| Commit message (Collapse) | Author | Age | Files | Lines |
|\
| |
| | |
SI-6541 valid wildcard existentials for case-module-unapply
|
| |
| |
| |
| |
| |
| |
| |
| | |
Instead of letting the compiler infer the return type of case module
unapply methods, provide them explicitly.
This is enabled only under -Xsource:2.12, because the change is not
source compatible.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The default value was NoScalaVersion before, because tryToSet (where
the default was supposed to be set) is not called at all if the option
is not specified.
The initial value of Xmigration is set to NoScalaVersion (which it was
before, the AnyScalaVersion argument was ignored). AnyScalaVersion
would be wrong, it would give a warning in `Map(1 -> "eis").values`
if the option was not specified. See tests.
|
|\ \
| | |
| | | |
SI-8916 Clean up unused imports, values and variables
|
| | | |
|
| | | |
|
|\ \ \
| | | |
| | | | |
SI-8711 ScalaVersion.unparse doesn't produce valid versions
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
There is no dot between `major.minor.rev` and `-build` in a scala
version, yet that's what unparse returns for
```
// was "2.11.3.-SNAPSHOT"
ScalaVersion("2.11.3-SNAPSHOT").unparse
```
|
|\ \ \ \
| | | | |
| | | | | |
SI-8893 Restore linear perf in TailCalls with nested matches
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The iteration is only needed to side effect on `instructionList`
and `code.touched`. This commit uses an iterator and `foreach`,
rather than creating throwaway Arrays.
% time (for f in {1..200}; do echo test/files/pos/t8893.scala; done | qscalac -Xresident)
Before: 30s
After : 21s
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Compilation perfomance of the enclosed test regressed when the
new pattern matcher was introduced, specifically when the tail
call elimination phase was made aware of its tree shapes in cd3d34203.
The code added in that commit detects an application to a tail label
in order to treat recursive calls in the argument as in tail position.
If the transform of that argument makes no change, it falls
back to `rewriteApply`, which transforms the argument again
(although this time on a non-tail-position context.)
This commit avoids the second transform by introducing a flag
to `rewriteApply` to mark the arguments are pre-transformed.
I don't yet see how that fixes the exponential performance, as
on the surface it seems like a constant factor improvement.
But the numbers don't lie, and we can now compile the test
case in seconds, rather than before when it was running for
> 10 minutes.
This test case was based on a code pattern generated by the Avro
serializer macro in:
https://github.com/paytronix/utils-open/tree/release/2014/ernststavrosgrouper
The exponential performance in that context is visualed here:
https://twitter.com/dridus/status/519544110173007872
Thanks for @rmacleod2 for minimizing the problem.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Save the Trees!
When a subtransform is an identity, we must
strive to return the identical tree to enable the lazy part of
LazyTreeCopier. If we get this wrong in a leaf, all parents are
wastefully copied.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
I believe ClonedTailContext was added to avoid the need to
mutate the Boolean `ctx.tailPos`. All other calls are forwarded
to a delegate context.
This commit tries to find a delegate context with the right
value of `tailPos` to reuse that, rather than creating a
wrapper each time we need to flip that bit.
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Cache the member symbols for `Boolean.{||, &&}` per-run, rather
than look them up repeatedly.
Based on profiling the tail calls phase in the program below, which
was distilled by @rmacleod2 from the code generated by macros in
https://github.com/paytronix/utils-open/tree/release/2014/ernststavrosgrouper
Wall clock time went from 12s to 6.5s.
```scala
object Test {
def a(): Option[String] = Some("a")
def main(args: Array[String]) {
a() match {
case Some(b1) =>
a() match {
case Some(b2) =>
a() match {
case Some(b3) =>
a() match {
case Some(b4) =>
a() match {
case Some(b5) =>
a() match {
case Some(b6) =>
a() match {
case Some(b7) =>
a() match {
case Some(b8) =>
a() match {
case Some(b9) =>
a() match {
case Some(b10) =>
a() match {
case Some(b11) =>
a() match {
case Some(b12) =>
a() match {
case Some(b13) =>
a() match {
case Some(b14) =>
a() match {
case Some(b15) =>
a() match {
case Some(b16) =>
a() match {
case Some(b17) =>
a() match {
case Some(b18) =>
a() match {
case Some(b19) =>
a() match {
case Some(b20) =>
println("yay")
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
case None => None
}
}
}
```
|
|\ \ \ \
| | | | |
| | | | | |
SI-8910 BitSet sometimes uses exponential memory.
