a kill to a view

is there a working alternative?

Am 17.12.2011 05:55, schrieb Paul Phillips:
> Speaking of things not pulling their weight: views.
>
> Library jar before and after ripping out views.
>
> % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
> -rw-r--r-- 1 paulp admin 7161826 Dec 16 19:53
> /scratch/trunk2/build/pack/lib/scala-library.jar
> % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
> -rw-r--r-- 1 paulp admin 6561034 Dec 16 19:51
> /scratch/trunk3/build/pack/lib/scala-library.jar
>
> Statistics from the scala code in my general ~/src directory.
>
> % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
> 904766
> % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
> -v ':(import|package)' | grep -v deprecated |wc -l
> 92
>
> One view usage per 10,000 lines of scala code. (And that may be
> generous, as some of those 92 hits are either copies of compiler/library
> source or things testing the compiler/library source. However I could
> have missed some dotless uses so we'll call it a wash.)
>
> So a feature used in .01% of lines of scala written is consuming 9% of
> the jar. I focus on 9% not because I'm obsessed with jar size but as a
> proxy for all the overhead associated with any large body of code. And
> views have not been small in the overhead department, and they're still
> not close to being a robust unit.
>
> https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&reque...
>
> Displaying issues 1 to 49 of 49 matching issues. [...]
>
> Solution: exile to separate jar, deprecate for 2.10, and retire
> them afterward. They can easily be pimped in by any who wish to use
> them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
> make things very slightly easier on the .01%. (I'm not even sure it
> makes things slightly easier.) And maybe many view designs will flourish
> in this brave new viewless world.
>
> If we were determined not to let those view methods go away, we could
> stub them as empty methods in the core library and have some way to
> configure them to invoke a library elsewhere. Anything to get them out of the
> main collections inheritance hierarchy.
>
>
> *** FOOTNOTES ***
>
> Enclosed for any who may wish to validate my experimental results.
>
> [1] the list of projects containing those 900,000 lines of scala
>
> Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
> InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
> a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
> argot ashlar asm bash-completion bash-completion-lib blueeyes
> blueprint borachio brepl caliper categories cccp chromium-tabs
> collection-tests config-typesafe coreen d3 derivative-combinators
> dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
> extractors factorie files.txt finagle findbugs git-prompt git-pulls
> git-sbt-processor git-scripts gitblit github-gem github-services
> github-terminal gitignore gll-combinators gmailr google-code-prettify
> google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
> guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
> izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
> jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
> kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
> magpie maven-sbt maven-scala-plugin meow metascala minikanren
> mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
> ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
> parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
> projectplus pythagoras quickcursor rapture-io recursivity-commons
> replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
> scala-cel scala-collection-test scala-datetime scala-graph scala-ide
> scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
> scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
> scala-time scalabot scalacheck scalaconsole scalacs scalafs
> scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
> scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
> scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
> sonatype-aether source.tmbundle spakle spark spde specs specs2
> specs2-test sperformance standard-project stopwatch talking-puffin texto
> threeten treetop twitter-util uniscala unplanned up vibrantinklion
> view-files.txt vscaladoc windmill yarrgs yeti zmpp2
>
>
> [2] the view hits.
>
>
>
> anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
> val b = VectorCase.newBuilder[B] ++= g.view.take(index)
> anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
> val els = depthFirst.view map { case (p,v) =>
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430: val
> condition = matchCondition(matchCase.view, names)
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427: matchCondition(target,
> targetType, caseAST.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169: matchCondition(it.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140: matchCondition(m.view,
> types),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299: it.view.values.foreach(
> arg => walkOps(ctx, arg) )
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303: it.view.checkMethod,
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305: (it.view.valueTypes.map(t
> => T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325: val
> bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
> case (tmpCtx, (expVar, idx)) =>
> blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
> TemporalCacheState(removed.view.map(t => (t._1,
> t._2.value)).toMap, TemporalCache(kept.toMap))
> factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
> override def toString = getClass.getName+"("+"len="+length+"
> 1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
> factorie/src/test/scala/cc/factorie/TestUtils.scala:32: def
> remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
> seq.view(lastElemZipIndex(s)+1, seq.length)
> mada/src/main/scala/sequence/vector/Scala.scala:17: for (e <- _1.view) {
> mada/src/test/scala/sequence/iterative/ConversionTest.scala:53: x.view
> mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
> candsAndLabels.view.unzip.zipped foreach emitBranch
> ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
> val indices = for ((param, idx) <- params.view.zipWithIndex
> ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28: *
> val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
> override def view = self.view
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
> override def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
> override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
> override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
> override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
> override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/SeqLike.scala:518: b ++=
> toCollection(rest).view drop replaced
> ozma/src/scalalib/scala/collection/SeqLike.scala:531: b ++=
> toCollection(rest).view.tail
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69: override
> def view = self.view
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70: override
> def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
> override def view = self.view
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
> override def view(from: Int, until: Int): TraversableView[A, Repr] =
> self.view(from, until)
> ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
> ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
> def result() = { builder.result.view }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
> def result() = { builder.result.view(min, max) }
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
> nodes.toList.view map nodeDegree filter degreeFilter min
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
> nodes.toList.view map nodeDegree filter degreeFilter max
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
> val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
> else nodes.view map nodeDegree filter
> degreeFilter)
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
> (for (jsonNode <- nodeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
> (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
> (for (jsonEdge <- edgeList.view) yield {
> scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
> val toRead = data.view.slice(position.toInt, min(position.toInt +
> dst.limit-dst.position, data.size))
> scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
> lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
> scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
> override def view = super.view
> scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
> val list = (f(expectedData().toList.view) map { i => expectedCount
> += 1; i })
> scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42: new
> Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
> => new Grouping(Node(c))).toList)
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
> m <- getClass().getMethods.view
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
> m <- getClass().getMethods.view
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
> tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
> {sigBytesFromAnnotation} headOption
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
> splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
> headOption
> Scalala/src/main/scala/scalala/library/Library.scala:213:
> matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Library.scala:223:
> matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Plotting.scala:464://
> val values = items.view.map(item =>
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:
>
> SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
> if (f.isInstanceOf[Transformer]) f else f.andThen((i :
> Iterable[String]) => i.view);
> scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
> peeked.view(0, n);
> scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
> for( datum <- range.view map data) {
> scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
> val (grad: DenseVector[Double],prob: Double) =
> eCounts.zipWithIndex.par.view.map { case (vec,c) =>
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
> val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
> ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
> scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
> context.view(model)
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
> context.view(it.get.asInstanceOf[AnyRef])
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
> override def requiredViewClass(): java.lang.Class[_] =
> classOf[org.fusesource.scalate.spring.view.ScalateView]
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
> file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
> directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
> children(file).view.flatMap(andDescendants(_)))
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
> constructors.view.map(c =>
> tryCreate(c)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
> sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
> scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
> sourceDirectories.view.map
> {findMatching(_)}.find(_.isDefined).getOrElse(None)
> scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23: if
> (len != -1) out ++= buffer.view(0, len)
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
> argumentsFragments.view.zip(nestedArguments).collect { case
> (ApplicableArguments(value), args) => (value, args) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
> def fragmentAndSpecNames: Seq[(T, SpecName)] =
> argumentsFragments.view.zip(nestedSpecNames).collect { case
> (ApplicableArguments(value), name) => (value, name) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
> argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
> { case ((ApplicableArguments(value), args), name) => (value, args,
> name) }
> specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
> specificationNames(path, pattern, basePath,
> verbose).view.filter(filter).flatMap(n => createSpecification(n))
> sperformance/src/test/scala/CollectionsShootout.scala:148:
> List.range(1, size).toList.view
> sperformance/src/test/scala/CollectionsShootout.scala:155:
> ArrayBuffer.range(1, size).view
> sperformance/src/test/scala/CollectionsShootout.scala:184:
> col.view.map(_ * 2).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:189:
> col.view.filter(_ % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:194:
> col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:236:
> performance of "Vector.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:239:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:244:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> sperformance/src/test/scala/CollectionsShootout.scala:260:
> performance of "ArrayBuffer.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:263:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:268:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
> val result = codecs.urlParamCodecs.view map { codec =>
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
> val result = codecs.urlParamCodecs.view map { codec =>
>

I don't understand the question :-(

On Dec 17, 2011 11:35 AM, "HamsterofDeath" <h-star@gmx.de> wrote:

is there a working alternative?

Am 17.12.2011 05:55, schrieb Paul Phillips:
> Speaking of things not pulling their weight: views.
>
> Library jar before and after ripping out views.
>
>     % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
>     -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
> /scratch/trunk2/build/pack/lib/scala-library.jar
>     % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
>     -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
> /scratch/trunk3/build/pack/lib/scala-library.jar
>
> Statistics from the scala code in my general ~/src directory.
>
>     % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
>       904766
>     % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
> -v ':(import|package)' | grep -v deprecated |wc -l
>           92
>
> One view usage per 10,000 lines of scala code. (And that may be
> generous, as some of those 92 hits are either copies of compiler/library
> source or things testing the compiler/library source. However I could
> have missed some dotless uses so we'll call it a wash.)
>
> So a feature used in .01% of lines of scala written is consuming 9% of
> the jar. I focus on 9% not because I'm obsessed with jar size but as a
> proxy for all the overhead associated with any large body of code. And
> views have not been small in the overhead department, and they're still
> not close to being a robust unit.
>
>   https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906
>
>   Displaying issues 1 to 49 of 49 matching issues. [...]
>
> Solution: exile to separate jar, deprecate for 2.10, and retire
> them afterward. They can easily be pimped in by any who wish to use
> them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
> make things very slightly easier on the .01%. (I'm not even sure it
> makes things slightly easier.) And maybe many view designs will flourish
> in this brave new viewless world.
>
> If we were determined not to let those view methods go away, we could
> stub them as empty methods in the core library and have some way to
> configure them to invoke a library elsewhere.  Anything to get them out of the
> main collections inheritance hierarchy.
>
>
> *** FOOTNOTES ***
>
> Enclosed for any who may wish to validate my experimental results.
>
> [1] the list of projects containing those 900,000 lines of scala
>
> Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
> InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
> a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
> argot ashlar asm bash-completion bash-completion-lib blueeyes
> blueprint borachio brepl caliper categories cccp chromium-tabs
> collection-tests config-typesafe coreen d3 derivative-combinators
> dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
> extractors factorie files.txt finagle findbugs git-prompt git-pulls
> git-sbt-processor git-scripts gitblit github-gem github-services
> github-terminal gitignore gll-combinators gmailr google-code-prettify
> google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
> guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
> izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
> jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
> kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
> magpie maven-sbt maven-scala-plugin meow metascala minikanren
> mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
> ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
> parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
> projectplus pythagoras quickcursor rapture-io recursivity-commons
> replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
> scala-cel scala-collection-test scala-datetime scala-graph scala-ide
> scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
> scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
> scala-time scalabot scalacheck scalaconsole scalacs scalafs
> scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
> scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
> scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
> sonatype-aether source.tmbundle spakle spark spde specs specs2
> specs2-test sperformance standard-project stopwatch talking-puffin texto
> threeten treetop twitter-util uniscala unplanned up vibrantinklion
> view-files.txt vscaladoc windmill yarrgs yeti zmpp2
>
>
> [2] the view hits.
>
>
>
> anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
> val b = VectorCase.newBuilder[B] ++= g.view.take(index)
> anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
>    val els = depthFirst.view map { case (p,v) =>
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430:              val
> condition =   matchCondition(matchCase.view, names)
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427:           matchCondition(target,
> targetType, caseAST.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169:                        matchCondition(it.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140:           matchCondition(m.view,
> types),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299:                  it.view.values.foreach(
> arg => walkOps(ctx, arg) )
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303:                          it.view.checkMethod,
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305:                          (it.view.valueTypes.map(t
> => T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325:                  val
> bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
> case (tmpCtx, (expVar, idx)) =>
> blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
>      TemporalCacheState(removed.view.map(t => (t._1,
> t._2.value)).toMap, TemporalCache(kept.toMap))
> factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
> override def toString = getClass.getName+"("+"len="+length+"
> 1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
> factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
> remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
> seq.view(lastElemZipIndex(s)+1, seq.length)
> mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <- _1.view) {
> mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:        x.view
> mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
>        candsAndLabels.view.unzip.zipped foreach emitBranch
> ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
>       val indices = for ((param, idx) <- params.view.zipWithIndex
> ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
>  val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
> override def view = self.view
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
> override def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
>    override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
>    override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
> override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
>  override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
> toCollection(rest).view drop replaced
> ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
> toCollection(rest).view.tail
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
> def view = self.view
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
> def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
> override def view = self.view
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
> override def view(from: Int, until: Int): TraversableView[A, Repr] =
> self.view(from, until)
> ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
> ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
>    def result() = { builder.result.view }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
>    def result() = { builder.result.view(min, max) }
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
>      nodes.toList.view map nodeDegree filter degreeFilter min
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
>      nodes.toList.view map nodeDegree filter degreeFilter max
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
>    val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
>    else                          nodes.view map nodeDegree filter
> degreeFilter)
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
>            (for (jsonNode <- nodeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
>                (for (jsonEdge <- edgeList.view) yield {
> scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
>      val toRead = data.view.slice(position.toInt, min(position.toInt +
> dst.limit-dst.position, data.size))
> scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
>   lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
> scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
>  override def view = super.view
> scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
>    val list = (f(expectedData().toList.view) map { i => expectedCount
> += 1; i })
> scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42:    new
> Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
> => new Grouping(Node(c))).toList)
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
> m <- getClass().getMethods.view
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
> m <- getClass().getMethods.view
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
>  tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
> {sigBytesFromAnnotation} headOption
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
>  splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
> headOption
> Scalala/src/main/scala/scalala/library/Library.scala:213:
> matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Library.scala:223:
> matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Plotting.scala:464://
> val values = items.view.map(item =>
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:
>
> SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
>      if (f.isInstanceOf[Transformer]) f else f.andThen((i :
> Iterable[String]) => i.view);
> scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
> peeked.view(0, n);
> scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
>        for( datum <- range.view map data) {
> scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
>    val (grad: DenseVector[Double],prob: Double) =
> eCounts.zipWithIndex.par.view.map { case (vec,c) =>
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
>    val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
>    ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
> scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
>    context.view(model)
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
>    context.view(it.get.asInstanceOf[AnyRef])
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
>  override def requiredViewClass(): java.lang.Class[_] =
> classOf[org.fusesource.scalate.spring.view.ScalateView]
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
>      file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
>    directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
>    children(file).view.flatMap(andDescendants(_)))
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
>    constructors.view.map(c =>
> tryCreate(c)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
>    sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
> scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
>    sourceDirectories.view.map
> {findMatching(_)}.find(_.isDefined).getOrElse(None)
> scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23:                       if
> (len != -1)   out ++= buffer.view(0, len)
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
> argumentsFragments.view.zip(nestedArguments).collect { case
> (ApplicableArguments(value), args) => (value, args) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
> def fragmentAndSpecNames: Seq[(T, SpecName)] =
> argumentsFragments.view.zip(nestedSpecNames).collect { case
> (ApplicableArguments(value), name) => (value, name) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
> argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
> { case ((ApplicableArguments(value), args), name) => (value, args,
> name) }
> specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
>    specificationNames(path, pattern, basePath,
> verbose).view.filter(filter).flatMap(n => createSpecification(n))
> sperformance/src/test/scala/CollectionsShootout.scala:148:
> List.range(1, size).toList.view
> sperformance/src/test/scala/CollectionsShootout.scala:155:
> ArrayBuffer.range(1, size).view
> sperformance/src/test/scala/CollectionsShootout.scala:184:
> col.view.map(_ * 2).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:189:
> col.view.filter(_ % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:194:
> col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:236:
> performance of "Vector.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:239:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:244:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> sperformance/src/test/scala/CollectionsShootout.scala:260:
> performance of "ArrayBuffer.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:263:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:268:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
>      val result = codecs.urlParamCodecs.view map { codec =>
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
>      val result = codecs.urlParamCodecs.view map { codec =>
>

i had a look at the various issues. how many methods need fixing? and
are all fixable, and do they even make sense? for example, there is one
issue about Set(1,2,3).view.map(_ => 1).size==3 being true. i see why
this is unexpected, but implementing map in such a way that it checks
for duplicated elements without using a temporary collection is
impossible - and using one defeats the purpose of views.

or the view.dropRight-problem. how is a view supposed to know when it
has reached the last-nth item without traversing the whole collection
first (or asking for its size which might be equal to traversing it)?

