I retained
the old behavior, but replaced getModule invocations with getPackage
where appropriate to be in line with staticModule and staticPackage.
Another important thing that follows from the discussion is that
both staticClass and staticModule prefer parent packages over parent objects
in cases of ambiguity. Say, if we have the following snippet of code:
object B { class C } next to package B { class C }
then staticClass("B.C") will never even consider a C inside the object B.
This is how scalac operates, so we decided to be consistent here.
Finally reification logic got changed to distinguish between
staticModule and staticPackage, and to allow for the fact that
staticClass and staticModule prefer parent packages to parent objects.
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Currently there are discrepancies between the behavior of
c.reify and c.universe.reify.
First step in fixing these problems is removing the duplication in the API.
That's why I'm cutting away the Context.reify shortcut.
Context.reify is a magic macro, hardwired in the fast track mechanism, so
removing it requires redeploying the starr (because an old starr will
crash if launched on sources that don't contain Context.reify).
To cleanly redeploy a starr I've left a Context.reify stub in sources,
but hidden it behind a `protected` modifier. When starr is redeployed
(in a subsequent commit) the stub will be removed.
I've also updated the tests to use c.universe.reify instead of c.reify.
This will break some of them, because c.universe.reify uses a standard
compiler mirror, which unlike a macro mirror doesn't like packageless classes.
That's an annoyance, but I think having clean separation of commits
is more important that being 100% consistent.
|
| |\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-5856 enables use of $this in string interpolation
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This pull request fixes SI-5856. The scanner has been modified
to return the correct token if keywords appear in $-expressions.
The parser has been modified to issue an error and to only accept
$<identifier>, $this and $block.
|
| |\ \ \ \ \ \ \
| |_|/ / / / /
|/| | | | | | |
SI-5739 store sub-patterns in local vals
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Also closes SI-5158 (debuggability), SI-6070 (soundness).
To improve both debuggability and soundness,
we now store the result of an extractor (user-defined and synthetic)
in local variables.
For the case class case, this also fixes the soundness bug SI-6070,
as this prevents post-match mutation of bound variables.
The core of the refactoring consisted of introducing the PreserveSubPatBinders
trait, which introduces local variables instead of substituting symbols
for the RHS of those variables (so this can be seen as reverting the premature optimization
of inline the case-class getters into the case body).
Since TreeMakerToCond fuses the substitutions performed in a match to find out
which symbolic values binders evaluate to, masquerade PreserveSubPatBinders's
binding of subPatBinders and subPatRefs as the corresponding substitution.
Consider `case class Foo(bar: Int)`, then `case y@Foo(x) => println(x)`
gives rise to `{val x = y.bar; println(x)}` (instead of `println(y.bar)`),
and `subPatternsAsSubstitution` pretends we still replace `x` by `y.bar`,
instead of storing it in a local variable so that the rest of the analysis need not be modified.
Misc notes:
- correct type for seq-subpattern
- more error resilience (ill-typed patterns sometimes slip past the typechecker -- reopened SI-4425)
TODO: come up with a more abstract framework for expressing bound symbols and their values
|
| |\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
fixes field mirrors and also improves docs and exceptions for all mirrors
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Factors out error raising code and introduces a special exception class
for Scala reflection errors.
Also adds membership sanity checks to reflectXXX. Previously
reflectField, reflectMethod, reflectClass and reflectModule
in InstanceMirror didn't check that the symbols being passed to them
actually correspond to some member of the related class.
|
| | | |_|_|_|/ /
| |/| | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
reflectField now accepts getters and setters along with the field symbols,
it also checks whether a field has a reasonable binary representation
(this is necessary, because ctor parameters that are unused outside of their
declaring constructors don't get compiled down to Java fields/methods).
|
| |\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
SI-4897 derive expected value from single type
|
| | | |_|_|_|_|/
| |/| | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
when the type in a type test is, say, `C.this.A.type`,
must use the corresponding term `C.this.A` to test for equality
if you use the naive REF(<C.this.A.type>.symbol), you'll
get a path like `OwnerOfA.this.A`,
where `OwnerOfA` might be a superclass of `C`, and
explicitouter won't like that
|
| |\ \ \ \ \ \ \
| |/ / / / / /
|/| | | | | | |
SI-6086 magic symbols strike back
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Some of the symbols inside the compiler get created on the fly,
because there are no physical entities in classfiles corresponding to them.
This curious fact needs to be taken into account when loading symbols,
so that the magic symbols get correctly loaded by reflective mirrors.
magicSymbols (as defined in Definitions.scala) include not only
top-level classes, but some other stuff (e.g. String_+ or methods on Any).
Hence a filtering was done to exclude the stuff that's irrelevant
to reflective symbol loading.
Unfortunately a filter was configured to accept only _.isClass,
which consequently ruled out scala.AnyRef (that is a type alias).
|
| |\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
SI-6092 don't leak in LazyAnnotationInfo, don't lift try{}finally{}
|
| | | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
which a try-catch is lifted
out to an inner method.
Less known fact: lazy values null-out their dependent values is their accessed only from
their initializer. The analysis is not context-dependent (meaning the owner where a reference
happens needs to be exactly that lazy value).
* Removed no-op code around positions in `LazyAnnotationInfo`
* Don't lift expressions that have no `catch` clause
The two changes combined fix a memory leak that's been plaguing the IDE: an annotation
(even when forced) would hang on to a namer, through the outer field of its call-by-name
parameter.