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Because of an off-by-one error in scala.collection.mutable.BitSet, where a
function (ensureCapacity) is passed a list length instead of an index, when
^=, &=, |=, or &~= are passed BitSets with the same internal capacity as the
set the method is being invoked on, the size of the first BitSet is needlessly
doubled.
This patch simply changes the ensureCapacity calls to pass the last index of
the list, instead of the raw length. In addition, add documentation to
ensureCapacity to try and stop something similar from happening in the future.
|
|\ \ \ \
| | | | |
| | | | | |
IntelliJ IDEA files for version 14
|
| | |_|/
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The latest version of the Scala plugin for IntelliJ IDEA introduces
a new project structure
http://blog.jetbrains.com/scala/2014/10/30/scala-plugin-update-for-intellij-idea-14-rc-is-out/
Due to a bug (https://youtrack.jetbrains.com/issue/SCL-7753), you
currently need to install the latest nightly build from here:
http://confluence.jetbrains.com/display/SCA/Scala+Plugin+Nightly+Builds+for+Cassiopeia
The new format doesn't allow scala compiler options per-module, so
the `-sourcepath src/libarary` is used for all modules.
|
|\ \ \ \
| | | | |
| | | | | |
SI-8875 showCode should print all class constructor modifiers.
|
| | |/ /
| |/| |
| | | |
| | | |
| | | | |
showCode used to print nothing when the only modifier was a change in
visibility scope (i.e. no flags but privateWithin is set).
|
|\ \ \ \
| |_|/ /
|/| | | |
fix typo. s/represenation/representation
|
| |/ / |
|
|\ \ \
| | | |
| | | | |
SI-8943 Handle non-public case fields in pres. compiler
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When a case class is type checked, synthetic methods are added,
such as the `hashCode`/`equals`, implementations of the `Product`
interface. At the same time, a case accessor method is added for
each non-public constructor parameter. This the accessor for a
parameter named `x` is named `x$n`, where `n` is a fresh suffix.
This is all done to retain universal pattern-matchability of
case classes, irrespective of access. What is the point of allowing
non-public parameters if pattern matching can subvert access? I
believe it is to enables private setters:
```
case class C(private var x: String)
scala> val x = new C("")
x: C = C()
scala> val c = new C("")
c: C = C()
scala> val C(x) = c
x: String = ""
scala> c.x
<console>:11: error: variable x in class C cannot be accessed in C
c.x
^
scala> c.x = ""
<console>:13: error: variable x in class C cannot be accessed in C
val $ires2 = c.x
^
<console>:10: error: variable x in class C cannot be accessed in C
c.x = ""
^
```
Perhaps I'm missing additional motivations.
If you think scheme sounds like a binary compatiblity nightmare,
you're right: https://issues.scala-lang.org/browse/SI-8944
`caseFieldAccessors` uses the naming convention to find the right
accessor; this in turn is used in pattern match translation.
The accessors are also needed in the synthetic `unapply` method
in the companion object. Here, we must tread lightly to avoid
triggering a typechecking cycles before; the synthesis of that method
is not allowed to force the info of the case class.
Instead, it uses a back channel, `renamedCaseAccessors` to see
which parameters have corresonding accessors.
This is pretty flaky: if the companion object is typechecked
before the case class, it uses the private param accessor directly,
which it happends to have access to, and which duly gets an
expanded name to allow JVM level access. If the companion
appears afterwards, it uses the case accessor method.
In the presentation compiler, it is possible to typecheck a source
file more than once, in which case we can redefine a case class. This
uses the same `Symbol` with a new type completer. Synthetics must
be re-added to its type.
The reported bug occurs when, during the second typecheck, an entry
in `renamedCaseAccessors` directs the unapply method to use `x$1`
before it has been added to the type of the case class symbol.
This commit clears corresponding entries from that map when we
detect that we are redefining a class symbol.
Case accessors are in need of a larger scale refactoring. But I'm
leaving that for SI-8944.
|
|\ \ \
| | | |
| | | | |
SI-8941 Idempotent presentation compilation of implicit classes
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
A retrospective test case which covers typechecking idempotency which
was introduced in 0b78a0196 / 148736c3df. It also tests the
implicit class handling, which was fixed in the previous commit.
It is difficult to test this using existing presentation compiler
testing infrastructure, as one can't control at which point during
the first typechecking the subesquent work item will be noticed.