my suggestion would be to fix the problems (at least the unsupported
operation exceptions) by splitting up traversable into
stricttraversable, nonstricttraversable and
idontcarewhichoneitistraversable which they both extend from. similar to
the .par & .seq collections. all unsupported operations like dropright
would simply not exist in the trait, and no one would try to call them.

i must admit it's just an idea and i haven't checked if it might work at all

Am 17.12.2011 11:35, schrieb HamsterofDeath:
> is there a working alternative?
>
> Am 17.12.2011 05:55, schrieb Paul Phillips:
>> Speaking of things not pulling their weight: views.
>>
>> Library jar before and after ripping out views.
>>
>> % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
>> -rw-r--r-- 1 paulp admin 7161826 Dec 16 19:53
>> /scratch/trunk2/build/pack/lib/scala-library.jar
>> % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
>> -rw-r--r-- 1 paulp admin 6561034 Dec 16 19:51
>> /scratch/trunk3/build/pack/lib/scala-library.jar
>>
>> Statistics from the scala code in my general ~/src directory.
>>
>> % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
>> 904766
>> % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
>> -v ':(import|package)' | grep -v deprecated |wc -l
>> 92
>>
>> One view usage per 10,000 lines of scala code. (And that may be
>> generous, as some of those 92 hits are either copies of compiler/library
>> source or things testing the compiler/library source. However I could
>> have missed some dotless uses so we'll call it a wash.)
>>
>> So a feature used in .01% of lines of scala written is consuming 9% of
>> the jar. I focus on 9% not because I'm obsessed with jar size but as a
>> proxy for all the overhead associated with any large body of code. And
>> views have not been small in the overhead department, and they're still
>> not close to being a robust unit.
>>
>> https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&reque...
>>
>> Displaying issues 1 to 49 of 49 matching issues. [...]
>>
>> Solution: exile to separate jar, deprecate for 2.10, and retire
>> them afterward. They can easily be pimped in by any who wish to use
>> them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
>> make things very slightly easier on the .01%. (I'm not even sure it
>> makes things slightly easier.) And maybe many view designs will flourish
>> in this brave new viewless world.
>>
>> If we were determined not to let those view methods go away, we could
>> stub them as empty methods in the core library and have some way to
>> configure them to invoke a library elsewhere. Anything to get them out of the
>> main collections inheritance hierarchy.
>>
>>
>> *** FOOTNOTES ***
>>
>> Enclosed for any who may wish to validate my experimental results.
>>
>> [1] the list of projects containing those 900,000 lines of scala
>>
>> Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
>> InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
>> a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
>> argot ashlar asm bash-completion bash-completion-lib blueeyes
>> blueprint borachio brepl caliper categories cccp chromium-tabs
>> collection-tests config-typesafe coreen d3 derivative-combinators
>> dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
>> extractors factorie files.txt finagle findbugs git-prompt git-pulls
>> git-sbt-processor git-scripts gitblit github-gem github-services
>> github-terminal gitignore gll-combinators gmailr google-code-prettify
>> google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
>> guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
>> izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
>> jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
>> kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
>> magpie maven-sbt maven-scala-plugin meow metascala minikanren
>> mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
>> ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
>> parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
>> projectplus pythagoras quickcursor rapture-io recursivity-commons
>> replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
>> scala-cel scala-collection-test scala-datetime scala-graph scala-ide
>> scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
>> scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
>> scala-time scalabot scalacheck scalaconsole scalacs scalafs
>> scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
>> scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
>> scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
>> sonatype-aether source.tmbundle spakle spark spde specs specs2
>> specs2-test sperformance standard-project stopwatch talking-puffin texto
>> threeten treetop twitter-util uniscala unplanned up vibrantinklion
>> view-files.txt vscaladoc windmill yarrgs yeti zmpp2
>>
>>
>> [2] the view hits.
>>
>>
>>
>> anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
>> val b = VectorCase.newBuilder[B] ++= g.view.take(index)
>> anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
>> val els = depthFirst.view map { case (p,v) =>
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430: val
>> condition = matchCondition(matchCase.view, names)
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427: matchCondition(target,
>> targetType, caseAST.view),
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169: matchCondition(it.view),
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140: matchCondition(m.view,
>> types),
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299: it.view.values.foreach(
>> arg => walkOps(ctx, arg) )
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303: it.view.checkMethod,
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305: (it.view.valueTypes.map(t
>> => T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
>> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325: val
>> bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
>> case (tmpCtx, (expVar, idx)) =>
>> blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
>> TemporalCacheState(removed.view.map(t => (t._1,
>> t._2.value)).toMap, TemporalCache(kept.toMap))
>> factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
>> override def toString = getClass.getName+"("+"len="+length+"
>> 1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
>> factorie/src/test/scala/cc/factorie/TestUtils.scala:32: def
>> remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
>> seq.view(lastElemZipIndex(s)+1, seq.length)
>> mada/src/main/scala/sequence/vector/Scala.scala:17: for (e <- _1.view) {
>> mada/src/test/scala/sequence/iterative/ConversionTest.scala:53: x.view
>> mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
>> candsAndLabels.view.unzip.zipped foreach emitBranch
>> ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
>> val indices = for ((param, idx) <- params.view.zipWithIndex
>> ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28: *
>> val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
>> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
>> override def view = self.view
>> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
>> override def view(from: Int, until: Int) = self.view(from, until)
>> ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
>> override def seq = self.seq.view
>> ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
>> override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
>> CollSeq]]
>> ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
>> override def seq = self.seq.view
>> ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
>> override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
>> CollSeq]]
>> ozma/src/scalalib/scala/collection/SeqLike.scala:518: b ++=
>> toCollection(rest).view drop replaced
>> ozma/src/scalalib/scala/collection/SeqLike.scala:531: b ++=
>> toCollection(rest).view.tail
>> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69: override
>> def view = self.view
>> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70: override
>> def view(from: Int, until: Int) = self.view(from, until)
>> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
>> override def view = self.view
>> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
>> override def view(from: Int, until: Int): TraversableView[A, Repr] =
>> self.view(from, until)
>> ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
>> ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
>> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
>> def result() = { builder.result.view }
>> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
>> def result() = { builder.result.view(min, max) }
>> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
>> nodes.toList.view map nodeDegree filter degreeFilter min
>> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
>> nodes.toList.view map nodeDegree filter degreeFilter max
>> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
>> val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
>> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
>> else nodes.view map nodeDegree filter
>> degreeFilter)
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
>> (for (jsonNode <- nodeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
>> (for (jsonEdge <- edgeList.view) yield {
>> scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
>> val toRead = data.view.slice(position.toInt, min(position.toInt +
>> dst.limit-dst.position, data.size))
>> scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
>> lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
>> scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
>> override def view = super.view
>> scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
>> val list = (f(expectedData().toList.view) map { i => expectedCount
>> += 1; i })
>> scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42: new
>> Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
>> => new Grouping(Node(c))).toList)
>> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
>> m <- getClass().getMethods.view
>> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
>> m <- getClass().getMethods.view
>> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
>> tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
>> {sigBytesFromAnnotation} headOption
>> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
>> splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
>> headOption
>> Scalala/src/main/scala/scalala/library/Library.scala:213:
>> matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
>> Scalala/src/main/scala/scalala/library/Library.scala:223:
>> matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
>> Scalala/src/main/scala/scalala/library/Plotting.scala:464://
>> val values = items.view.map(item =>
>> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:
>>
>> SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
>> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
>> if (f.isInstanceOf[Transformer]) f else f.andThen((i :
>> Iterable[String]) => i.view);
>> scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
>> peeked.view(0, n);
>> scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
>> for( datum <- range.view map data) {
>> scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
>> val (grad: DenseVector[Double],prob: Double) =
>> eCounts.zipWithIndex.par.view.map { case (vec,c) =>
>> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
>> val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
>> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
>> ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
>> scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
>> context.view(model)
>> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
>> context.view(it.get.asInstanceOf[AnyRef])
>> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
>> override def requiredViewClass(): java.lang.Class[_] =
>> classOf[org.fusesource.scalate.spring.view.ScalateView]
>> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
>> file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
>> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
>> directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
>> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
>> children(file).view.flatMap(andDescendants(_)))
>> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
>> constructors.view.map(c =>
>> tryCreate(c)).find(_.isDefined).getOrElse(None)
>> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
>> sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
>> scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
>> sourceDirectories.view.map
>> {findMatching(_)}.find(_.isDefined).getOrElse(None)
>> scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23: if
>> (len != -1) out ++= buffer.view(0, len)
>> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
>> argumentsFragments.view.zip(nestedArguments).collect { case
>> (ApplicableArguments(value), args) => (value, args) }
>> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
>> def fragmentAndSpecNames: Seq[(T, SpecName)] =
>> argumentsFragments.view.zip(nestedSpecNames).collect { case
>> (ApplicableArguments(value), name) => (value, name) }
>> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
>> argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
>> { case ((ApplicableArguments(value), args), name) => (value, args,
>> name) }
>> specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
>> specificationNames(path, pattern, basePath,
>> verbose).view.filter(filter).flatMap(n => createSpecification(n))
>> sperformance/src/test/scala/CollectionsShootout.scala:148:
>> List.range(1, size).toList.view
>> sperformance/src/test/scala/CollectionsShootout.scala:155:
>> ArrayBuffer.range(1, size).view
>> sperformance/src/test/scala/CollectionsShootout.scala:184:
>> col.view.map(_ * 2).take(100).force
>> sperformance/src/test/scala/CollectionsShootout.scala:189:
>> col.view.filter(_ % 2 == 0).take(100).force
>> sperformance/src/test/scala/CollectionsShootout.scala:194:
>> col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
>> sperformance/src/test/scala/CollectionsShootout.scala:236:
>> performance of "Vector.view" in {
>> sperformance/src/test/scala/CollectionsShootout.scala:239:
>> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
>> (x._2%3==0)).map(_._1).force
>> sperformance/src/test/scala/CollectionsShootout.scala:244:
>> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
>> sperformance/src/test/scala/CollectionsShootout.scala:260:
>> performance of "ArrayBuffer.view" in {
>> sperformance/src/test/scala/CollectionsShootout.scala:263:
>> a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
>> (x._2%3==0)).map(_._1).force
>> sperformance/src/test/scala/CollectionsShootout.scala:268:
>> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
>> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
>> val result = codecs.urlParamCodecs.view map { codec =>
>> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
>> val result = codecs.urlParamCodecs.view map { codec =>
>>
>

-1000 to removing views.

Cleaning them up and reducing overhead is great, but views are crucial to the collections library.  My guess is that people don't know them well enough to use them or are suffering from view's poor zip performance ...

In any case, something that removes views I am against.  They're a perfect solution when you need them.

On Dec 16, 2011 11:55 PM, "Paul Phillips" <paulp@improving.org> wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

   % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
   % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

   % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
     904766
   % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
         92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

 https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

 Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.


*** FOOTNOTES ***

Enclosed for any who may wish to validate my experimental results.

[1] the list of projects containing those 900,000 lines of scala

Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
argot ashlar asm bash-completion bash-completion-lib blueeyes
blueprint borachio brepl caliper categories cccp chromium-tabs
collection-tests config-typesafe coreen d3 derivative-combinators
dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
extractors factorie files.txt finagle findbugs git-prompt git-pulls
git-sbt-processor git-scripts gitblit github-gem github-services
github-terminal gitignore gll-combinators gmailr google-code-prettify
google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
magpie maven-sbt maven-scala-plugin meow metascala minikanren
mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
projectplus pythagoras quickcursor rapture-io recursivity-commons
replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
scala-cel scala-collection-test scala-datetime scala-graph scala-ide
scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
scala-time scalabot scalacheck scalaconsole scalacs scalafs
scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
sonatype-aether source.tmbundle spakle spark spde specs specs2
specs2-test sperformance standard-project stopwatch talking-puffin texto
threeten treetop twitter-util uniscala unplanned up vibrantinklion
view-files.txt vscaladoc windmill yarrgs yeti zmpp2


[2] the view hits.



anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
val b = VectorCase.newBuilder[B] ++= g.view.take(index)
anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
  val els = depthFirst.view map { case (p,v) =>
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430:                val
condition =     matchCondition(matchCase.view, names)
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427:             matchCondition(target,
targetType, caseAST.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169:                  matchCondition(it.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140:             matchCondition(m.view,
types),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299:                    it.view.values.foreach(
arg => walkOps(ctx, arg) )
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303:                            it.view.checkMethod,
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305:                            (it.view.valueTypes.map(t
=> T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325:                    val
bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
case (tmpCtx, (expVar, idx)) =>
blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
    TemporalCacheState(removed.view.map(t => (t._1,
t._2.value)).toMap, TemporalCache(kept.toMap))
factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
override def toString = getClass.getName+"("+"len="+length+"
1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
seq.view(lastElemZipIndex(s)+1, seq.length)
mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <- _1.view) {
mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:        x.view
mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
      candsAndLabels.view.unzip.zipped foreach emitBranch
ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
     val indices = for ((param, idx) <- params.view.zipWithIndex
ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
 val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
override def view = self.view
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
override def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
  override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
  override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
 override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
toCollection(rest).view drop replaced
ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
toCollection(rest).view.tail
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
def view = self.view
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
override def view = self.view
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
override def view(from: Int, until: Int): TraversableView[A, Repr] =
self.view(from, until)
ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
  def result() = { builder.result.view }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
  def result() = { builder.result.view(min, max) }
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
    nodes.toList.view map nodeDegree filter degreeFilter min
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
    nodes.toList.view map nodeDegree filter degreeFilter max
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
  val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
  else                          nodes.view map nodeDegree filter
degreeFilter)
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
          (for (jsonNode <- nodeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
              (for (jsonEdge <- edgeList.view) yield {
scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
    val toRead = data.view.slice(position.toInt, min(position.toInt +
dst.limit-dst.position, data.size))
scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
 lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
 override def view = super.view
scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
  val list = (f(expectedData().toList.view) map { i => expectedCount
+= 1; i })
scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42:    new
Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
=> new Grouping(Node(c))).toList)
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
m <- getClass().getMethods.view
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
m <- getClass().getMethods.view
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
 tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
{sigBytesFromAnnotation} headOption
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
 splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
headOption
Scalala/src/main/scala/scalala/library/Library.scala:213:
matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Library.scala:223:
matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Plotting.scala:464://
val values = items.view.map(item =>
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:

SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
    if (f.isInstanceOf[Transformer]) f else f.andThen((i :
Iterable[String]) => i.view);
scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
peeked.view(0, n);
scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
      for( datum <- range.view map data) {
scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
  val (grad: DenseVector[Double],prob: Double) =
eCounts.zipWithIndex.par.view.map { case (vec,c) =>
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
  val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
  ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
  context.view(model)
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
  context.view(it.get.asInstanceOf[AnyRef])
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
 override def requiredViewClass(): java.lang.Class[_] =
classOf[org.fusesource.scalate.spring.view.ScalateView]
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
    file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
  directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
  children(file).view.flatMap(andDescendants(_)))
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
  constructors.view.map(c =>
tryCreate(c)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
  sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
  sourceDirectories.view.map
{findMatching(_)}.find(_.isDefined).getOrElse(None)
scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23:                 if
(len != -1)     out ++= buffer.view(0, len)
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
argumentsFragments.view.zip(nestedArguments).collect { case
(ApplicableArguments(value), args) => (value, args) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
def fragmentAndSpecNames: Seq[(T, SpecName)] =
argumentsFragments.view.zip(nestedSpecNames).collect { case
(ApplicableArguments(value), name) => (value, name) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
{ case ((ApplicableArguments(value), args), name) => (value, args,
name) }
specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
  specificationNames(path, pattern, basePath,
verbose).view.filter(filter).flatMap(n => createSpecification(n))
sperformance/src/test/scala/CollectionsShootout.scala:148:
List.range(1, size).toList.view
sperformance/src/test/scala/CollectionsShootout.scala:155:
ArrayBuffer.range(1, size).view
sperformance/src/test/scala/CollectionsShootout.scala:184:
col.view.map(_ * 2).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:189:
col.view.filter(_ % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:194:
col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:236:
performance of "Vector.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:239:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:244:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
sperformance/src/test/scala/CollectionsShootout.scala:260:
performance of "ArrayBuffer.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:263:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:268:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
    val result = codecs.urlParamCodecs.view map { codec =>
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
    val result = codecs.urlParamCodecs.view map { codec =>

On Sat, Dec 17, 2011 at 1:37 PM, Josh Suereth wrote:
> -1000 to removing views.
>
> Cleaning them up and reducing overhead is great, but views are crucial to
> the collections library.  My guess is that people don't know them well
> enough to use them or are suffering from view's poor zip performance ...
>
> In any case, something that removes views I am against.  They're a perfect
> solution when you need them.
>

I agree. Reasons to keep views:

- They are hard to do, so rolling your own is a pain
- They convert some algorithms from N space and time to constant
space and time
- They are the equivalent of the standard behavior of Java's collections.