The test for the memory leak is in the IDE project (couldn't find a simple way to reproduce it outside
the IDE), but there's a test checking that the field is null after initialization.
|
| |\ \ \ \ \ \ \ \
| |_|_|_|_|_|/ /
|/| | | | | | | |
SI-6089 better tail position analysis for matches
|
| | | |_|/ / / /
| |/| | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
we mistakenly went into apply nodes to look for matchEnd-labeldefs
in tail positions -- only apply nodes that jump to labels in tailpos
should be traversed (this arises for nested matches)
|
| |\ \ \ \ \ \ \
| |_|_|_|_|_|/
|/| | | | | | |
SI-5937: Vector updated, +:, and :+ ignore provided builder, breaking type safety
|
| | | |_|_|/ /
| |/| | | | |
|
| |\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6061 adds weakly conformance for number types to resolveOverloaded
|
| | | |_|_|_|/
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
This fix changes resolveOverloaded in scala.reflect.internal
so that numeric widening on the actual argument types is also
considered for primitive number types. Needs changes in functions
applicable and mostSpecific.
|
| |\ \ \ \ \ \
| | | | | | |
| | | | | | | |
move test files that fail spuriously to pending
|
| | | |_|_|_|/
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
I have this sneaky suspicion that part of these spurious failures
are caused by the recent partest optimizations.
@axel22 already checked that compiler instances are not shared between test runs.
However, except for the benchmark test, they all have a distinct
race condition in symbol loading/type checking feel to them.
Since, in the end, the tests and/or their corresponding fixes are as likely
a culprit as the test framework, moving them out of the way until their owners
can get them back in line and they stop throwing primate wenches into our build.
We should bring them back as soon as possible, though.
|
| |\ \ \ \ \ \
| |_|_|_|_|/
|/| | | | | |
SI-5731 a few fixes for value classes
|
| | | |_|_|/
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
I've faced two gotchas.
First of all posterasure, which is supposed to erase ErasedValueType types,
didn't look into ConstantType.value that is known to be smuggling types
(hi Paul that's a plus one).
Secondly ClassManifest.classType[T] assumed that its T is bound by AnyRef,
which is not the case for value types. Here I had two choices:
a) introduce a special method for manifests of value types,
b) remove the upper bound of the type parameter and call it a day.
Since manifests are already deprecated and there's no difference which
method was used to create which manifest, I went for option b).
|
| |\ \ \ \ \
| |_|_|_|/
|/| | | | |
Fixes SI-5588. Correct compare for Enumeration.
|
| | |/ / /
| | | |
| | | |
| | | |
| | | |
| | | | |
Slower than the original one but does comparison in the same spirit as IntOrdering.
Review by @axel22.
|
| |\ \ \ \
| |/ / /
|/| | | |
SI-6077 more conservative TreeMakersToConds for CSE
|
| | | |/
| |/|
| | |
| | |
| | |
| | |
| | | |
we were dropping irrefutable extractor calls (implemented for SI-4691)
in common subcondition elimination, which caused bad code to be generated
clean up TreeMakersToConds significantly in the process
|
| |\ \ \
| |_|/
|/| | |
SI-6028 Avoid needless symbol renaming in lambdalift
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Preserve names of all referenced free vars. Only the proxy symbols
have the fresh names.
The resulting natural beauty is evident in the diff of t6028.check.
This subsumes the treatment in 0e170e4b that ensured named parameter
calls cannot see mangled names; pos/t6028 confirms as much.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
No doubt there are plenty of additional variations that could be
added to exercise more code paths.
This is the unflattering "before" photo; the next commit will
make over the name mangling and reveal the simple beauty of
unmangled names.
References SI-6028
|
| |/ /
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A complete overhaul. The original implementation in SI-5830 (#821) was pretty buggy.
[from the doc comment of `collapseGuardedCases`:]
Collapse guarded cases that switch on the same constant (the last case may be unguarded).
Cases with patterns A and B switch on the same constant iff for all values x that match A also match B and vice versa.
(This roughly corresponds to equality on trees modulo alpha renaming and reordering of alternatives.)
The rewrite only applies if some of the cases are guarded (this must be checked before invoking this method).
The rewrite goes through the switch top-down and merges each case with the subsequent cases it is implied by
(i.e. it matches if they match, not taking guards into account)
If there are no unreachable cases, all cases can be uniquely assigned to a partition of such 'overlapping' cases,
save for the default case (thus we jump to it rather than copying it several times).
(The cases in a partition are implied by the principal element of the partition.)
The overlapping cases are merged into one case with their guards pushed into the body as follows
(with P the principal element of the overlapping patterns Pi):
`{case Pi if(G_i) => B_i }*` is rewritten to `case P => {if(G_i) B_i}*`
The rewrite fails (and returns Nil) when:
(1) there is a subsequence of overlapping cases that has an unguarded case in the middle;
only the last case of each subsequence of overlapping cases may be unguarded (this is implied by unreachability)
(2) there are overlapping cases that differ (tested by `caseImpliedBy`)
cases with patterns A and B are overlapping if for SOME value x, A matches x implies B matches y OR vice versa <-- note the difference with case equality defined above
for example `case 'a' | 'b' =>` and `case 'b' =>` are different and overlapping (overlapping and equality disregard guards)
Improved by @retronym's feedback in the following ways:
- fix patternEquals (it's now quadratic, but correct)
- update neg/t6011 to test the improved patternEquals
- remove side-effect-in-condition ugliness
- introduce isGuardedCase
- docs & various code clarity
Also closes SI-6048 (duplicate).
|