Instead, I've created a test with a custom subclass of
`interactive.Global` that allows precise, deterministic control
of when this happens. It overrides `signalDone`, which is called
after each tree is typechecked, and watches for a defintion with
a well known name. At that point, it triggers a targetted typecheck
of the tree marked with a special comment.
It is likely that this approach can be generalized to a reusable
base class down the track. In particular, I expect that some of
the nasty interactive ScalaDoc bugs could use this single-threaded
approach to testing the presentation compiler.
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When we name an implicit class, `enterImplicitWrapper` is called,
which enters the symbol for the factory method into the
owning scope. The tree defining this factory method is stowed into
`unit.synthetics`, from whence it will be retrieved and incorporated
into the enclosing tree during typechecking (`addDerivedTrees`).
The entry in `unit.synthetics` is removed at that point.
However, in the presentation compiler, we can typecheck a unit
more than once in a single run. For example, if, as happens
in the enclosed test, a call to ask for a type at a given
position interrupts type checking of the entire unit, we
can get into a situation whereby the first type checking
invocation has consumed the entry from `unit.synthetics`,
and the second will crash when it can't find an entry.
Similar problems have been solved in the past in
`enterExistingSym` in the presentation compiler. This method
is called when the namer encounters a tree that already has
a symbol attached. See 0b78a0196 / 148736c3df.
This commit takes a two pronged approach.
First, `enterExistingSym` is extended to handle implicit classes.
Any previous factory method in the owning scope is removed, and
`enterImplicitWrapper` is called to place a new tree for the factory
into `unit.synthetics` and to enter its symbol into the owning scope.
Second, the assertions that could be tripped in `addDerivedTrees`
and in `ImplicitClassWrapper#derivedSym` have been converted to
positioned errors.
The first change is sufficient to fix this bug, but the second
is also enough to make the enclosed test pass, and has been retained
as an extra layer of defence.
|
|\ \ \
| | | |
| | | | |
SI-3439 Fix use of implicit constructor params in super call
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When typechecking the primary constructor body, the symbols of
constructor parameters of a class are owned by the class's owner.
This is done make scoping work; you shouldn't be able to refer to
class members in that position.
However, other parts of the compiler weren't so happy about
this arrangement. The enclosed test case shows that our
checks for invalid, top-level implicits was spuriously triggered,
and implicit search itself would fail.
Furthermore, we had to hack `Run#compiles` to special case
top-level early-initialized symbols. See SI-7264 / 86e6e9290.
This commit:
- introduces an intermediate local dummy term symbol which
will act as the owner for constructor parameters and early
initialized members
- adds this to the `Run#symSource` map if it is top level
- simplifies `Run#compiles` accordingly
- tests this all in a top-level class, and one nested in
another class.
|
|\ \ \
| | | |
| | | | |
SI-8934 Fix whitebox extractor macros in the pres. compiler
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The code that "aligns" patterns and extractors assumes that it can
look at the type of the unapply method to figure the arity of the
extractor.
However, the result type of a whitebox macro does not tell the
whole story, only after expanding an application of that macro do
we know the result type.
During regular compilation, this isn't a problem, as the macro
application is expanded to a call to a synthetic unapply:
{
class anon$1 {
def unapply(tree: Any): Option[(Tree, List[Treed])]
}
new anon$1
}.unapply(<unapply selector>)
In the presentation compiler, however, we now use
`-Ymacro-expand:discard`, which expands macros only to compute
the type of the application (and to allow the macro to issue
warnings/errors). The original application is retained in the
typechecked tree, modified only by attributing the potentially-sharper
type taken from the expanded macro.
This was done to improve hyperlinking support in the IDE.
This commit passes `sel.tpe` (which is the type computed by the
macro expansion) to `unapplyMethodTypes`, rather than using
the `finalResultType` of the unapply method.
This is tested with a presentation compiler test (which closely
mimics the reported bug), and with a pos test that also exercises
`-Ymacro-expand:discard`. Prior to this patch, they used to fail with:
too many patterns for trait api: expected 1, found 2
|
|\ \ \
| |/ /
|/| | |
FSC / REPL Bug Bonanza
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
We needed to hop from the enum's owner to its companion
module in an early phase of the compiler.