Views seem to be underused. But I think that reflects the fact that, often,
performance is not critical, so the standard behavior of collections
is fine. But when you need them, you really do need them.

Cheers

On Sat, Dec 17, 2011 at 4:37 AM, Josh Suereth wrote:
> Cleaning them up and reducing overhead is great, but views are crucial to
> the collections library.  My guess is that people don't know them well
> enough to use them or are suffering from view's poor zip performance ...

Maybe they just prefer not to roll the dice on runtime errors:

https://issues.scala-lang.org/browse/SI-4332

I agree with the last comment in that ticket.

"For now, scala collection views are not usable at all, see errors
below. It is very bad thing, since there are NO WARNINGS about broken
feature in API docs. I thought Scala API is more or less stable,
disregarding small bugs. I believe it is wrong to leave broken
features in Scala releases."

> In any case, something that removes views I am against. They're a perfect solution when you need them.

Nobody is suggesting removing them from the world.

On Sat, Dec 17, 2011 at 6:33 AM, martin odersky wrote:
> But when you need them, you really do need them.

How often is it really that you can't get the same effect - more
predictably, no runtime errors, no 10^7 factor performance regressions
- with an iterator? I realize you can construct an example, that's not
the question. I think features should have to earn their keep far
more effectively than this one does.

On Sat, Dec 17, 2011 at 6:33 AM, martin odersky wrote:
> I agree. Reasons to keep views:

Those are all reasons to keep them in the world. They aren't reasons
to keep them in the core scala library.

Reasons to remove them: they're huge, they're dangerous, and they're
barely used (with good reason, which I can elaborate in more detail if
it's really necessary but I don't think it should be necessary to
belabor it.) And nobody is going to fix them. I feel very safe in
saying that, because I'm not: they're too complicated, I don't like
the design, I don't feel it can be made robust, and I don't want to
sink any more time into them.

I say break them out, put them in a dedicated jar that can optionally be left out of a production app if the feature isn't used.
Aside from specific concerns about views, modularisation is a worthy goal.  From what Paul is saying, it sounds like this is low-hanging fruit for purposes of getting more modularity, so it's a good place to start.
Do I use views? Yes, and I'm happy to make sure the required jar is on my classpath when I do.
Do/Would I use iterators and collection.breakOut instead to give equivalent behaviour?  Absolutely! I'm seriously looking into Scala for Android dev, so *anything* that pulls down the  library size and makes the oh-so-vital proguard step shorter is a couple of orders of magnitude more useful to me than views at this stage.
I'm willing to bet that people interested in Android dev is a *much* larger group right now than people who couldn't live without views, so I'll take Josh's -1000 and counter with a +2000.
Just so long as they remain an option and not yanked entirely.


On 17 December 2011 14:43, Paul Phillips <paulp@improving.org> wrote:

On Sat, Dec 17, 2011 at 6:33 AM, martin odersky <martin.odersky@epfl.ch> wrote:
> But when you need them, you really do need them.

How often is it really that you can't get the same effect - more
predictably, no runtime errors, no 10^7 factor performance regressions
- with an iterator? I realize you can construct an example, that's not
the question.  I think features should have to earn their keep far
more effectively than this one does.

--
Kevin Wright
mail: kevin.wright@scalatechnology.com
gtalk / msn : kev.lee.wright@gmail.com quora: http://www.quora.com/Kevin-Wrightgoogle+: http://gplus.to/thecoda
kev.lee.wright@gmail.com twitter: @thecoda
vibe / skype: kev.lee.wrightsteam: kev_lee_wright
"My point today is that, if we wish to count lines of code, we should not regard them as "lines produced" but as "lines spent": the current conventional wisdom is so foolish as to book that count on the wrong side of the ledger" ~ Dijkstra

Hey lurkers, you think this is irrelevant to you if you don't use
views, think again. Drink in the trait signature below and, if you
have the courage, follow the thread of interlocking type parameters
and bounds as far as it goes. Then try to imagine what kind of tax
that places on changing anything in the collections. Oh, you don't
want to change anything in the collections? Can I close all the
tickets then?

trait ParSeqViewLike[+T,
+Coll <: Parallel,
+CollSeq,
+This <: ParSeqView[T, Coll, CollSeq] with
ParSeqViewLike[T, Coll, CollSeq, This, ThisSeq],
+ThisSeq <: SeqView[T, CollSeq] with
SeqViewLike[T, CollSeq, ThisSeq]]
extends GenSeqView[T, Coll]
with GenSeqViewLike[T, Coll, This]
with ParIterableView[T, Coll, CollSeq]
with ParIterableViewLike[T, Coll, CollSeq, This, ThisSeq]
with ParSeq[T]
with ParSeqLike[T, This, ThisSeq]
{
self =>
import tasksupport._

trait Transformed[+S] extends ParSeqView[S, Coll, CollSeq]
with super[ParIterableView].Transformed[S] with
super[GenSeqViewLike].Transformed[S] {
override def splitter: SeqSplitter[S]
override def iterator = splitter
override def size = length
}

trait Sliced extends super[GenSeqViewLike].Sliced with
super[ParIterableViewLike].Sliced with Transformed[T] {
// override def slice(from1: Int, until1: Int): This =
newSliced(from1 max 0, until1 max 0).asInstanceOf[This]
override def splitter = self.splitter.psplit(from, until - from)(1)
override def seq = self.seq.slice(from, until)
}

trait Mapped[S] extends super[GenSeqViewLike].Mapped[S] with
super[ParIterableViewLike].Mapped[S] with Transformed[S] {
override def splitter = self.splitter.map(mapping)
override def seq = self.seq.map(mapping).asInstanceOf[SeqView[S, CollSeq]]
}

trait Appended[U >: T] extends super[GenSeqViewLike].Appended[U]
with super[ParIterableViewLike].Appended[U] with Transformed[U] {
override def restPar: ParSeq[U] = rest.asParSeq
override def splitter = self.splitter.appendParSeq[U,
SeqSplitter[U]](restPar.splitter)
override def seq = self.seq.++(rest).asInstanceOf[SeqView[U, CollSeq]]
}

trait Forced[S] extends super[GenSeqViewLike].Forced[S] with
super[ParIterableViewLike].Forced[S] with Transformed[S] {
override def forcedPar: ParSeq[S] = forced.asParSeq
override def splitter: SeqSplitter[S] = forcedPar.splitter
override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S, CollSeq]]
}

trait Zipped[S] extends super[GenSeqViewLike].Zipped[S] with
super[ParIterableViewLike].Zipped[S] with Transformed[(T, S)] {
override def splitter = self.splitter zipParSeq otherPar.splitter
override def seq = (self.seq zip other).asInstanceOf[SeqView[(T,
S), CollSeq]]
}

trait ZippedAll[U >: T, S] extends
super[GenSeqViewLike].ZippedAll[U, S] with
super[ParIterableViewLike].ZippedAll[U, S] with Transformed[(U, S)] {
override def splitter: SeqSplitter[(U, S)] =
self.splitter.zipAllParSeq(otherPar.splitter, thisElem, thatElem)
override def seq = (self.seq.zipAll(other, thisElem,
thatElem)).asInstanceOf[SeqView[(U, S), CollSeq]]
}

trait Reversed extends super.Reversed with Transformed[T] {
override def splitter: SeqSplitter[T] = self.splitter.reverse
override def seq = self.seq.reverse.asInstanceOf[SeqView[T, CollSeq]]
}

// use only with ParSeq patches, otherwise force
trait Patched[U >: T] extends super.Patched[U] with Transformed[U] {
def patchPar: ParSeq[U] = patch.asInstanceOf[ParSeq[U]]
override def splitter: SeqSplitter[U] =
self.splitter.patchParSeq[U](from, patchPar.splitter, replaced)
override def seq = self.seq.patch(from, patch,
replaced).asInstanceOf[SeqView[U, CollSeq]]
}

// !!!
//
// What is up with this trait and method, why are they here doing
// nothing but throwing exceptions, without even being deprecated?
// They're not implementing something abstract; why aren't they
// just removed?
//
// use Patched instead
trait Prepended[U >: T] extends super.Prepended[U] with Transformed[U] {
unsupported
}
protected def newPrepended[U >: T](elem: U): Transformed[U] = unsupported

/* wrapper virtual ctors */

protected override def newSliced(_endpoints: SliceInterval):
Transformed[T] = new { val endpoints = _endpoints } with Sliced
protected override def newAppended[U >: T](that: GenIterable[U]):
Transformed[U] = {
// we only append if `that` is a parallel sequence, i.e. it has a
precise splitter
if (that.isParSeq) new Appended[U] { val rest = that }
else newForced(mutable.ParArray.fromTraversables(this, that))
}
protected override def newForced[S](xs: => GenSeq[S]): Transformed[S] = {
if (xs.isParSeq) new Forced[S] { val forced = xs }
else new Forced[S] { val forced = mutable.ParArray.fromTraversables(xs) }
}
protected override def newMapped[S](f: T => S): Transformed[S] = new
Mapped[S] { val mapping = f }
protected override def newZipped[S](that: GenIterable[S]):
Transformed[(T, S)] = new Zipped[S] { val other = that }
protected override def newZippedAll[U >: T, S](that: GenIterable[S],
_thisElem: U, _thatElem: S): Transformed[(U, S)] = new ZippedAll[U, S]
{
val other = that
val thisElem = _thisElem
val thatElem = _thatElem
}
protected def newReversed: Transformed[T] = new Reversed { }
protected def newPatched[U >: T](_from: Int, _patch: GenSeq[U],
_replaced: Int): Transformed[U] = new {
val from = _from;
val patch = _patch;
val replaced = _replaced
} with Patched[U]

/* operation overrides */

/* sliced */
override def slice(from: Int, until: Int): This =
newSliced(SliceInterval(from, until)).asInstanceOf[This]
override def take(n: Int): This = newSliced(SliceInterval(0,
n)).asInstanceOf[This]
override def drop(n: Int): This = newSliced(SliceInterval(n,
length)).asInstanceOf[This]
override def splitAt(n: Int): (This, This) = (take(n), drop(n))

/* appended */
override def ++[U >: T, That](xs: GenTraversableOnce[U])(implicit
bf: CanBuildFrom[This, U, That]): That =
newAppended(xs.toTraversable).asInstanceOf[That]
override def :+[U >: T, That](elem: U)(implicit bf:
CanBuildFrom[This, U, That]): That = ++(Iterator.single(elem))(bf)
//override def union[U >: T, That](that: GenSeq[U])(implicit bf:
CanBuildFrom[This, U, That]): That = this ++ that

/* misc */
override def map[S, That](f: T => S)(implicit bf: CanBuildFrom[This,
S, That]): That = newMapped(f).asInstanceOf[That]
override def zip[U >: T, S, That](that: GenIterable[S])(implicit bf:
CanBuildFrom[This, (U, S), That]): That =
newZippedTryParSeq(that).asInstanceOf[That]
override def zipWithIndex[U >: T, That](implicit bf:
CanBuildFrom[This, (U, Int), That]): That =
newZipped(ParRange(0, splitter.remaining, 1, false)).asInstanceOf[That]
override def reverse: This = newReversed.asInstanceOf[This]
override def reverseMap[S, That](f: T => S)(implicit bf:
CanBuildFrom[This, S, That]): That = reverse.map(f)

/* patched */
override def updated[U >: T, That](index: Int, elem: U)(implicit bf:
CanBuildFrom[This, U, That]): That = {
require(0 <= index && index < length)
patch(index, List(elem), 1)(bf)
}
override def padTo[U >: T, That](len: Int, elem: U)(implicit bf:
CanBuildFrom[This, U, That]): That = patch(length, Seq.fill(len -
length)(elem), 0)
override def +:[U >: T, That](elem: U)(implicit bf:
CanBuildFrom[This, U, That]): That = patch(0,
mutable.ParArray.fromTraversables(Iterator.single(elem)), 0)
override def patch[U >: T, That](from: Int, patch: GenSeq[U],
replace: Int)(implicit bf: CanBuildFrom[This, U, That]): That =
newPatched(from, patch, replace).asInstanceOf[That]

/* forced */
// override def diff[U >: T](that: GenSeq[U]): This =
newForced(thisParSeq diff that).asInstanceOf[This]
// override def intersect[U >: T](that: GenSeq[U]): This =
newForced(thisParSeq intersect that).asInstanceOf[This]
// override def sorted[U >: T](implicit ord: Ordering[U]): This =
newForced(thisParSeq sorted ord).asInstanceOf[This]
override def collect[S, That](pf: PartialFunction[T, S])(implicit
bf: CanBuildFrom[This, S, That]): That =
filter(pf.isDefinedAt).map(pf)(bf)
override def scanLeft[S, That](z: S)(op: (S, T) => S)(implicit bf:
CanBuildFrom[This, S, That]): That =
newForced(thisParSeq.scanLeft(z)(op)).asInstanceOf[That]
override def scanRight[S, That](z: S)(op: (T, S) => S)(implicit bf:
CanBuildFrom[This, S, That]): That =
newForced(thisParSeq.scanRight(z)(op)).asInstanceOf[That]
override def groupBy[K](f: T => K): immutable.ParMap[K, This] =
thisParSeq.groupBy(f).map(kv => (kv._1,
newForced(kv._2).asInstanceOf[This]))
override def force[U >: T, That](implicit bf: CanBuildFrom[Coll, U,
That]) = bf ifParallel { pbf =>
executeAndWaitResult(new Force(pbf,
splitter).mapResult(_.result).asInstanceOf[Task[That, _]])
} otherwise {
val b = bf(underlying)
b ++= this.iterator
b.result
}

/* tasks */

protected[this] class Force[U >: T, That](cbf: CanCombineFrom[Coll,
U, That], protected[this] val pit: SeqSplitter[T])
extends Transformer[Combiner[U, That], Force[U, That]] {
var result: Combiner[U, That] = null
def leaf(prev: Option[Combiner[U, That]]) = result =
pit.copy2builder[U, That, Combiner[U, That]](reuse(prev,
cbf(self.underlying)))
protected[this] def newSubtask(p: SuperParIterator) = new
Force(cbf, down(p))
override def merge(that: Force[U, That]) = result = result combine
that.result
}
}

I used views, and their 2.7 version, projections, all the time in my
code. In practically every for comprehension I wrote, I would begin
with a view, and then take the final result and transform it into the
collection I was interested in to avoid creating any intermediate
collections. Code would often look like (for(a <- c.view; ....) yield
v).ToArray(), or Map.empty ++ (for(a <- c.view; ...) yield v). That
said, it seems like most of my use cases could also have been
satisfied by using an iterator. The point of views is to get non-
strict behavior on operations, but if you consume a view multiple
times,you end up performing the calculations in those operations
redundantly, which is usually not what you want anyways.

What use cases do exist right now that are satisfiable by views but
not by iterators? Is it possible to have a simpler version of views
that support less operations? In my mind, views for the most part
should just be Iterables, regardless of their originating collection.

