| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
| |
Also, added some docs variables to Gen* traits that were missing.
No review.
|
| |
|
|
|
| |
and renames file to avoid ant's brainlessness, no review.
|
| |
|
|
|
| |
No review.
|
| |
|
|
|
| |
No review.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Refactoring the collections api to support differentiation between
referring to a sequential collection and a parallel collection, and to
support referring to both types of collections.
New set of traits Gen* are now superclasses of both their * and Par* subclasses. For example, GenIterable is a superclass of both Iterable and ParIterable. Iterable and ParIterable are not in a subclassing relation. The new class hierarchy is illustrated below (simplified, not all relations and classes are shown):
TraversableOnce --> GenTraversableOnce
^ ^
| |
Traversable --> GenTraversable
^ ^
| |
Iterable --> GenIterable <-- ParIterable
^ ^ ^
| | |
Seq --> GenSeq <-- ParSeq
(the *Like, *View and *ViewLike traits have a similar hierarchy)
General views extract common view functionality from parallel and
sequential collections.
This design also allows for more flexible extensions to the collections
framework. It also allows slowly factoring out common functionality up
into Gen* traits.
From now on, it is possible to write this:
import collection._
val p = parallel.ParSeq(1, 2, 3)
val g: GenSeq[Int] = p // meaning a General Sequence
val s = g.seq // type of s is Seq[Int]
for (elem <- g) {
// do something without guarantees on sequentiality of foreach
// this foreach may be executed in parallel
}
for (elem <- s) {
// do something with a guarantee that foreach is executed in order, sequentially
}
for (elem <- p) {
// do something concurrently, in parallel
}
This also means that some signatures had to be changed. For example,
method `flatMap` now takes `A => GenTraversableOnce[B]`, and `zip` takes
a `GenIterable[B]`.
Also, there are mutable & immutable Gen* trait variants. They have
generic companion functionality.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
More fun with xml and elements not appearing equal. Closes #4387, no
review.
|
| |
|
|
|
|
|
| |
The beautification of repl pasting had the problem that the new
beautiful output was not itself pastable. Now I have achieved "paste
idempotence". No review.
|
| |
|
|
|
|
|
|
|
|
| |
Enhancing the repl-testing code by turning it into a transcript
producing machine. "Here's some code." "Here's a transcript!" "Good day
to you, sir!" "No, good day to YOU!"
These changes are awesome. Look at the checkfile diffs for god's sake,
they'll make you weep with joy. No review.
|
| |
|
|
|
|
| |
Fix for interpreter issue with missing newlines wreaking havoc. No
review.
|
| |
|
|
|
|
| |
A change to a test to defend against output change when run with a
plugin enabled. No review.
|
| |
|
|
|
|
| |
And so my attempt to have a performance test draws the final curtain, no
review.
|
| |
|
|
|
|
|
|
| |
Learned an interesting lesson about having an implicit object with a
"def apply(x: Any): List[String]" method imported into power mode. Let's
just say this is not advised. This and other rough edges eliminated from
power mode, no review.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fix for linked lists closes #4080 and proves my desire not to ship
obviously broken code is even greater than my will to hold out for any
help. I threw in a free fix for this which I noticed while in there.
scala> scala.collection.mutable.LinkedList[Int]().head
res0: Int = 0
Also was reminded how useless tests can be:
val ten = DoubleLinkedList(1 to 10: _*)
ten.insert(DoubleLinkedList(11))
// Post-insert position test
require(ten.last == 11)
Fortunately a test confirming buggy behavior still serves a purpose by
breaking when you fix the bug which allowed it to pass, thus letting you
fix the broken test too. Life's (very) little compensations. Linked list
code should still be presumed broken. No review.
|
| |
|
|
|
|
|
|
|
| |
I guess windows can show a < 500x difference and still give us
confidence there isn't a 100,000x difference. Testing situation
continues to confound and amaze. Trying to make 4279 not fail on
windows, one more time and I will delete it and burn the bodies. No
review.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Added a temporary fix for #4351, but disabled it because the extend
specialized class with nonspecialized type-parameters is used in the
stdlib already.
Disabling scala.parallel package, adding the currently disabled
scala.concurrent package which will be implemented in some of the next
releases.
