[Truffle] Don't calculate the entire byte[] when fetching a single byte.

The original approach works well when String#getbyte is called multiple times on the same string, since the rope pays the cost to calculate the array once and then can do quick access. However, this approach fell apart miserably when single byte access is performed over large ropes with sparse byte[]. As an example, pr-zlib does this frequently in its pure Ruby implementation of zlib.
jruby · Mar 28, 2016 · 893d2e5 · 893d2e5
1 parent dd1552e
commit 893d2e5
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/truffle/src/main/java/org/jruby/truffle/core/rope/RopeNodes.java b/truffle/src/main/java/org/jruby/truffle/core/rope/RopeNodes.java
@@ -653,7 +653,7 @@ public int getByte(Rope rope, int index) {
         @Specialization(guards = "rope.getRawBytes() == null")
         @TruffleBoundary
         public int getByteSlow(Rope rope, int index) {
-            return rope.getBytes()[index] & 0xff;
+            return rope.get(index) & 0xff;
         }
 
     }