On Dec 17, 6:37 am, Josh Suereth wrote:
> -1000 to removing views.
>
> Cleaning them up and reducing overhead is great, but views are crucial to
> the collections library.  My guess is that people don't know them well
> enough to use them or are suffering from view's poor zip performance ...
>
> In any case, something that removes views I am against.  They're a perfect
> solution when you need them.
> On Dec 16, 2011 11:55 PM, "Paul Phillips" wrote:
>
>
>
>
>
>
>
> > Speaking of things not pulling their weight: views.
>
> > Library jar before and after ripping out views.
>
> >    % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
> >    -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
> > /scratch/trunk2/build/pack/lib/scala-library.jar
> >    % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
> >    -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
> > /scratch/trunk3/build/pack/lib/scala-library.jar
>
> > Statistics from the scala code in my general ~/src directory.
>
> >    % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
> >      904766
> >    % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
> > -v ':(import|package)' | grep -v deprecated |wc -l
> >          92
>
> > One view usage per 10,000 lines of scala code. (And that may be
> > generous, as some of those 92 hits are either copies of compiler/library
> > source or things testing the compiler/library source. However I could
> > have missed some dotless uses so we'll call it a wash.)
>
> > So a feature used in .01% of lines of scala written is consuming 9% of
> > the jar. I focus on 9% not because I'm obsessed with jar size but as a
> > proxy for all the overhead associated with any large body of code. And
> > views have not been small in the overhead department, and they're still
> > not close to being a robust unit.
>
> >https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&re...
>
> >  Displaying issues 1 to 49 of 49 matching issues. [...]
>
> > Solution: exile to separate jar, deprecate for 2.10, and retire
> > them afterward. They can easily be pimped in by any who wish to use
> > them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
> > make things very slightly easier on the .01%. (I'm not even sure it
> > makes things slightly easier.) And maybe many view designs will flourish
> > in this brave new viewless world.
>
> > If we were determined not to let those view methods go away, we could
> > stub them as empty methods in the core library and have some way to
> > configure them to invoke a library elsewhere.  Anything to get them out of
> > the
> > main collections inheritance hierarchy.
>
> > *** FOOTNOTES ***
>
> > Enclosed for any who may wish to validate my experimental results.
>
> > [1] the list of projects containing those 900,000 lines of scala
>
> > Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
> > InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
> > a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
> > argot ashlar asm bash-completion bash-completion-lib blueeyes
> > blueprint borachio brepl caliper categories cccp chromium-tabs
> > collection-tests config-typesafe coreen d3 derivative-combinators
> > dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
> > extractors factorie files.txt finagle findbugs git-prompt git-pulls
> > git-sbt-processor git-scripts gitblit github-gem github-services
> > github-terminal gitignore gll-combinators gmailr google-code-prettify
> > google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
> > guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
> > izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
> > jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
> > kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
> > magpie maven-sbt maven-scala-plugin meow metascala minikanren
> > mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
> > ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
> > parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
> > projectplus pythagoras quickcursor rapture-io recursivity-commons
> > replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
> > scala-cel scala-collection-test scala-datetime scala-graph scala-ide
> > scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
> > scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
> > scala-time scalabot scalacheck scalaconsole scalacs scalafs
> > scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
> > scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
> > scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
> > sonatype-aether source.tmbundle spakle spark spde specs specs2
> > specs2-test sperformance standard-project stopwatch talking-puffin texto
> > threeten treetop twitter-util uniscala unplanned up vibrantinklion
> > view-files.txt vscaladoc windmill yarrgs yeti zmpp2
>
> > [2] the view hits.
>
> > anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
> > val b = VectorCase.newBuilder[B] ++= g.view.take(index)
> > anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
> >   val els = depthFirst.view map { case (p,v) =>
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/Canoni calNameAST.scala:430:
> >                val
> > condition =     matchCondition(matchCase.view, names)
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/Classi fyAST.scala:427:
> >             matchCondition(target,
> > targetType, caseAST.view),
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST .scala:169:
> >                  matchCondition(it.view),
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedR awAST.scala:140:
> >             matchCondition(m.view,
> > types),
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/Cojen Walker.scala:299:
> >                    it.view.values.foreach(
> > arg => walkOps(ctx, arg) )
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/Cojen Walker.scala:303:
> >                            it.view.checkMethod,
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/Cojen Walker.scala:305:
> >                            (it.view.valueTypes.map(t
> > => T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
> > ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/Cojen Walker.scala:325:
> >                    val
> > bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
> > case (tmpCtx, (expVar, idx)) =>
>
> > blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache .scala:25:
> >     TemporalCacheState(removed.view.map(t => (t._1,
> > t._2.value)).toMap, TemporalCache(kept.toMap))
> > factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
> > override def toString = getClass.getName+"("+"len="+length+"
> > 1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
> > factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
> > remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
> > seq.view(lastElemZipIndex(s)+1, seq.length)
> > mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <-
> > _1.view) {
> > mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:
> >  x.view
>
> > mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala :310:
> >       candsAndLabels.view.unzip.zipped foreach emitBranch
> > ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
> >      val indices = for ((param, idx) <- params.view.zipWithIndex
> > ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
> >  val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
> > ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
> > override def view = self.view
> > ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
> > override def view(from: Int, until: Int) = self.view(from, until)
> > ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
> >   override def seq = self.seq.view
> > ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
> >   override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
> > CollSeq]]
> > ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
> > override def seq = self.seq.view
> > ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
> >  override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
> > CollSeq]]
> > ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
> > toCollection(rest).view drop replaced
> > ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
> > toCollection(rest).view.tail
> > ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
> > def view = self.view
> > ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
> > def view(from: Int, until: Int) = self.view(from, until)
> > ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
> > override def view = self.view
> > ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
> > override def view(from: Int, until: Int): TraversableView[A, Repr] =
> > self.view(from, until)
> > ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
> > ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
>
> > scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableS pec.scala:531:
> >   def result() = { builder.result.view }
>
> > scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableS pec.scala:543:
> >   def result() = { builder.result.view(min, max) }
> > scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
> >     nodes.toList.view map nodeDegree filter degreeFilter min
> > scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
> >     nodes.toList.view map nodeDegree filter degreeFilter max
> > scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
> >   val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
> > scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
>
> ...
>
> read more »

I'm for splitting them out of the core library. I tried to use them
many times, they seldom did what I expected. I'm now most of the times
using Iterators or another custom solution.

On Sat, Dec 17, 2011 at 3:43 PM, Paul Phillips wrote:
> On Sat, Dec 17, 2011 at 6:33 AM, martin odersky wrote:
>> But when you need them, you really do need them.
>
> How often is it really that you can't get the same effect - more
> predictably, no runtime errors, no 10^7 factor performance regressions
> - with an iterator? I realize you can construct an example, that's not
> the question.  I think features should have to earn their keep far
> more effectively than this one does.

On Sat, Dec 17, 2011 at 4:24 PM, Kevin Wright wrote:
> I say break them out, put them in a dedicated jar that can optionally be
> left out of a production app if the feature isn't used.
>
> Aside from specific concerns about views, modularisation is a worthy goal.
>  From what Paul is saying, it sounds like this is low-hanging fruit for
> purposes of getting more modularity, so it's a good place to start.
>
> Do I use views? Yes, and I'm happy to make sure the required jar is on my
> classpath when I do.
>
> Do/Would I use iterators and collection.breakOut instead to give equivalent
> behaviour?  Absolutely! I'm seriously looking into Scala for Android dev, so
> *anything* that pulls down the  library size and makes the oh-so-vital
> proguard step shorter is a couple of orders of magnitude more useful to me
> than views at this stage.
>
> I'm willing to bet that people interested in Android dev is a *much* larger
> group right now than people who couldn't live without views, so I'll take
> Josh's -1000 and counter with a +2000.
>
I do not think that any Android app ships with the standard library,
with or without views present. So I do not think this is an argument
for or against having views.

Cheers

On Sat, Dec 17, 2011 at 8:31 AM, martin odersky wrote:
> I do not think that any Android app ships with the standard library,
> with or without views present. So I do not think this is an argument
> for or against having views.

It's impossible to ship with the standard library because it's so big
(or so I hear.) This is like having a cage full of lions, and when
someone wants to free the lions, saying that you don't see any other
lions roaming free, so why should these lions get special treatment?
There are no lions roaming free because they're all in the cage!

On Sat, Dec 17, 2011 at 5:31 PM, martin odersky wrote:
> I do not think that any Android app ships with the standard library,
> with or without views present. So I do not think this is an argument
> for or against having views.

No, but proguard isn't able to remove view classes in many cases even
if they are not used any where in user code. And they are the current
road block for using the Scala standard library with Android.

On Sat, Dec 17, 2011 at 3:36 PM, Paul Phillips wrote:
> On Sat, Dec 17, 2011 at 4:37 AM, Josh Suereth wrote:
>> Cleaning them up and reducing overhead is great, but views are crucial to
>> the collections library.  My guess is that people don't know them well
>> enough to use them or are suffering from view's poor zip performance ...
>
> Maybe they just prefer not to roll the dice on runtime errors:
>
>  https://issues.scala-lang.org/browse/SI-4332
>
> I agree with the last comment in that ticket.
>
> "For now, scala collection views are not usable at all, see errors
> below. It is very bad thing, since there are NO WARNINGS about broken
> feature in API docs. I thought Scala API is more or less stable,
> disregarding small bugs. I believe it is wrong to leave broken
> features in Scala releases."
>
>> In any case, something that removes views I am against.  They're a perfect solution when you need them.
>
> Nobody is suggesting removing them from the world.

I agree views are a huge pain to implement correctly, but that does
not invalidate the fact that they are essential in some scenarios.

Here's the fundamental problem with views: To do them correctly you
need to re-implement a lot of classes in every layer of your
inheritance library. Each class captures the behavior of one common
collection operation and possible a couple of variants. We used to
have 4 such layers: Traversable, Iterable, Seq, and Linear/IndexedSeq.
Views were a pain to implement then, but it was sort of manageable. We
now have almost twice the number of layers because of the Gen...
types. This seemed to have pushed views over the edge where no person
can understand the code anymore and bugs like SI-4332 creep in and are
almost impossible to fix (I am pretty sure that you could take a slice
of a slice before 2.9, the whole point of views is that this should be
effcient!)

So, here's an alternate proposal to get back some sanity: Let's
completely separate the implementations of sequential and parallel
collections, including their views. This means we still keep GenSeq,
GenIterable as common supertypes, but these would be pure interfaces
without any implementation. And traits GenIterableLike, GenSeqLike,
etc would be removed.

This means that both parallel and sequential collections have to
implement their own view hierarchies, but at least there's no
interference between them. And we collapse the number of levels from 7
to 4 or maybe 3 (if Iterable and Traversable get merged). Hopefully
this makes views manageable again. And it would have other benefits
such as reducing the number of supertraits of common types such as
List.

Cheers

On Sat, Dec 17, 2011 at 8:53 AM, martin odersky wrote:
> So, here's an alternate proposal to get back some sanity: Let's
> completely separate the implementations of sequential and parallel
> collections, including their views. This means we still keep GenSeq,
> GenIterable as common supertypes, but these would be pure interfaces
> without any implementation. And traits GenIterableLike, GenSeqLike,
> etc would be removed.

I will travel backward in time eight months to endorse this idea.

...

On 4/17/11 12:47 PM, martin odersky wrote:
> Agree in principle, but wondering how to do better?

That's one of those tough open ended questions, because it seems
unlikely I have any ideas which are going to be news to you. Broadly
speaking, looser coupling and a cleaner separation of implementation and
interface. I feel like traits have been something of a trap in that
they have reduced the motivation to cleanly define interfaces, and they
have made it too easy to reuse code (and actually, too hard not to) so
code ends up being reused in places where the implementation for that
purpose is questionable at best; and in other cases the "reused" code
becomes a hostile entity which one must defend against by being sure to
override the existing implementation.

References: various issues which have arisen with mutable vs. immutable,
linear vs. indexed access, parallel vs. sequential traversal, lazy vs.
eager evaluation, by-name vs. by-value calls, view vs. non-view monadic
operations.

I think mixin composition where resulting behavior depends on the
linearization is something which can only be gotten right in small
doses, and that we are way past the maximum dose in the collections.

I have lots of ideas, but assuming we are only in the market for
short-term plausibly viable approaches with the parallel collections, I
think we should try to completely divorce them in terms of inheritance
and provide one or more mechanisms for hopping back and forth
(conversions and/or "views" of one as the other.) They can reuse some
pure interfaces.

I wish we had language support for pure interfaces, by the way. There's
no way to write a trait and require that it provide no implementations.
We could make "abstract trait" mean this.

On 17/12/2011 16:47, Johannes Rudolph wrote:
> I'm for splitting them out of the core library. I tried to use them
> many times, they seldom did what I expected. I'm now most of the times
> using Iterators or another custom solution.

Same for me. When I read about view for the first time, I was thrilled
by the feature, and was wondering why it was not the default behaviour
for all collection iteration. I did try to use them, sometimes getting
completlly unexpected results, most of the time hitting a bug along the
road. After some time, I realize that they weren't the free lunch I was
hopping for and I found that it was far safer to just ignore them, and
build custom solution when the need happened.

On the other hand, I'm all for a modularized and/or lighter Scala-lib,
so if my vote count, I'm for removing them.
Ok, well, either that, or make what is needed to have them be really
useful (useful as in "in Scala, the default pattern for fast iteration
is "for { elt <- mycollection.view ... }, that a no brainer")

Cheers,

On 17/12/2011 05:55, Paul Phillips wrote:
> Speaking of things not pulling their weight: views.
>
> [...]
> % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
> -v ':(import|package)' | grep -v deprecated |wc -l
> 92

How many of these 92 view usages can't be work-arounded with a
iterator-like solution ? That would help to see to what extends view
usage are mandatory. If not even only one of all the view usages in the
wild are not (better) coded with something else, that will be a serious
wondering of their value/cost ratio.

Thanks,

On Sat, Dec 17, 2011 at 9:23 AM, Francois wrote:
> How many of these 92 view usages can't be work-arounded with a iterator-like
> solution ? That would help to see to what extends view usage are mandatory.
> If not even only one of all the view usages in the wild are not (better)
> coded with something else, that will be a serious wondering of their
> value/cost ratio.

92 was a little much to verify individually. Instead I enumerated all
the uses of views in the compiler and library. As the perpetrators of
views, one could reasonably expect us to have at least one compelling
use.

src/compiler/scala/tools/nsc/interactive/Global.scala:420:
source.content.view.drop(start).take(length).mkString+" :
"+source.path+" ("+start+", "+end+

source.content.slice(start, start + length)

src/compiler/scala/tools/nsc/interactive/tests/Tester.scala:161:
changes.view.reverse foreach (_.insertAll())

changes reverseMap (_.insertAll())

src/compiler/scala/tools/nsc/symtab/classfile/ClassfileParser.scala:395:
bytesBuffer ++= in.buf.view(start + 3, start + 3 + len)

bytesBuffer ++= in.buf.iterator.slice(start + 3, start + 3 + len)

src/library/scala/collection/IterableProxyLike.scala:99: override def
view = self.view
src/library/scala/collection/IterableProxyLike.scala:100: override
def view(from: Int, until: Int) = self.view(from, until)
src/library/scala/collection/parallel/ParIterableLike.scala:791:
override def seq = self.seq.view
src/library/scala/collection/parallel/ParIterableViewLike.scala:88:
override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
CollSeq]]
src/library/scala/collection/parallel/ParSeqLike.scala:350:
override def seq = self.seq.view
src/library/scala/collection/parallel/ParSeqViewLike.scala:70:
override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
CollSeq]]

Baton-passing calls.

src/library/scala/collection/SeqLike.scala:525: b ++=
toCollection(rest).view drop replaced

b ++= (toCollection(rest).iterator drop replaced)

src/library/scala/collection/SeqLike.scala:538: b ++=
toCollection(rest).view.tail

b ++= toCollection(rest).iterator.tail

src/library/scala/collection/SeqLike.scala:780: if
(S.view.slice(m0, m1) == W.view.slice(n0, n1)) m0

if (S.iterator.slice(m0, m1) == W.iterator.slice(n0, n1))

src/library/scala/collection/SeqProxyLike.scala:68: override def view
= self.view
src/library/scala/collection/SeqProxyLike.scala:69: override def
view(from: Int, until: Int) = self.view(from, until)

More batons.

src/library/scala/xml/factory/NodeFactory.scala:32:
ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }

(ch1 corresponds ch2)(_ eq _)

src/compiler/scala/tools/nsc/doc/html/page/ReferenceIndex.scala:55:
entry(groups._1, groups._2.view)

Not sure, this one might be legit.

So, somewhere in the neighborhood of zero.

On Sat, Dec 17, 2011 at 6:07 PM, Paul Phillips wrote:
> So, somewhere in the neighborhood of zero.

Does dropping views in favour of iterators in the compiler have any
effect on compilation performance?

Cheers,

Miles

On Sat, Dec 17, 2011 at 10:15 AM, Miles Sabin wrote:
> Does dropping views in favour of iterators in the compiler have any
> effect on compilation performance?

My guess is yes, for the better, especially if we properly account for
the occasional five order of magnitude slowdown in views. But I don't
know, one can only gather so much evidence and I feel like I've
already shown enough to put the guy in the chair while simultaneously
being lethally injected and hanged.

https://github.com/scala/scala/commit/be49752855

A patch for views. Most relevant change:

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.

Amazing. Some people have some incredible insight, and even the ability
to articulate it.

Thanks Paul.