The enclosed test used to fail when this lookup returned
NoSymbol on the second run of the resident compiler when
triggered from `MixinTransformer`: the lookup didn't work
after the flatten info transform. This is related to the
fact that module classes are flattened into the enclosing
package, but module accessors remain in the enclosing class.
|
| | |
| | |
| | |
| | |
| | |
| | | |
08505bd added a workaround for an undiagnosed interaction
between resident compilation and specialzation. This is no
longer needed after the previous commit.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The transformation of applications to specialized methods
relies on the owner of said method having had the specialization
info transform run which stashes a bunch of related data into
per-run caches such as `SpecializeTypes#{typeEnv}`.
Recently, we found that per-run caches didn't quite live up
to there name, and in fact weren't being cleaned up before
a new run. This was remedied in 00e11ff.
However, no good deed goes unpunished, and this led to a
regression in specialization in the REPL and FSC.
This commit makes two changes:
- change the specialization info tranformer to no longer
directly enter specialized methods into the `info` of whatever
the current phase happens to be. This stops them showing up
`enteringTyper` of the following run.
- change `adaptInfos` to simply discard all but the oldest
entry in the type history when bringing a symbol from one
run into the next. This generalizes the approach taken to
fix SI-7801. The specialization info transformer will now
execute in each run, and repopulate `typeEnv` and friends.
I see that we have a seemingly related bandaid for this sort
of problem since 08505bd4ec. In a followup, I'll try to revert
that.
|
|\ \ \
| | | |
| | | | |
SI-8900 Don't assert !isDelambdafyFunction, it may not be accurate
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The implementations of isAnonymousClass, isAnonymousFunction,
isDelambdafyFunction and isDefaultGetter check if a specific substring
(eg "$lambda") exists in the symbol's name.
SI-8900 shows an example where a class ends up with "$lambda" in its
name even though it's not a delambdafy lambda class. In this case the
conflict seems to be introduced by a macro. It is possible that the
compiler itself never introduces such names, but in any case, the
above methods should be implemented more robustly.
This commit is band-aid, it fixes one specific known issue, but there
are many calls to the mentioned methods across the compiler which
are potentially wrong.
Thanks to Jason for the test case!
|
|\ \ \
| | | |
| | | | |
SI-8926 default visbility RUNTIME for java annotations
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When parsed from source, java annotation class symbol are lacking the
`@Retention` annotation. In mixed compilation, java annotations are
therefore emitted with visibility CLASS.
This patch conservatively uses the RUNTIME visibility in case there is
no @Retention annotation.
The real solution is to fix the Java parser, logged in SI-8928.
|
|\ \ \
| | | |
| | | | |
[nomerge] SI-8899 Revert "SI-8627 make Stream.filterNot non-eager"
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This reverts commit 9276a1205f74fdec74206209712831913e93f359.
The change is not binary compatible, See discussion on SI-8899. Making
filterImpl non-private changes its call-sites (within TraversableLike)
from INVOKESTATIC to INVOKEINTERFACE. Subclasses of TraversableLike
compiled before this change don't have a mixin for filterImpl.
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Test case by Jason.
RefChecks adds the lateMETHOD flag lazily in its info transformer.
This means that forcing the `sym.info` may change the value of
`sym.isMethod`.
0ccdb151f introduced a check to force the info in isModuleNotMethod,
but it turns out this leads to errors on stub symbols (SI-8907).
The responsibility to force info is transferred to callers, which
is the case for other operations on symbols, too.
|
|\ \
| | |
| | | |
SI-8894 dealias when looking at tuple components
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Classic bait-and-switch: `isTupleType` dealiases, but `typeArgs` does not.
When deciding with `isTupleType`, process using `tupleComponents`.
Similar for other combos. We should really enforce this using extractors,
and only decouple when performance is actually impacted.
|
|\ \ \
| | | |
| | | | |
Color REPL under -Dscala.color
|
| | | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
* Errors are red
* Warnings are yellow
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
We already use -Dscala.color when using -Ytyper-debug
This tries to reuse the colors chosen from the debug flag:
* Bold blue for vals (e.g. "res0")
* Bold green for types (e.g. "Int")
* Magenta for the shell prompt (e.g. "scala>")
|
| |/ /
|/| |
| | |
| | |
| | |
| | |
| | | |
When buffering, we must report the ambiguity error to avoid a stack overflow.
When the error refers to erroneous types/symbols,
we don't report it directly to the user,
because there will be an underlying error that's the root cause.
|
|\ \ \
| | | |
| | | | |
Make Scaladoc actually exit with non-zero exit code in case of errors, as its docs say it does
|