Review by phaller.
|
| |
|
|
|
|
|
|
| |
Moved failing tests to pending after having diagnosed that the only
problem is a difference in output between Java 1.5 and 1.6 reflection
libraries. Can we moved back once that's fixed. For now, it's more
important to ghet the build back.
|
| |
|
|
|
|
| |
Changed App-using tests to use main() to see if that's our trouble. No
review.
|
| |
|
|
|
|
| |
Spiced up the signature test infrastructure a bunch, wrote some more
tests, restored the tests in pending. No review.
|
| |
|
|
|
|
| |
Moving signature tests to pending, because they operate under wrong
assumptions now that mixed in members are bridges. Review by extempore.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
My early attempts to implement non-integral ranges in a way which was
useful without having lots of floating point traps were less than
successful. One of the bigger backfires is that the requirement not to
round (trying, and failing anyway, to avoid surprises with methods like
"contains") inflicts runtime errors.
The simple way to improve this, which seems a good idea anyway, is to
make the default math context something less inclined to exceptions.
Default BigDecimal mc is now DECIMAL128. References #1812, #4201 and
puts #4201 back to normal priority. Review by community.
|
| |
|
|
|
| |
No review.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Implementing foreach to work in parallel in ParIterableLike.
Doing a bunch of refactoring around in the collection framework to
ensure a parallel foreach is never called with a side-effecting method.
This still leaves other parts of the standard library and the compiler
unguarded.
No review.
|
| |
|
|
|
|
|
| |
Oh the irony, disabling the failing test made the build fail, because
another test is hardcoded to use its paths. Disabled that test too.
We'll put humpty back together again. No review.
|
| |
|
|
|
|
| |
We need a successful build, not sure why this hasn't been disabled yet.
Disabled failing coder test, no review.
|
| |
|
|
|
|
| |
One hundred! One hundred times slower on windows! Ah, ha, ha! Adjusted
test case, no review.
|
| |
|
|
|
|
|
| |
Fix for a big bug in lastIndexOfSlice and some latent negative index
bugs in both that and indexOfSlice. This stuff is taxing. Closes #4348,
no review.
|
| | |
|
| |
|
|
|
|
|
| |
infix implicits to Integral and Fractional. As a bonus this patch
knocked 10,000 long boxings off a specialized test. Who knew. Closes
#4308, #4321, review by community.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Removing toPar* methods, since we've agreed they're difficult to: -
underestand - maintain
Also, changed the docs and some tests appropriately.
Description:
1) Every collection is now parallelizable - switch to the parallel version of the collection is done via `par`.
- Traversable collections and iterators have `par` return a parallel
- collection of type `ParIterable[A]` with the implementation being the
- representative of `ParIterable`s (currently, `ParArray`). Iterable
- collections do the same thing. Sequences refine `par`'s returns type
- to `ParSeq[A]`. Maps and sets do a similar thing.
The above means that the contract for `par` changed - it is no longer guaranteed to be O(1), nor reflect the same underlying data, as was the case for mutable collections before. Method `par` is now at worst linear.
Furthermore, specific collection implementations override `par` to a more efficient alternative - instead of copying the dataset, the dataset is shared between the old and the new version. Implementation complexity may be sublinear or constant in these cases, and the underlying data structure may be shared. Currently, these data structures include parallel arrays, maps and sets, vectors, hash trie maps and sets, and ranges.
Finally, parallel collections implement `par` trivially.
2) Methods `toMap`, `toSet`, `toSeq` and `toIterable` have been refined
for parallel collections to switch between collection types, however,
they never switch an implementation from parallel to sequential. They
may or may not copy the elements, as is the case with sequential
variants of these methods.
3) The preferred way to switch between different collection types,
whether maps, sets and seqs, or parallel and sequential, is now via use
of methods `toIterable`, `toSeq`, `toSet` and `toMap` in combination
with `par` and `seq`.