On 2011-12-16 8:55 PM, Paul Phillips wrote:
> Speaking of things not pulling their weight: views.
>
> Library jar before and after ripping out views.
>
> % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
> -rw-r--r-- 1 paulp admin 7161826 Dec 16 19:53
> /scratch/trunk2/build/pack/lib/scala-library.jar
> % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
> -rw-r--r-- 1 paulp admin 6561034 Dec 16 19:51
> /scratch/trunk3/build/pack/lib/scala-library.jar
>
> Statistics from the scala code in my general ~/src directory.
>
> % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
> 904766
> % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
> -v ':(import|package)' | grep -v deprecated |wc -l
> 92
>
> One view usage per 10,000 lines of scala code. (And that may be
> generous, as some of those 92 hits are either copies of compiler/library
> source or things testing the compiler/library source. However I could
> have missed some dotless uses so we'll call it a wash.)
>
> So a feature used in .01% of lines of scala written is consuming 9% of
> the jar. I focus on 9% not because I'm obsessed with jar size but as a
> proxy for all the overhead associated with any large body of code. And
> views have not been small in the overhead department, and they're still
> not close to being a robust unit.
>
> https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&reque...
>
> Displaying issues 1 to 49 of 49 matching issues. [...]
>
> Solution: exile to separate jar, deprecate for 2.10, and retire
> them afterward. They can easily be pimped in by any who wish to use
> them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
> make things very slightly easier on the .01%. (I'm not even sure it
> makes things slightly easier.) And maybe many view designs will flourish
> in this brave new viewless world.
>
> If we were determined not to let those view methods go away, we could
> stub them as empty methods in the core library and have some way to
> configure them to invoke a library elsewhere. Anything to get them out of the
> main collections inheritance hierarchy.
>
>
> *** FOOTNOTES ***
>
> Enclosed for any who may wish to validate my experimental results.
>
> [1] the list of projects containing those 900,000 lines of scala
>
> Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
> InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
> a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
> argot ashlar asm bash-completion bash-completion-lib blueeyes
> blueprint borachio brepl caliper categories cccp chromium-tabs
> collection-tests config-typesafe coreen d3 derivative-combinators
> dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
> extractors factorie files.txt finagle findbugs git-prompt git-pulls
> git-sbt-processor git-scripts gitblit github-gem github-services
> github-terminal gitignore gll-combinators gmailr google-code-prettify
> google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
> guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
> izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
> jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
> kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
> magpie maven-sbt maven-scala-plugin meow metascala minikanren
> mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
> ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
> parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
> projectplus pythagoras quickcursor rapture-io recursivity-commons
> replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
> scala-cel scala-collection-test scala-datetime scala-graph scala-ide
> scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
> scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
> scala-time scalabot scalacheck scalaconsole scalacs scalafs
> scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
> scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
> scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
> sonatype-aether source.tmbundle spakle spark spde specs specs2
> specs2-test sperformance standard-project stopwatch talking-puffin texto
> threeten treetop twitter-util uniscala unplanned up vibrantinklion
> view-files.txt vscaladoc windmill yarrgs yeti zmpp2
>
>
> [2] the view hits.
>
>
>
> anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
> val b = VectorCase.newBuilder[B] ++= g.view.take(index)
> anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
> val els = depthFirst.view map { case (p,v) =>
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430: val
> condition = matchCondition(matchCase.view, names)
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427: matchCondition(target,
> targetType, caseAST.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169: matchCondition(it.view),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140: matchCondition(m.view,
> types),
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299: it.view.values.foreach(
> arg => walkOps(ctx, arg) )
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303: it.view.checkMethod,
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305: (it.view.valueTypes.map(t
> => T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
> ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325: val
> bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
> case (tmpCtx, (expVar, idx)) =>
> blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
> TemporalCacheState(removed.view.map(t => (t._1,
> t._2.value)).toMap, TemporalCache(kept.toMap))
> factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
> override def toString = getClass.getName+"("+"len="+length+"
> 1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
> factorie/src/test/scala/cc/factorie/TestUtils.scala:32: def
> remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
> seq.view(lastElemZipIndex(s)+1, seq.length)
> mada/src/main/scala/sequence/vector/Scala.scala:17: for (e<- _1.view) {
> mada/src/test/scala/sequence/iterative/ConversionTest.scala:53: x.view
> mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
> candsAndLabels.view.unzip.zipped foreach emitBranch
> ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
> val indices = for ((param, idx)<- params.view.zipWithIndex
> ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28: *
> val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
> override def view = self.view
> ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
> override def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
> override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
> override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
> override def seq = self.seq.view
> ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
> override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
> CollSeq]]
> ozma/src/scalalib/scala/collection/SeqLike.scala:518: b ++=
> toCollection(rest).view drop replaced
> ozma/src/scalalib/scala/collection/SeqLike.scala:531: b ++=
> toCollection(rest).view.tail
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69: override
> def view = self.view
> ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70: override
> def view(from: Int, until: Int) = self.view(from, until)
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
> override def view = self.view
> ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
> override def view(from: Int, until: Int): TraversableView[A, Repr] =
> self.view(from, until)
> ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
> ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
> def result() = { builder.result.view }
> scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
> def result() = { builder.result.view(min, max) }
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
> nodes.toList.view map nodeDegree filter degreeFilter min
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
> nodes.toList.view map nodeDegree filter degreeFilter max
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
> val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
> scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
> else nodes.view map nodeDegree filter
> degreeFilter)
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
> (for (jsonNode<- nodeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
> (for (jsonEdge<- edgeList.view) yield {
> scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
> (for (jsonEdge<- edgeList.view) yield {
> scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
> val toRead = data.view.slice(position.toInt, min(position.toInt +
> dst.limit-dst.position, data.size))
> scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
> lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
> scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
> override def view = super.view
> scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
> val list = (f(expectedData().toList.view) map { i => expectedCount
> += 1; i })
> scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42: new
> Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
> => new Grouping(Node(c))).toList)
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
> m<- getClass().getMethods.view
> scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
> m<- getClass().getMethods.view
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
> tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
> {sigBytesFromAnnotation} headOption
> scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
> splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
> headOption
> Scalala/src/main/scala/scalala/library/Library.scala:213:
> matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Library.scala:223:
> matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
> Scalala/src/main/scala/scalala/library/Plotting.scala:464://
> val values = items.view.map(item =>
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:
>
> SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
> scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
> if (f.isInstanceOf[Transformer]) f else f.andThen((i :
> Iterable[String]) => i.view);
> scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
> peeked.view(0, n);
> scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
> for( datum<- range.view map data) {
> scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
> val (grad: DenseVector[Double],prob: Double) =
> eCounts.zipWithIndex.par.view.map { case (vec,c) =>
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
> val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
> scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
> ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
> scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
> context.view(model)
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
> context.view(it.get.asInstanceOf[AnyRef])
> scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
> override def requiredViewClass(): java.lang.Class[_] =
> classOf[org.fusesource.scalate.spring.view.ScalateView]
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
> file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
> directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
> children(file).view.flatMap(andDescendants(_)))
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
> constructors.view.map(c =>
> tryCreate(c)).find(_.isDefined).getOrElse(None)
> scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
> sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
> scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
> sourceDirectories.view.map
> {findMatching(_)}.find(_.isDefined).getOrElse(None)
> scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23: if
> (len != -1) out ++= buffer.view(0, len)
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
> argumentsFragments.view.zip(nestedArguments).collect { case
> (ApplicableArguments(value), args) => (value, args) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
> def fragmentAndSpecNames: Seq[(T, SpecName)] =
> argumentsFragments.view.zip(nestedSpecNames).collect { case
> (ApplicableArguments(value), name) => (value, name) }
> specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
> argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
> { case ((ApplicableArguments(value), args), name) => (value, args,
> name) }
> specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
> specificationNames(path, pattern, basePath,
> verbose).view.filter(filter).flatMap(n => createSpecification(n))
> sperformance/src/test/scala/CollectionsShootout.scala:148:
> List.range(1, size).toList.view
> sperformance/src/test/scala/CollectionsShootout.scala:155:
> ArrayBuffer.range(1, size).view
> sperformance/src/test/scala/CollectionsShootout.scala:184:
> col.view.map(_ * 2).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:189:
> col.view.filter(_ % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:194:
> col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
> sperformance/src/test/scala/CollectionsShootout.scala:236:
> performance of "Vector.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:239:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1")&&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:244:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> sperformance/src/test/scala/CollectionsShootout.scala:260:
> performance of "ArrayBuffer.view" in {
> sperformance/src/test/scala/CollectionsShootout.scala:263:
> a.view.zipWithIndex.filter(x => x._1.startsWith("1")&&
> (x._2%3==0)).map(_._1).force
> sperformance/src/test/scala/CollectionsShootout.scala:268:
> a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
> val result = codecs.urlParamCodecs.view map { codec =>
> uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
> val result = codecs.urlParamCodecs.view map { codec =>

I really like the idea of views, but the implementation leaves so much to be desired that I'm inclined to throw them all out and perhaps at some nebulous point in the future try again.

For example, views and mutable classes create exactly the kinds of bookkeeping problems that you would expect from a fragile user-based implementation:

    import collection.mutable.ArrayBuffer
    val a = ArrayBuffer(1,2,3,4)
    val b = a.view.slice(2,4)
    b(0) = 5     // a is now (1,2,5,4) as it should be
    a -= 1
    b(0) = 6     // a is now (2,5,6)?  Did we really mean that?
    a.slice(2,4).foreach(println)      // All okay
    a.view.slice(2,4).foreach(println) // Still okay
    b.foreach(println)                 // IndexOutOfBoundsException?!

Having views in the library _could_ allow them to be very clever and help with safety and so on, but they're mostly not.  (Iterators are broken in this way, too, incidentally.  But the exposure is lower since they're generally fire-and-forget, so you don't keep them around long enough, typically, to get into this kind of problem.)

I wonder how much of the functionality of views could be captured by allowing complex sets of operations to be specified in advance, something like

  xs.magicUnicorns(
    Drop(3), Filter(_<5), Map(_.toString), Filter(_.length>2)
  )

which would apply all the operations in one pass.  The advantage of this over iterators is that you could more easily carry around chains of operations for repeated use:

  val myUnicorn = Unicorns(Drop(3), Filter(_<5), Map(_.toString))
  xs.magicUnicorns(myUnicorn).filter(_.length > 2)
  xs.magicUnicorns(myUnicorn).reduce(_ + _)

and it would _probably_ be easier to perform automatic simplifications.

Then again, this gets back into the whole internal vs. external iterators thing that came up in another context, and I'm not sure that this avoids the problems of externality.  (Then again, neither do views currently.)

  --Rex



On Fri, Dec 16, 2011 at 11:55 PM, Paul Phillips <paulp@improving.org> wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

   % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
   % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

   % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
     904766
   % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
         92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

 https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

 Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.


*** FOOTNOTES ***

Enclosed for any who may wish to validate my experimental results.

[1] the list of projects containing those 900,000 lines of scala

Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
argot ashlar asm bash-completion bash-completion-lib blueeyes
blueprint borachio brepl caliper categories cccp chromium-tabs
collection-tests config-typesafe coreen d3 derivative-combinators
dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
extractors factorie files.txt finagle findbugs git-prompt git-pulls
git-sbt-processor git-scripts gitblit github-gem github-services
github-terminal gitignore gll-combinators gmailr google-code-prettify
google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
magpie maven-sbt maven-scala-plugin meow metascala minikanren
mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
projectplus pythagoras quickcursor rapture-io recursivity-commons
replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
scala-cel scala-collection-test scala-datetime scala-graph scala-ide
scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
scala-time scalabot scalacheck scalaconsole scalacs scalafs
scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
sonatype-aether source.tmbundle spakle spark spde specs specs2
specs2-test sperformance standard-project stopwatch talking-puffin texto
threeten treetop twitter-util uniscala unplanned up vibrantinklion
view-files.txt vscaladoc windmill yarrgs yeti zmpp2


[2] the view hits.



anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
val b = VectorCase.newBuilder[B] ++= g.view.take(index)
anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
  val els = depthFirst.view map { case (p,v) =>
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430:                val
condition =     matchCondition(matchCase.view, names)
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427:             matchCondition(target,
targetType, caseAST.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169:                  matchCondition(it.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140:             matchCondition(m.view,
types),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299:                    it.view.values.foreach(
arg => walkOps(ctx, arg) )
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303:                            it.view.checkMethod,
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305:                            (it.view.valueTypes.map(t
=> T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325:                    val
bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
case (tmpCtx, (expVar, idx)) =>
blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
    TemporalCacheState(removed.view.map(t => (t._1,
t._2.value)).toMap, TemporalCache(kept.toMap))
factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
override def toString = getClass.getName+"("+"len="+length+"
1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
seq.view(lastElemZipIndex(s)+1, seq.length)
mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <- _1.view) {
mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:        x.view
mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
      candsAndLabels.view.unzip.zipped foreach emitBranch
ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
     val indices = for ((param, idx) <- params.view.zipWithIndex
ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
 val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
override def view = self.view
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
override def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
  override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
  override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
 override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
toCollection(rest).view drop replaced
ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
toCollection(rest).view.tail
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
def view = self.view
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
override def view = self.view
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
override def view(from: Int, until: Int): TraversableView[A, Repr] =
self.view(from, until)
ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
  def result() = { builder.result.view }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
  def result() = { builder.result.view(min, max) }
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
    nodes.toList.view map nodeDegree filter degreeFilter min
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
    nodes.toList.view map nodeDegree filter degreeFilter max
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
  val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
  else                          nodes.view map nodeDegree filter
degreeFilter)
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
          (for (jsonNode <- nodeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
              (for (jsonEdge <- edgeList.view) yield {
scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
    val toRead = data.view.slice(position.toInt, min(position.toInt +
dst.limit-dst.position, data.size))
scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
 lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
 override def view = super.view
scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
  val list = (f(expectedData().toList.view) map { i => expectedCount
+= 1; i })
scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42:    new
Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
=> new Grouping(Node(c))).toList)
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
m <- getClass().getMethods.view
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
m <- getClass().getMethods.view
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
 tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
{sigBytesFromAnnotation} headOption
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
 splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
headOption
Scalala/src/main/scala/scalala/library/Library.scala:213:
matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Library.scala:223:
matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Plotting.scala:464://
val values = items.view.map(item =>
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:

SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
    if (f.isInstanceOf[Transformer]) f else f.andThen((i :
Iterable[String]) => i.view);
scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
peeked.view(0, n);
scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
      for( datum <- range.view map data) {
scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
  val (grad: DenseVector[Double],prob: Double) =
eCounts.zipWithIndex.par.view.map { case (vec,c) =>
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
  val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
  ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
  context.view(model)
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
  context.view(it.get.asInstanceOf[AnyRef])
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
 override def requiredViewClass(): java.lang.Class[_] =
classOf[org.fusesource.scalate.spring.view.ScalateView]
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
    file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
  directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
  children(file).view.flatMap(andDescendants(_)))
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
  constructors.view.map(c =>
tryCreate(c)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
  sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
  sourceDirectories.view.map
{findMatching(_)}.find(_.isDefined).getOrElse(None)
scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23:                 if
(len != -1)     out ++= buffer.view(0, len)
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
argumentsFragments.view.zip(nestedArguments).collect { case
(ApplicableArguments(value), args) => (value, args) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
def fragmentAndSpecNames: Seq[(T, SpecName)] =
argumentsFragments.view.zip(nestedSpecNames).collect { case
(ApplicableArguments(value), name) => (value, name) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
{ case ((ApplicableArguments(value), args), name) => (value, args,
name) }
specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
  specificationNames(path, pattern, basePath,
verbose).view.filter(filter).flatMap(n => createSpecification(n))
sperformance/src/test/scala/CollectionsShootout.scala:148:
List.range(1, size).toList.view
sperformance/src/test/scala/CollectionsShootout.scala:155:
ArrayBuffer.range(1, size).view
sperformance/src/test/scala/CollectionsShootout.scala:184:
col.view.map(_ * 2).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:189:
col.view.filter(_ % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:194:
col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:236:
performance of "Vector.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:239:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:244:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
sperformance/src/test/scala/CollectionsShootout.scala:260:
performance of "ArrayBuffer.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:263:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:268:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
    val result = codecs.urlParamCodecs.view map { codec =>
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
    val result = codecs.urlParamCodecs.view map { codec =>

How about moving them to a jar that is distributed with scala but not part of the core like swing?

If we treat views like swing, I'm all for that.  Same with parallel collections.   However, the definition of "core scala" is where I gave my -1000.

I'd be all for moving them into a src / collection / views directory and support and release the library with core scala.

Making them a completely separate project is what concerns me.

Also, I'll happily start maintaining views at the expense of other projects if it's a "they leave standard district if no one takes them" deal.

On Dec 17, 2011 6:11 PM, "Rex Kerr" <ichoran@gmail.com> wrote:

I really like the idea of views, but the implementation leaves so much to be desired that I'm inclined to throw them all out and perhaps at some nebulous point in the future try again.

