Merge pull request #3860 from gourlaysama/wip/t7710-backport

[backport] SI-7710 fix memory performance of RegexParsers in jdk7u6+

1 files changed, 32 insertions, 0 deletions

diff --git a/src/library/scala/util/parsing/combinator/SubSequence.scala b/src/library/scala/util/parsing/combinator/SubSequence.scala
new file mode 100644
index 0000000000..79c8acac0f
--- /dev/null
+++ b/src/library/scala/util/parsing/combinator/SubSequence.scala
@@ -0,0 +1,32 @@
+/*                     __                                               *\
+**     ________ ___   / /  ___     Scala API                            **
+**    / __/ __// _ | / /  / _ |    (c) 2006-2013, LAMP/EPFL             **
+**  __\ \/ /__/ __ |/ /__/ __ |    http://scala-lang.org/               **
+** /____/\___/_/ |_/____/_/ | |                                         **
+**                          |/                                          **
+\*                                                                      */
+
+
+package scala
+package util.parsing.combinator
+
+// A shallow wrapper over another CharSequence (usually a String)
+//
+// See SI-7710: in jdk7u6 String.subSequence stopped sharing the char array of the original
+// string and began copying it.
+// RegexParsers calls subSequence twice per input character: that's a lot of array copying!
+private[combinator] class SubSequence(s: CharSequence, start: Int, val length: Int) extends CharSequence {
+  def this(s: CharSequence, start: Int) = this(s, start, s.length - start)
+
+  def charAt(i: Int) =
+    if (i >= 0 && i < length) s.charAt(start + i) else throw new IndexOutOfBoundsException(s"index: $i, length: $length")
+
+  def subSequence(_start: Int, _end: Int) = {
+    if (_start < 0 || _end < 0 || _end > length || _start > _end)
+      throw new IndexOutOfBoundsException(s"start: ${_start}, end: ${_end}, length: $length")
+
+    new SubSequence(s, start + _start, _end - _start)
+  }
+
+  override def toString = s.subSequence(start, start + length).toString
+}