Review by odersky.
|
| | |
|
| |
|
|
|
| |
Disabled test not actually doing anything anyway, no review.
|
| |
|
|
|
|
|
|
|
| |
Always forget that checking system properties causes exceptions in
applets and such. Made the system property wrapper wrap its access
checks in some more wrapping. I spent a long time trying to write a
test for the security manager but it's hopeless without knowing all the
details of the test environment. Closes #4346, no review.
|
| |
|
|
|
|
|
|
|
|
|
| |
warnings due to my changing a map from mutable to immutable (which
ought to be the good direction) because "def update" still lingers on
immutable maps. I counted the days elapsed since it was marked for death
(before 2.8.0) and added in the bugliness of what I was looking at and
bid it farewell.
Now removed: def update on immutable maps. No review.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Various chronic annoyances with the repl addressed. Much improved
transcript pasting. Now goes back in time to fix the transcript if it
contains self-referential "res0, res1" etc. so that it works as it
originally did. Shows which commands it is running, and places the
commands with their result in a manner suitable for framing.
Also, a new :paste command which accepts input up to ctrl-D, so you can
enter companions without gyrations, or code from people who write in a
repl unfriendly fashion by putting their curly braces on the next line
(I'm looking at you mark harrah) or you name it, it's not picky.
No review.
|
| | |
|
| |
|
|
|
|
|
|
| |
Raised the threshold for the test looking at view performance. I guess
if views are only 50x as slow on windows that's still a lot better than
100,000. (The real issue is more like "it's windows" plus "it's windows
running on virtualbox".) No review.
|
| |
|
|
|
|
|
|
|
|
|
| |
A small addition to the library to address something bugging me forever.
It's a light interface to system properties. It's not intended to solve
all property issues for all time, only to greatly improve on the overly
ad-hoc ways things are presently done. Feedback welcome. Sorry it's
coming in this late but it arises from writing the tools to fix the bugs
to allow that release to happen. That's nature's circle of bugs. Review
by community.
|
| |
|
|
|
| |
for pointing me toward the problem. No review.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Almost all view classes now list parents like
trait Appended[B >: A] extends super.Appended[B] with Transformed[B]
instead of the former
trait Appended[B >: A] extends Transformed[B] with super.Appended[B]
because as it was, the implementation of foreach in
TraversableViewLike#Transformed was repeatedly trumping overrides found
in e.g. IterableLike. This change was not without its own consequences,
and much of the rest of the patch is dealing with that. A more general
issue is clearly revealed here: there is no straightforward way to deal
with trait composition and overrides when some methods should prefer B
over A and some the reverse. (It's more like A through Z in this case.)
That closes #4279, with some views being five orders of magnitude slower
than necessary. There is a test that confirms they'll stay performance
neighbors.
In the view classes (Zipped, Mapped, etc.) I attended to them with
comb and brush until they were reasonably consistent. I only use
"override" where necessary and throw in some "final" in the interests
of trying to anchor the composition outcome. I also switched the
newSliced, newZipped, etc. methods to use early init syntax since a
number have abstract vals and I found at least one bug originating with
uninitialized access.
There was a piece of a parallel collections scalacheck test failing,
which
I disabled out of expedience - am emailing prokopec.
There is plenty of work left to do but paulp must get back to other 2.9
issues. This is the Zurich->SF airplane patch. No review.
|
| |
|
|
|
|
|
|
| |
Adding special take and drop for numeric ranges, and a test. Parallel
numeric ranges are added, but currently disabled.
Review by extempore.
|
| |
|
|
|
|
|
|
|
|
|
| |
An overhaul of slice and related implementations (primarily that is
drop and take.) In the course of trying to get it working consistently
(mostly with respect to negative indices, which were dealt with
arbitrarily differently across the 25+ concrete implementations) I fixed
various bugs.
Closes #4288, no review.
|
| |
|
|
|
| |
No review.
|
| |
|
|
|
|
|
| |
More signature fixes and tests and generally being more sophisticated
about our primitive squashing. These signatures belong in world-class
museums but we are their shepherds for now. Closes #4317, no review.
|
| |
|
|
|
|
| |
More signature work arising from actually looking at the signatures.
Slightly improved the coverage of the tests. No review.
|
| |
|
|
|
|
|
| |
And extempore's original plan carries the day: null.## no longer throws
an NPE. Took advantage of feature to simplify a bunch of sharp-sharp
calculations. Closes #4311, no review.
|
| |
|
|
|
| |
Changed class name from `Main` to `Test` in a failing test case.
|