For example, views and mutable classes create exactly the kinds of bookkeeping problems that you would expect from a fragile user-based implementation:

    import collection.mutable.ArrayBuffer
    val a = ArrayBuffer(1,2,3,4)
    val b = a.view.slice(2,4)
    b(0) = 5     // a is now (1,2,5,4) as it should be
    a -= 1
    b(0) = 6     // a is now (2,5,6)?  Did we really mean that?
    a.slice(2,4).foreach(println)      // All okay
    a.view.slice(2,4).foreach(println) // Still okay
    b.foreach(println)                 // IndexOutOfBoundsException?!

Having views in the library _could_ allow them to be very clever and help with safety and so on, but they're mostly not.  (Iterators are broken in this way, too, incidentally.  But the exposure is lower since they're generally fire-and-forget, so you don't keep them around long enough, typically, to get into this kind of problem.)

I wonder how much of the functionality of views could be captured by allowing complex sets of operations to be specified in advance, something like

  xs.magicUnicorns(
    Drop(3), Filter(_<5), Map(_.toString), Filter(_.length>2)
  )

which would apply all the operations in one pass.  The advantage of this over iterators is that you could more easily carry around chains of operations for repeated use:

  val myUnicorn = Unicorns(Drop(3), Filter(_<5), Map(_.toString))
  xs.magicUnicorns(myUnicorn).filter(_.length > 2)
  xs.magicUnicorns(myUnicorn).reduce(_ + _)

and it would _probably_ be easier to perform automatic simplifications.

Then again, this gets back into the whole internal vs. external iterators thing that came up in another context, and I'm not sure that this avoids the problems of externality.  (Then again, neither do views currently.)

  --Rex



On Fri, Dec 16, 2011 at 11:55 PM, Paul Phillips <paulp@improving.org> wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

   % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
   % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

   % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
     904766
   % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
         92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

 https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

 Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.


*** FOOTNOTES ***

Enclosed for any who may wish to validate my experimental results.

[1] the list of projects containing those 900,000 lines of scala

Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
argot ashlar asm bash-completion bash-completion-lib blueeyes
blueprint borachio brepl caliper categories cccp chromium-tabs
collection-tests config-typesafe coreen d3 derivative-combinators
dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
extractors factorie files.txt finagle findbugs git-prompt git-pulls
git-sbt-processor git-scripts gitblit github-gem github-services
github-terminal gitignore gll-combinators gmailr google-code-prettify
google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
magpie maven-sbt maven-scala-plugin meow metascala minikanren
mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
projectplus pythagoras quickcursor rapture-io recursivity-commons
replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
scala-cel scala-collection-test scala-datetime scala-graph scala-ide
scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
scala-time scalabot scalacheck scalaconsole scalacs scalafs
scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
sonatype-aether source.tmbundle spakle spark spde specs specs2
specs2-test sperformance standard-project stopwatch talking-puffin texto
threeten treetop twitter-util uniscala unplanned up vibrantinklion
view-files.txt vscaladoc windmill yarrgs yeti zmpp2


[2] the view hits.



anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
val b = VectorCase.newBuilder[B] ++= g.view.take(index)
anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
  val els = depthFirst.view map { case (p,v) =>
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430:                val
condition =     matchCondition(matchCase.view, names)
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427:             matchCondition(target,
targetType, caseAST.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169:                  matchCondition(it.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140:             matchCondition(m.view,
types),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299:                    it.view.values.foreach(
arg => walkOps(ctx, arg) )
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303:                            it.view.checkMethod,
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305:                            (it.view.valueTypes.map(t
=> T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325:                    val
bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
case (tmpCtx, (expVar, idx)) =>
blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
    TemporalCacheState(removed.view.map(t => (t._1,
t._2.value)).toMap, TemporalCache(kept.toMap))
factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
override def toString = getClass.getName+"("+"len="+length+"
1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
seq.view(lastElemZipIndex(s)+1, seq.length)
mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <- _1.view) {
mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:        x.view
mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
      candsAndLabels.view.unzip.zipped foreach emitBranch
ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
     val indices = for ((param, idx) <- params.view.zipWithIndex
ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
 val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
override def view = self.view
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
override def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
  override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
  override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
 override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
toCollection(rest).view drop replaced
ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
toCollection(rest).view.tail
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
def view = self.view
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
override def view = self.view
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
override def view(from: Int, until: Int): TraversableView[A, Repr] =
self.view(from, until)
ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
  def result() = { builder.result.view }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
  def result() = { builder.result.view(min, max) }
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
    nodes.toList.view map nodeDegree filter degreeFilter min
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
    nodes.toList.view map nodeDegree filter degreeFilter max
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
  val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
  else                          nodes.view map nodeDegree filter
degreeFilter)
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
          (for (jsonNode <- nodeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
              (for (jsonEdge <- edgeList.view) yield {
scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
    val toRead = data.view.slice(position.toInt, min(position.toInt +
dst.limit-dst.position, data.size))
scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
 lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
 override def view = super.view
scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
  val list = (f(expectedData().toList.view) map { i => expectedCount
+= 1; i })
scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42:    new
Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
=> new Grouping(Node(c))).toList)
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
m <- getClass().getMethods.view
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
m <- getClass().getMethods.view
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
 tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
{sigBytesFromAnnotation} headOption
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
 splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
headOption
Scalala/src/main/scala/scalala/library/Library.scala:213:
matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Library.scala:223:
matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Plotting.scala:464://
val values = items.view.map(item =>
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:

SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
    if (f.isInstanceOf[Transformer]) f else f.andThen((i :
Iterable[String]) => i.view);
scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
peeked.view(0, n);
scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
      for( datum <- range.view map data) {
scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
  val (grad: DenseVector[Double],prob: Double) =
eCounts.zipWithIndex.par.view.map { case (vec,c) =>
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
  val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
  ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
  context.view(model)
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
  context.view(it.get.asInstanceOf[AnyRef])
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
 override def requiredViewClass(): java.lang.Class[_] =
classOf[org.fusesource.scalate.spring.view.ScalateView]
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
    file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
  directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
  children(file).view.flatMap(andDescendants(_)))
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
  constructors.view.map(c =>
tryCreate(c)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
  sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
  sourceDirectories.view.map
{findMatching(_)}.find(_.isDefined).getOrElse(None)
scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23:                 if
(len != -1)     out ++= buffer.view(0, len)
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
argumentsFragments.view.zip(nestedArguments).collect { case
(ApplicableArguments(value), args) => (value, args) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
def fragmentAndSpecNames: Seq[(T, SpecName)] =
argumentsFragments.view.zip(nestedSpecNames).collect { case
(ApplicableArguments(value), name) => (value, name) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
{ case ((ApplicableArguments(value), args), name) => (value, args,
name) }
specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
  specificationNames(path, pattern, basePath,
verbose).view.filter(filter).flatMap(n => createSpecification(n))
sperformance/src/test/scala/CollectionsShootout.scala:148:
List.range(1, size).toList.view
sperformance/src/test/scala/CollectionsShootout.scala:155:
ArrayBuffer.range(1, size).view
sperformance/src/test/scala/CollectionsShootout.scala:184:
col.view.map(_ * 2).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:189:
col.view.filter(_ % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:194:
col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:236:
performance of "Vector.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:239:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:244:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
sperformance/src/test/scala/CollectionsShootout.scala:260:
performance of "ArrayBuffer.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:263:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:268:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
    val result = codecs.urlParamCodecs.view map { codec =>
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
    val result = codecs.urlParamCodecs.view map { codec =>

Mostly I want the Traversable class and TraversableView....

On Dec 17, 2011 6:11 PM, "Rex Kerr" <ichoran@gmail.com> wrote:

I really like the idea of views, but the implementation leaves so much to be desired that I'm inclined to throw them all out and perhaps at some nebulous point in the future try again.

For example, views and mutable classes create exactly the kinds of bookkeeping problems that you would expect from a fragile user-based implementation:

    import collection.mutable.ArrayBuffer
    val a = ArrayBuffer(1,2,3,4)
    val b = a.view.slice(2,4)
    b(0) = 5     // a is now (1,2,5,4) as it should be
    a -= 1
    b(0) = 6     // a is now (2,5,6)?  Did we really mean that?
    a.slice(2,4).foreach(println)      // All okay
    a.view.slice(2,4).foreach(println) // Still okay
    b.foreach(println)                 // IndexOutOfBoundsException?!

Having views in the library _could_ allow them to be very clever and help with safety and so on, but they're mostly not.  (Iterators are broken in this way, too, incidentally.  But the exposure is lower since they're generally fire-and-forget, so you don't keep them around long enough, typically, to get into this kind of problem.)

I wonder how much of the functionality of views could be captured by allowing complex sets of operations to be specified in advance, something like

  xs.magicUnicorns(
    Drop(3), Filter(_<5), Map(_.toString), Filter(_.length>2)
  )

which would apply all the operations in one pass.  The advantage of this over iterators is that you could more easily carry around chains of operations for repeated use:

  val myUnicorn = Unicorns(Drop(3), Filter(_<5), Map(_.toString))
  xs.magicUnicorns(myUnicorn).filter(_.length > 2)
  xs.magicUnicorns(myUnicorn).reduce(_ + _)

and it would _probably_ be easier to perform automatic simplifications.

Then again, this gets back into the whole internal vs. external iterators thing that came up in another context, and I'm not sure that this avoids the problems of externality.  (Then again, neither do views currently.)

  --Rex



On Fri, Dec 16, 2011 at 11:55 PM, Paul Phillips <paulp@improving.org> wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

   % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
   % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
   -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

   % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
     904766
   % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
         92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

 https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

 Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.


*** FOOTNOTES ***

Enclosed for any who may wish to validate my experimental results.

[1] the list of projects containing those 900,000 lines of scala

Asynchronous-Functional-Programming Autoproxy-Lite Gitalist Hooks
InteractiveHelp Play20 SBT.tmbundle Scalala Scalog ShiftIt Xus
a.sh alacs amzn-cloudinit annotation-tools annovention anti-xml
argot ashlar asm bash-completion bash-completion-lib blueeyes
blueprint borachio brepl caliper categories cccp chromium-tabs
collection-tests config-typesafe coreen d3 derivative-combinators
dotfiles.repository.steve.org.uk dracula egit egit-github ensime escabot
extractors factorie files.txt finagle findbugs git-prompt git-pulls
git-sbt-processor git-scripts gitblit github-gem github-services
github-terminal gitignore gll-combinators gmailr google-code-prettify
google-diff-match-patch google-toolbox-for-mac grizzled-scala groovy
guava hammersmith hawtdispatch helix homebrew hudson ii irccat ivy
izpack jdepend jenkins jgit jhbuild jline jruby jruby-launcher
jsonpretty jsr292-cookbook jvm-language-runtime jvm-verifier jython
kawa launch4j learn_you_a_scalaz leiningen logula lookatgit mada
magpie maven-sbt maven-scala-plugin meow metascala minikanren
mirah mnemonic mozrepl my-benchmark nailgun neercs nestedvm
ninja~indify-repo noboxing-plugin ostrich osx-window-sizing ozma
parameterized-trigger-plugin parboiled pegdown perf4j play play-scala
projectplus pythagoras quickcursor rapture-io recursivity-commons
replhtml repo rugu samskivert sbt-idea sbt-twt sbteclipse scala-arm
scala-cel scala-collection-test scala-datetime scala-graph scala-ide
scala-io scala-javautils scala-json scala-proxy scala-query scala-redis
scala-refactoring scala-reflectionator scala-stm scala-symbol-browser
scala-time scalabot scalacheck scalaconsole scalacs scalafs
scalaj-collection scalaj-reflect scalamd scalang scalanlp-core
scalariform scalastyle scalate scalatra scalaz scalaz-camel scalex
scamacs scutil simile-butterfly sing slf4j smock snakeyaml solarized
sonatype-aether source.tmbundle spakle spark spde specs specs2
specs2-test sperformance standard-project stopwatch talking-puffin texto
threeten treetop twitter-util uniscala unplanned up vibrantinklion
view-files.txt vscaladoc windmill yarrgs yeti zmpp2


[2] the view hits.



anti-xml/src/main/scala/com/codecommit/antixml/Group.scala:205:
val b = VectorCase.newBuilder[B] ++= g.view.take(index)
anti-xml/src/main/scala/com/codecommit/antixml/ZipperHoleMap.scala:129:
  val els = depthFirst.view map { case (p,v) =>
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/CanonicalNameAST.scala:430:                val
condition =     matchCondition(matchCase.view, names)
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/ClassifyAST.scala:427:             matchCondition(target,
targetType, caseAST.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/RawAST.scala:169:                  matchCondition(it.view),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/ast/TypedRawAST.scala:140:             matchCondition(m.view,
types),
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:299:                    it.view.values.foreach(
arg => walkOps(ctx, arg) )
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:303:                            it.view.checkMethod,
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:305:                            (it.view.valueTypes.map(t
=> T.forClass(t)) ++ Seq(TypeDesc.OBJECT)).toArray
ashlar/compiler/src/main/scala/com/smokejumperit/ashlar/compiler/flow/CojenWalker.scala:325:                    val
bodyCtx = matchCase.view.exposedVariables.zipWithIndex.foldLeft(ctx) {
case (tmpCtx, (expVar, idx)) =>
blueeyes/src/main/scala/blueeyes/persistence/cache/functional/TemporalCache.scala:25:
    TemporalCacheState(removed.view.map(t => (t._1,
t._2.value)).toMap, TemporalCache(kept.toMap))
factorie/src/main/scala/cc/factorie/la/SparseBinaryVector.scala:182:
override def toString = getClass.getName+"("+"len="+length+"
1s=("+ind.view(0, _size).mkString("[", ", ", "]")+"))"
factorie/src/test/scala/cc/factorie/TestUtils.scala:32:    def
remainingSeq[T](s:Seq[(T, Int)], seq:Seq[(T, Int)]) =
seq.view(lastElemZipIndex(s)+1, seq.length)
mada/src/main/scala/sequence/vector/Scala.scala:17:        for (e <- _1.view) {
mada/src/test/scala/sequence/iterative/ConversionTest.scala:53:        x.view
mnemonic/bytecode/src/main/scala/net/virtualvoid/bytecode/backend/ASM.scala:310:
      candsAndLabels.view.unzip.zipped foreach emitBranch
ozma/src/compiler/scala/tools/nsc/backend/ozcode/OzCodes.scala:123:
     val indices = for ((param, idx) <- params.view.zipWithIndex
ozma/src/scalalib/scala/collection/generic/SliceInterval.scala:28:   *
 val coll = (1 to 100).view.slice(10, 30).slice(1, 3)
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:38:
override def view = self.view
ozma/src/scalalib/scala/collection/IterableProxyLike.scala:39:
override def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/parallel/ParIterableLike.scala:777:
  override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParIterableViewLike.scala:88:
  override def seq = forcedPar.seq.view.asInstanceOf[IterableView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/parallel/ParSeqLike.scala:344:
override def seq = self.seq.view
ozma/src/scalalib/scala/collection/parallel/ParSeqViewLike.scala:70:
 override def seq = forcedPar.seq.view.asInstanceOf[SeqView[S,
CollSeq]]
ozma/src/scalalib/scala/collection/SeqLike.scala:518:    b ++=
toCollection(rest).view drop replaced
ozma/src/scalalib/scala/collection/SeqLike.scala:531:    b ++=
toCollection(rest).view.tail
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:69:  override
def view = self.view
ozma/src/scalalib/scala/collection/SeqProxyLike.scala:70:  override
def view(from: Int, until: Int) = self.view(from, until)
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:94:
override def view = self.view
ozma/src/scalalib/scala/collection/TraversableProxyLike.scala:95:
override def view(from: Int, until: Int): TraversableView[A, Repr] =
self.view(from, until)
ozma/src/scalalib/scala/xml/factory/NodeFactory.scala:32:
ch1.view.zipAll(ch2.view, null, null) forall { case (x,y) => x eq y }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:531:
  def result() = { builder.result.view }
scala-collection-test/src/test/scala/com/bizo/scala/collection/TraversableSpec.scala:543:
  def result() = { builder.result.view(min, max) }
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:75:
    nodes.toList.view map nodeDegree filter degreeFilter min
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:89:
    nodes.toList.view map nodeDegree filter degreeFilter max
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:100:
  val v = nodes.toList.view map nodeDegree sorted IntReverseOrdering
scala-graph/core/src/main/scala/scalax/collection/GraphDegree.scala:115:
  else                          nodes.view map nodeDegree filter
degreeFilter)
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:36:
          (for (jsonNode <- nodeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:71:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:90:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:107:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:124:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:142:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:161:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:178:
              (for (jsonEdge <- edgeList.view) yield {
scala-graph/json/src/main/scala/scalax/collection/io/json/imp/Stream.scala:195:
              (for (jsonEdge <- edgeList.view) yield {
scala-io/core/src/main/scala/scalax/io/ArrayBufferSeekableChannel.scala:52:
    val toRead = data.view.slice(position.toInt, min(position.toInt +
dst.limit-dst.position, data.size))
scala-io/core/src/main/scala/scalax/io/JavaConverters.scala:68:
 lazy val chars = codec.decode(t.view.map{_.toByte}.toArray)
scala-io/core/src/main/scala/scalax/io/LongTraversableLike.scala:687://
 override def view = super.view
scala-io/core/src/test/scala/scalaio/test/LongTraversableTest.scala:338:
  val list = (f(expectedData().toList.view) map { i => expectedCount
+= 1; i })
scala-query/src/main/scala/org/scalaquery/ql/Query.scala:42:    new
Query[E, U](unpackable, cond, condHaving, modifiers ::: by.view.map(c
=> new Grouping(Node(c))).toList)
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:45:
m <- getClass().getMethods.view
scala-query/src/main/scala/org/scalaquery/ql/TableBase.scala:61:
m <- getClass().getMethods.view
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:17:
 tpe.getAnnotations.view collect { case x: ScalaSignature => x } map
{sigBytesFromAnnotation} headOption
scalaj-reflect/src/main/scala/scalaj/reflect/AutographBook.scala:41:
 splits.view flatMap { case (l,r) => safeGetClass(l) map (_ -> r) }
headOption
Scalala/src/main/scala/scalala/library/Library.scala:213:
matrix.domain._1.view.map(sliceRows(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Library.scala:223:
matrix.domain._2.view.map(sliceCols(matrix,_)).reduceLeft(opAdd);
Scalala/src/main/scala/scalala/library/Plotting.scala:464://
val values = items.view.map(item =>
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/PTBTokenizer.scala:92:

SL(pref.mkString("") + d + tail.view.map(_.chars).mkString(""))
scalanlp-core/data/src/main/scala/scalanlp/text/tokenize/Tokenizer.scala:102:
    if (f.isInstanceOf[Transformer]) f else f.andThen((i :
Iterable[String]) => i.view);
scalanlp-core/data/src/main/scala/scalanlp/util/ScalaQL.scala:60:
peeked.view(0, n);
scalanlp-core/learn/src/main/scala/scalanlp/classify/LogisticClassifier.scala:101:
      for( datum <- range.view map data) {
scalanlp-core/learn/src/main/scala/scalanlp/maxent/MaxEntObjectiveFunction.scala:154:
  val (grad: DenseVector[Double],prob: Double) =
eCounts.zipWithIndex.par.view.map { case (vec,c) =>
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:42:
  val baseDiff = diff(lpairs.view.map(_._1),lpairs.view.map(_._2));
scalanlp-core/learn/src/main/scala/scalanlp/stats/RandomizationTest.scala:72:
  ) apply (dataset.view.map(Example.lift(c1)).map(_.label).force,dataset.view.map(Example.lift(c2)).map(_.features).force);
scalate/samples/scalate-sample/src/main/scala/org/fusesource/scalate/sample/ServletRendersView.scala:40:
  context.view(model)
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateView.scala:62:
  context.view(it.get.asInstanceOf[AnyRef])
scalate/scalate-spring-mvc/src/main/scala/org/fusesource/scalate/spring/view/ScalateViewResolver.scala:29:
 override def requiredViewClass(): java.lang.Class[_] =
classOf[org.fusesource.scalate.spring.view.ScalateView]
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:32:
    file.listFiles.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:40:
  directories.view.map(recursiveFind(_)(filter)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Files.scala:48:
  children(file).view.flatMap(andDescendants(_)))
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/Objects.scala:88:
  constructors.view.map(c =>
tryCreate(c)).find(_.isDefined).getOrElse(None)
scalate/scalate-util/src/main/scala/org/fusesource/scalate/util/ResourceLoader.scala:89:
  sourceDirectories.view.map(new File(_, uri)).find(_.exists) match {
scalate/scalate-wikitext/src/main/scala/org/fusesource/scalate/wikitext/SwizzleLinkFilter.scala:94:
  sourceDirectories.view.map
{findMatching(_)}.find(_.isDefined).getOrElse(None)
scutil/src/main/scala/scutil/ext/InputStreamImplicits.scala:23:                 if
(len != -1)     out ++= buffer.view(0, len)
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:51:
argumentsFragments.view.zip(nestedArguments).collect { case
(ApplicableArguments(value), args) => (value, args) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:55:
def fragmentAndSpecNames: Seq[(T, SpecName)] =
argumentsFragments.view.zip(nestedSpecNames).collect { case
(ApplicableArguments(value), name) => (value, name) }
specs2/src/main/scala/org/specs2/reporter/SpecsArguments.scala:61:
argumentsFragments.view.zip(nestedArguments).zip(nestedSpecNames).collect
{ case ((ApplicableArguments(value), args), name) => (value, args,
name) }
specs2/src/main/scala/org/specs2/runner/SpecificationsFinder.scala:23:
  specificationNames(path, pattern, basePath,
verbose).view.filter(filter).flatMap(n => createSpecification(n))
sperformance/src/test/scala/CollectionsShootout.scala:148:
List.range(1, size).toList.view
sperformance/src/test/scala/CollectionsShootout.scala:155:
ArrayBuffer.range(1, size).view
sperformance/src/test/scala/CollectionsShootout.scala:184:
col.view.map(_ * 2).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:189:
col.view.filter(_ % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:194:
col.view.zipWithIndex.filter(_._2 % 2 == 0).take(100).force
sperformance/src/test/scala/CollectionsShootout.scala:236:
performance of "Vector.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:239:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:244:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
sperformance/src/test/scala/CollectionsShootout.scala:260:
performance of "ArrayBuffer.view" in {
sperformance/src/test/scala/CollectionsShootout.scala:263:
a.view.zipWithIndex.filter(x => x._1.startsWith("1") &&
(x._2%3==0)).map(_._1).force
sperformance/src/test/scala/CollectionsShootout.scala:268:
a.view.filter(_.startsWith("1")).foldLeft("")( (prev, cur) => cur)
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:61:
    val result = codecs.urlParamCodecs.view map { codec =>
uniscala/uniscala-granite/granite-core/src/main/scala/net/uniscala/granite/DefaultIDUrlParamCodec.scala:86:
    val result = codecs.urlParamCodecs.view map { codec =>

Hi everyone,

while I don't think view should be removed completely, I understand that it is under-used, a huge source of bugs, too complicated and people would prefer spending their time on better things.
I think separating the view stuff into a separate JAR file would be a good first step. Then just wait and see how it works out, e. g. if usage stays low, if bugs keep coming, etc. Moving it out of the standard lib would probably make it possible to implement necessary changes faster.

Whatever the outcome is, I would prefer if

• The "Traversable" thing
  
• The "This" thing
• The "View" thing

would happen in the same release. If we are concerned to break compatibility when merging/removing Traversable, I would really prefer going to whole way in one step. Not breaking (in the worst case) stuff in three separate releases.

Thanks and bye,

Simon

On Sun, Dec 18, 2011 at 10:39 PM, Simon Ochsenreither <simon.ochsenreither@googlemail.com> wrote:

Hi everyone,

while I don't think view should be removed completely, I understand that it is under-used, a huge source of bugs, too complicated and people would prefer spending their time on better things.
I think separating the view stuff into a separate JAR file would be a good first step. Then just wait and see how it works out, e. g. if usage stays low, if bugs keep coming, etc. Moving it out of the standard lib would probably make it possible to implement necessary changes faster.

Whatever the outcome is, I would prefer if

• The "Traversable" thing
  
• The "This" thing
• The "View" thing

would happen in the same release. If we are concerned to break compatibility when merging/removing Traversable, I would really prefer going to whole way in one step. Not breaking (in the worst case) stuff in three separate releases.

Sorry to be off-topic, but what's this about removing traversable? I find it extremely useful to have something where I just have to implement foreach to get full collection support (e.g. tree traversal, where writing foreach is trivial, but writing an efficient iterator is a PITA). 

On Sat, Dec 17, 2011 at 9:09 AM, Francois <fanf42@gmail.com> wrote:

On 17/12/2011 16:47, Johannes Rudolph wrote:

I'm for splitting them out of the core library. I tried to use them
many times, they seldom did what I expected. I'm now most of the times
using Iterators or another custom solution.

Same for me. When I read about view for the first time, I was thrilled by the feature, and was wondering why it was not the default behaviour for all collection iteration. I did try to use them, sometimes getting completlly unexpected results, most of the time hitting a bug along the road. After some time, I realize that they weren't the free lunch I was hopping for and I found that it was far safer to just ignore them, and build custom solution when the need happened.

On the other hand, I'm all for a modularized and/or lighter Scala-lib, so if my vote count, I'm for removing them.
Ok, well, either that, or make what is needed to have them be really useful (useful as in "in Scala, the default pattern for fast iteration is "for { elt <- mycollection.view ... }, that a no brainer")

This pretty much matches my experience and recommendation for views.
+1
alex

Thanks for such a good analysis. I like the idea of views, they are conceptually useful I think. I can't speak to the bugginess of them, because I don't use them extensively.
I do have one question though: If they were to be removed, would this break binary compatibility between major versions?

Matthew Farwell.
On Saturday, 17 December 2011 05:55:13 UTC+1, Paul Phillips wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

    % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
    -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
    % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
    -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

    % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
      904766
    % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
          92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

  https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

  Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.

Scala usually breaks binary compatibility between major releases and retains it for bug fix releases.

When 2.10 is released we'll try to jumpstart the community libraries (those that opt in) so that 2.10.x versions are immediately available.

On Dec 20, 2011 9:18 AM, "Matthew Farwell" <matthew@farwell.co.uk> wrote:

Thanks for such a good analysis. I like the idea of views, they are conceptually useful I think. I can't speak to the bugginess of them, because I don't use them extensively.
I do have one question though: If they were to be removed, would this break binary compatibility between major versions?

Matthew Farwell.
On Saturday, 17 December 2011 05:55:13 UTC+1, Paul Phillips wrote:

Speaking of things not pulling their weight: views.

Library jar before and after ripping out views.

    % ls -l /scratch/trunk2/build/pack/lib/scala-library.jar
    -rw-r--r--  1 paulp  admin  7161826 Dec 16 19:53
/scratch/trunk2/build/pack/lib/scala-library.jar
    % ls -l /scratch/trunk3/build/pack/lib/scala-library.jar
    -rw-r--r--  1 paulp  admin  6561034 Dec 16 19:51
/scratch/trunk3/build/pack/lib/scala-library.jar

Statistics from the scala code in my general ~/src directory.

    % find . -name '*.scala' -print0 |xargs -0 cat | wc -l
      904766
    % find . -name '*.scala' -print0 |xargs -0 ack '\.view\b' |egrep
-v ':(import|package)' | grep -v deprecated |wc -l
          92

One view usage per 10,000 lines of scala code. (And that may be
generous, as some of those 92 hits are either copies of compiler/library
source or things testing the compiler/library source. However I could
have missed some dotless uses so we'll call it a wash.)

So a feature used in .01% of lines of scala written is consuming 9% of
the jar. I focus on 9% not because I'm obsessed with jar size but as a
proxy for all the overhead associated with any large body of code. And
views have not been small in the overhead department, and they're still
not close to being a robust unit.

  https://issues.scala-lang.org/secure/IssueNavigator.jspa?mode=hide&requestId=10906

  Displaying issues 1 to 49 of 49 matching issues. [...]

Solution: exile to separate jar, deprecate for 2.10, and retire
them afterward. They can easily be pimped in by any who wish to use
them, so the 99.99% of lines shouldn't have to pay the inheritance tax to
make things very slightly easier on the .01%. (I'm not even sure it
makes things slightly easier.) And maybe many view designs will flourish
in this brave new viewless world.

If we were determined not to let those view methods go away, we could
stub them as empty methods in the core library and have some way to
configure them to invoke a library elsewhere.  Anything to get them out of the
main collections inheritance hierarchy.

Hi Paul,
I have used views in some places where without them the code would have needed to be written very differently to avoid unacceptable slow-down and to regain acceptable memory use. I have also hit weird bugs where adding a .view() call has broken my code. I've had a brief look at the code that implements views, and some of it is fairly gnarley, and it makes the inheritance (implementation vs interface) network far to complex for me to be able to hold in my head.
The ops I use with view tend to be: zipWithIndex;map; flatMap; filter; foreach. In my code, chains of ops on a view tend to be terminated with an explicit toSet or toMap or something to materialize the result.
It is not hard to see how something performing optimization at a higher level can rework a chain of maps and filters, fusing the loops and functions, at least for some of the collection types. If we had compiler support for this that was reliable then there would almost certainly be no need for views in the first place. This would be my preferred fix. After that, moving views out of the core collections jar into a views jar with pimps would work for me.
Matthew
On 17 December 2011 18:24, Paul Phillips <paulp@improving.org> wrote:

On Sat, Dec 17, 2011 at 10:15 AM, Miles Sabin <miles@milessabin.com> wrote:
> Does dropping views in favour of iterators in the compiler have any
> effect on compilation performance?

My guess is yes, for the better, especially if we properly account for
the occasional five order of magnitude slowdown in views.  But I don't
know, one can only gather so much evidence and I feel like I've
already shown enough to put the guy in the chair while simultaneously
being lethally injected and hanged.

 https://github.com/scala/scala/commit/be49752855

A patch for views. Most relevant change:

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.

--
Dr Matthew PocockIntegrative Bioinformatics Group, School of Computing Science, Newcastle Universitymailto: turingatemyhamster@gmail.com gchat: turingatemyhamster@gmail.commsn: matthew_pocock@yahoo.co.uk irc.freenode.net: drdozerskype: matthew.pococktel: (0191) 2566550mob: +447535664143

p, li { white-space: pre-wrap; }

Am Dienstag, 20. Dezember 2011, 15:37:04 schrieb Matthew Pocock:

> Hi Paul,

>

...

> It is not hard to see how something performing optimization at a higher

> level can rework a chain of maps and filters, fusing the loops and

> functions, at least for some of the collection types. If we had compiler

> support for this that was reliable then there would almost certainly be no

> need for views in the first place. This would be my preferred fix. After

> that, moving views out of the core collections jar into a views jar with

> pimps would work for me.

This would be fantastic!

I thought the same as I had a brief look at the compile{} statement example - I thought: why only for GPU why not also on a standard architecture? Just have a code block (e.g. a "typical" collection processing chain) and instruct the compiler to work especially hard to optimize this thing by transformations - at the cost of a longer compile time and a reduced instruction set within such a block.

I think a clever optimization of any and all scala code is way beyond reach (maybe I am wrong) but optimizing a distinct block with a reduced language model could be feasible. At the same time this would allow to control where the compiler should spend extra time to get things as fast as possible.

Just my 5 cents.

Greetings

Bernd

Data point: in my 4 years of Scala, I've never used views in real code
because (as a watcher of the ticket database) I've never trusted them.
The main issue was that I didn't trust them to actually speed anything
up in practice, and I still don't. But also, I didn't trust them to
even work, and I still don't.

I would rejoice to see them removed or at least separated from the
main collections code and downplayed. For many reasons, not the least
of which is jar size.

Martin points out that sometimes using a view will improve the
asymptotic complexity of a piece of code. That's true, but aren't
there satisfactory alternative ways to get the same speedup? (Someone
mentioned iterators.) I'd need to see a strong collection of examples
to be convinced otherwise, especially given that having views around
is so expensive for the collections implementers.

On Mon, Jan 2, 2012 at 4:45 PM, Seth Tisue <seth@tisue.net> wrote:

Martin points out that sometimes using a view will improve the
asymptotic complexity of a piece of code. That's true, but aren't
there satisfactory alternative ways to get the same speedup? (Someone
mentioned iterators.)

I use the concept of views--but not the collections views--extensively in image processing code.  Very often, one wants to focus in on a small subset of an image and do all sorts of things to it--count pixels, find objects, whatever--without having to worry that it is just a subset of the real image, or that the data type is a different size, or whatever.  (Java's BufferedImage does similar things, though in a clunky and not-very-well-explained way.)  This ends up being absolutely essential for not having dozens of copies of the image present all over the place; for demanding tasks, it translates into needing a machine with 2-4x more memory (or more, depending on the length of the transformation chain).

If views had worked and collections had been specialized, I would have used views here, and iterators would have been useless, since one does need to refer to the same data multiple times (sometimes with random access).  However, since views were sorta broken, and the collections are not specialized, I didn't.  Nonetheless, the approach is extremely valuable whenever you're working with large data sets and need to perform multiple transformations on the full data set and/or small subsets in order to generate what you eventually want.

Thus, although I agree that views should be moved out of the standard core library for the time being, I don't think it's wise to just throw the capability away and not look back.

  --Rex

P.S. ImgLib 2.0, the image processing back end to (the next version of) ImageJ, also uses view-like capability extensively.

Here's what I'm looking at right now.  Proof of concept only.  These classes would be for the implementation, you'd use views the same way you use collections.  Although one could make the case that the world is upside down, and we ought to compose functions like this and apply the composite to a collection.  Programming with functions... oh, it'll never catch on.
scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o DropWhile(_ < 10) o Map(_.toString)op: listView.ViewOp[List[Int],List[String]] = <function1>
scala> op.describe("coll")res0: String = coll.takeWhile(<function1>).filter(<function1>).dropWhile(<function1>).map(<function1>)
scala> op(1 to 100 toList)res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48)
scala> op(-50 to 10 toList)res2: List[String] = List(10)
scala> Slice(100, 500) o Slice(10, 20)res4: listView.ViewOp[List[Int],List[Int]] = <function1>
scala> res4.describe("xs")res5: String = xs.slice(110, 120)
scala> res4(1 to 1000 toList)res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119, 120)

This is very much along the lines of what I was thinking.  If it scales to everything we want to cover, I'm all in favor!

  --Rex

P.S. I like the lower-case o trick.

On Tue, Jan 3, 2012 at 2:24 AM, Paul Phillips <paulp@improving.org> wrote:

Here's what I'm looking at right now.  Proof of concept only.  These classes would be for the implementation, you'd use views the same way you use collections.  Although one could make the case that the world is upside down, and we ought to compose functions like this and apply the composite to a collection.  Programming with functions... oh, it'll never catch on.
scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o DropWhile(_ < 10) o Map(_.toString)op: listView.ViewOp[List[Int],List[String]] = <function1>
scala> op.describe("coll")res0: String = coll.takeWhile(<function1>).filter(<function1>).dropWhile(<function1>).map(<function1>)
scala> op(1 to 100 toList)res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48)
scala> op(-50 to 10 toList)res2: List[String] = List(10)
scala> Slice(100, 500) o Slice(10, 20)res4: listView.ViewOp[List[Int],List[Int]] = <function1>
scala> res4.describe("xs")res5: String = xs.slice(110, 120)
scala> res4(1 to 1000 toList)res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119, 120)

I find it extremely ironic that this essentially introduces point-free
programming for collections. And, yet, I think modeling views as
classes on their own -- dissociated from data -- makes a lot of sense.
Like, to intentionally co-opt a term(*), a lens that is put in front
of a collection, which modifies what we look at. At any rate, I
thought of two ways that might enhance it.

First, one could write it method-like, same as collections:

new SeqView takeWhile (_ < 50) filter (_ % 2 == 0) dropWhile (_ < 10)
map (_.toString)

I expect this could be a layer over what you have:

class SeqView(val op: ViewOp = Id) {
def takeWhile(f: Function1) = new SeqView(op o TakeWhile f)
def filter(f: Fucntion1) = new SeqView(op o Filter f)
// etc
}

I'm not sure how you managed the type parameters/type inference, however.

The second thing I think would made this propose better would be to
define a method on the collection to apply the view:

class List {
def getView(op: ViewOp) = op(this)
def getView(view: SeqView) = getView(view.op)
}

Taking both together, we can write this:

val myView = new SeqView takeWhile (_ < 50) filter (_ % 2 == 0)
dropWhile (_ < 10) map (_.toString)
(1 to 100).toList getView myView

Or even

(1 to 100).toList getView (new SeqView takeWhile (_ < 50) filter (_ %
2 == 0) dropWhile (_ < 10) map (_.toString))

(*) According to
http://www.cis.upenn.edu/~bcpierce/papers/lenses-etapsslides.pdf, if
the composition is made entirely out of slices (take/drop variants),
it could be a well-behaved lens. Of course, that assumes a putback
(update/updated) functionality is provided. I'm guessing putback could
be made available, but it also could be made available *later*.

On Tue, Jan 3, 2012 at 05:24, Paul Phillips wrote:
> Here's what I'm looking at right now.  Proof of concept only.  These classes
> would be for the implementation, you'd use views the same way you use
> collections.  Although one could make the case that the world is upside
> down, and we ought to compose functions like this and apply the composite to
> a collection.  Programming with functions... oh, it'll never catch on.
>
> scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o DropWhile(_ <
> 10) o Map(_.toString)
> op: listView.ViewOp[List[Int],List[String]] =
>
> scala> op.describe("coll")
> res0: String =
> coll.takeWhile().filter().dropWhile().map()
>
> scala> op(1 to 100 toList)
> res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32,
> 34, 36, 38, 40, 42, 44, 46, 48)
>
> scala> op(-50 to 10 toList)
> res2: List[String] = List(10)
>
> scala> Slice(100, 500) o Slice(10, 20)
> res4: listView.ViewOp[List[Int],List[Int]] =
>
> scala> res4.describe("xs")
> res5: String = xs.slice(110, 120)
>
> scala> res4(1 to 1000 toList)
> res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119, 120)
>

I think the aim is to keep the current ".view" and ".force" methods to convert to and from views, but use the new 'ops' class to implement/optimise view behavior for all collections.
So, the way we use views won't be leaving, it'll just be a lot simpler and more powerful.
- Josh

On Tue, Jan 3, 2012 at 7:39 AM, Daniel Sobral <dcsobral@gmail.com> wrote:

I find it extremely ironic that this essentially introduces point-free
programming for collections. And, yet, I think modeling views as
classes on their own -- dissociated from data -- makes a lot of sense.
Like, to intentionally co-opt a term(*), a lens that is put in front
of a collection, which modifies what we look at. At any rate, I
thought of two ways that might enhance it.

First, one could write it method-like, same as collections:

new SeqView takeWhile (_ < 50) filter (_ % 2 == 0) dropWhile (_ < 10)
map (_.toString)

I expect this could be a layer over what you have:

class SeqView(val op: ViewOp = Id) {
 def takeWhile(f: Function1) = new SeqView(op o TakeWhile f)
 def filter(f: Fucntion1) = new SeqView(op o Filter f)
 // etc
}

I'm not sure how you managed the type parameters/type inference, however.

The second thing I think would made this propose better would be to
define a method on the collection to apply the view:

class List {
 def getView(op: ViewOp) = op(this)
 def getView(view: SeqView) = getView(view.op)
}

Taking both together, we can write this:

val myView = new SeqView takeWhile (_ < 50) filter (_ % 2 == 0)
dropWhile (_ < 10) map (_.toString)
(1 to 100).toList getView myView

Or even

(1 to 100).toList getView (new SeqView takeWhile (_ < 50) filter (_ %
2 == 0) dropWhile (_ < 10) map (_.toString))


(*) According to
http://www.cis.upenn.edu/~bcpierce/papers/lenses-etapsslides.pdf, if
the composition is made entirely out of slices (take/drop variants),
it could be a well-behaved lens. Of course, that assumes a putback
(update/updated) functionality is provided. I'm guessing putback could
be made available, but it also could be made available *later*.

On Tue, Jan 3, 2012 at 05:24, Paul Phillips <paulp@improving.org> wrote:
> Here's what I'm looking at right now.  Proof of concept only.  These classes
> would be for the implementation, you'd use views the same way you use
> collections.  Although one could make the case that the world is upside
> down, and we ought to compose functions like this and apply the composite to
> a collection.  Programming with functions... oh, it'll never catch on.
>
> scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o DropWhile(_ <
> 10) o Map(_.toString)
> op: listView.ViewOp[List[Int],List[String]] = <function1>
>
> scala> op.describe("coll")
> res0: String =
> coll.takeWhile(<function1>).filter(<function1>).dropWhile(<function1>).map(<function1>)
>
> scala> op(1 to 100 toList)
> res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32,
> 34, 36, 38, 40, 42, 44, 46, 48)
>
> scala> op(-50 to 10 toList)
> res2: List[String] = List(10)
>
> scala> Slice(100, 500) o Slice(10, 20)
> res4: listView.ViewOp[List[Int],List[Int]] = <function1>
>
> scala> res4.describe("xs")
> res5: String = xs.slice(110, 120)
>
> scala> res4(1 to 1000 toList)
> res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119, 120)
>



--
Daniel C. Sobral

I travel to the future all the time.

Is there any good reading out there about code transformations, from these kind of chained operations to "imperative" equivalents ? I assume some patterns are well-known and could be easily converted thanks to category theory, but is it the case for all (or the most-used) combinations of the collection methods ? Wouldn't that imply a case-by-case implementation of the 'o' combinator, which would grow exponentially in the number of combinators that collection themselves already use as methods ?

2012/1/3 Josh Suereth <joshua.suereth@gmail.com>

I think the aim is to keep the current ".view" and ".force" methods to convert to and from views, but use the new 'ops' class to implement/optimise view behavior for all collections.
So, the way we use views won't be leaving, it'll just be a lot simpler and more powerful.
- Josh

On Tue, Jan 3, 2012 at 7:39 AM, Daniel Sobral <dcsobral@gmail.com> wrote:

I find it extremely ironic that this essentially introduces point-free
programming for collections. And, yet, I think modeling views as
classes on their own -- dissociated from data -- makes a lot of sense.
Like, to intentionally co-opt a term(*), a lens that is put in front
of a collection, which modifies what we look at. At any rate, I
thought of two ways that might enhance it.

First, one could write it method-like, same as collections:

new SeqView takeWhile (_ < 50) filter (_ % 2 == 0) dropWhile (_ < 10)
map (_.toString)

I expect this could be a layer over what you have:

class SeqView(val op: ViewOp = Id) {
 def takeWhile(f: Function1) = new SeqView(op o TakeWhile f)
 def filter(f: Fucntion1) = new SeqView(op o Filter f)
 // etc
}

I'm not sure how you managed the type parameters/type inference, however.

The second thing I think would made this propose better would be to
define a method on the collection to apply the view:

class List {
 def getView(op: ViewOp) = op(this)
 def getView(view: SeqView) = getView(view.op)
}

Taking both together, we can write this:

val myView = new SeqView takeWhile (_ < 50) filter (_ % 2 == 0)
dropWhile (_ < 10) map (_.toString)
(1 to 100).toList getView myView

Or even

(1 to 100).toList getView (new SeqView takeWhile (_ < 50) filter (_ %
2 == 0) dropWhile (_ < 10) map (_.toString))


(*) According to
http://www.cis.upenn.edu/~bcpierce/papers/lenses-etapsslides.pdf, if
the composition is made entirely out of slices (take/drop variants),
it could be a well-behaved lens. Of course, that assumes a putback
(update/updated) functionality is provided. I'm guessing putback could
be made available, but it also could be made available *later*.

On Tue, Jan 3, 2012 at 05:24, Paul Phillips <paulp@improving.org> wrote:
> Here's what I'm looking at right now.  Proof of concept only.  These classes
> would be for the implementation, you'd use views the same way you use
> collections.  Although one could make the case that the world is upside
> down, and we ought to compose functions like this and apply the composite to
> a collection.  Programming with functions... oh, it'll never catch on.
>
> scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o DropWhile(_ <
> 10) o Map(_.toString)
> op: listView.ViewOp[List[Int],List[String]] = <function1>
>
> scala> op.describe("coll")
> res0: String =
> coll.takeWhile(<function1>).filter(<function1>).dropWhile(<function1>).map(<function1>)
>
> scala> op(1 to 100 toList)
> res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32,
> 34, 36, 38, 40, 42, 44, 46, 48)
>
> scala> op(-50 to 10 toList)
> res2: List[String] = List(10)
>
> scala> Slice(100, 500) o Slice(10, 20)
> res4: listView.ViewOp[List[Int],List[Int]] = <function1>
>
> scala> res4.describe("xs")
> res5: String = xs.slice(110, 120)
>
> scala> res4(1 to 1000 toList)
> res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119, 120)
>



--
Daniel C. Sobral

I travel to the future all the time.

Yes, have a look at all the deforestation papers.

On Wed, Jan 4, 2012 at 01:59, Alex Repain wrote:
> Is there any good reading out there about code transformations, from these
> kind of chained operations to "imperative" equivalents ? I assume some
> patterns are well-known and could be easily converted thanks to category
> theory, but is it the case for all (or the most-used) combinations of the
> collection methods ? Wouldn't that imply a case-by-case implementation of
> the 'o' combinator, which would grow exponentially in the number of
> combinators that collection themselves already use as methods ?
>
>
> 2012/1/3 Josh Suereth
>>
>> I think the aim is to keep the current ".view" and ".force" methods to
>> convert to and from views, but use the new 'ops' class to implement/optimise
>> view behavior for all collections.
>>
>> So, the way we use views won't be leaving, it'll just be a lot simpler and
>> more powerful.
>>
>> - Josh
>>
>>
>> On Tue, Jan 3, 2012 at 7:39 AM, Daniel Sobral wrote:
>>>
>>> I find it extremely ironic that this essentially introduces point-free
>>> programming for collections. And, yet, I think modeling views as
>>> classes on their own -- dissociated from data -- makes a lot of sense.
>>> Like, to intentionally co-opt a term(*), a lens that is put in front
>>> of a collection, which modifies what we look at. At any rate, I
>>> thought of two ways that might enhance it.
>>>
>>> First, one could write it method-like, same as collections:
>>>
>>> new SeqView takeWhile (_ < 50) filter (_ % 2 == 0) dropWhile (_ < 10)
>>> map (_.toString)
>>>
>>> I expect this could be a layer over what you have:
>>>
>>> class SeqView(val op: ViewOp = Id) {
>>>  def takeWhile(f: Function1) = new SeqView(op o TakeWhile f)
>>>  def filter(f: Fucntion1) = new SeqView(op o Filter f)
>>>  // etc
>>> }
>>>
>>> I'm not sure how you managed the type parameters/type inference, however.
>>>
>>> The second thing I think would made this propose better would be to
>>> define a method on the collection to apply the view:
>>>
>>> class List {
>>>  def getView(op: ViewOp) = op(this)
>>>  def getView(view: SeqView) = getView(view.op)
>>> }
>>>
>>> Taking both together, we can write this:
>>>
>>> val myView = new SeqView takeWhile (_ < 50) filter (_ % 2 == 0)
>>> dropWhile (_ < 10) map (_.toString)
>>> (1 to 100).toList getView myView
>>>
>>> Or even
>>>
>>> (1 to 100).toList getView (new SeqView takeWhile (_ < 50) filter (_ %
>>> 2 == 0) dropWhile (_ < 10) map (_.toString))
>>>
>>>
>>> (*) According to
>>> http://www.cis.upenn.edu/~bcpierce/papers/lenses-etapsslides.pdf, if
>>> the composition is made entirely out of slices (take/drop variants),
>>> it could be a well-behaved lens. Of course, that assumes a putback
>>> (update/updated) functionality is provided. I'm guessing putback could
>>> be made available, but it also could be made available *later*.
>>>
>>> On Tue, Jan 3, 2012 at 05:24, Paul Phillips wrote:
>>> > Here's what I'm looking at right now.  Proof of concept only.  These
>>> > classes
>>> > would be for the implementation, you'd use views the same way you use
>>> > collections.  Although one could make the case that the world is upside
>>> > down, and we ought to compose functions like this and apply the
>>> > composite to
>>> > a collection.  Programming with functions... oh, it'll never catch on.
>>> >
>>> > scala> val op = Id o TakeWhile(_ < 50) o Filter(_ % 2 == 0) o
>>> > DropWhile(_ <
>>> > 10) o Map(_.toString)
>>> > op: listView.ViewOp[List[Int],List[String]] =
>>> >
>>> > scala> op.describe("coll")
>>> > res0: String =
>>> >
>>> > coll.takeWhile().filter().dropWhile().map()
>>> >
>>> > scala> op(1 to 100 toList)
>>> > res1: List[String] = List(10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30,
>>> > 32,
>>> > 34, 36, 38, 40, 42, 44, 46, 48)
>>> >
>>> > scala> op(-50 to 10 toList)
>>> > res2: List[String] = List(10)
>>> >
>>> > scala> Slice(100, 500) o Slice(10, 20)
>>> > res4: listView.ViewOp[List[Int],List[Int]] =
>>> >
>>> > scala> res4.describe("xs")
>>> > res5: String = xs.slice(110, 120)
>>> >
>>> > scala> res4(1 to 1000 toList)
>>> > res6: List[Int] = List(111, 112, 113, 114, 115, 116, 117, 118, 119,
>>> > 120)
>>> >
>>>
>>>
>>>
>>> --
>>> Daniel C. Sobral
>>>
>>> I travel to the future all the time.
>>
>>
>