[GR-32765] Pass an interop array to TruffleRegex if we need to give a substring for startPos like with strscan

* Avoids materializing many substring byte[], which is very inefficient.
* Fixes #2407

Author: eregon

src/main/java/org/truffleruby/core/regexp/TruffleRegexpNodes.java

@@ -436,7 +436,7 @@ public abstract static class MatchInRegionTRegexNode extends PrimitiveArrayArgum
         @Child DispatchNode stringDupNode;
         @Child TranslateInteropExceptionNode translateInteropExceptionNode;
 
-        @Child RopeNodes.SubstringNode substringNode;
+        @Child RopeNodes.GetBytesObjectNode getBytesObjectNode;
 
         @Specialization(guards = "libString.isRubyString(string)")
         protected Object matchInRegionTRegex(
@@ -466,16 +466,27 @@ protected Object matchInRegionTRegex(
             }
 
             int fromIndex = fromPos;
+            final Object interopByteArray;
+            final String execMethod;
             if (startPosNotZeroProfile.profile(startPos > 0)) {
-                rope = substring(rope, startPos, toPos - startPos);
+                // GR-32765: When adopting TruffleString, use a TruffleString substring here instead
                 // If startPos != 0, then fromPos == startPos.
                 assert fromPos == startPos;
                 fromIndex = 0;
+
+                if (getBytesObjectNode == null) {
+                    CompilerDirectives.transferToInterpreterAndInvalidate();
+                    getBytesObjectNode = insert(RopeNodes.GetBytesObjectNode.create());
+                }
+                interopByteArray = getBytesObjectNode.getRange(rope, startPos, toPos);
+                execMethod = "exec";
+            } else {
+                final byte[] bytes = bytesNode.execute(rope);
+                interopByteArray = getContext().getEnv().asGuestValue(bytes);
+                execMethod = "execBytes";
             }
 
-            final byte[] bytes = bytesNode.execute(rope);
-            final Object interopByteArray = getContext().getEnv().asGuestValue(bytes);
-            final Object result = invoke(regexInterop, tRegex, "execBytes", interopByteArray, fromIndex);
+            final Object result = invoke(regexInterop, tRegex, execMethod, interopByteArray, fromIndex);
 
             final boolean isMatch = (boolean) readMember(resultInterop, result, "isMatch");
 
@@ -570,15 +581,6 @@ private Object dupString(Object string) {
 
             return stringDupNode.call(string, "dup");
         }
-
-        private Rope substring(Rope rope, int byteOffset, int byteLength) {
-            if (substringNode == null) {
-                CompilerDirectives.transferToInterpreterAndInvalidate();
-                substringNode = insert(RopeNodes.SubstringNode.create());
-            }
-
-            return substringNode.executeSubstring(rope, byteOffset, byteLength);
-        }
     }
 
     public abstract static class MatchNode extends RubyContextNode {

src/main/java/org/truffleruby/core/rope/Bytes.java

@@ -10,9 +10,17 @@
 package org.truffleruby.core.rope;
 
 import com.oracle.truffle.api.CompilerDirectives.ValueType;
+import com.oracle.truffle.api.dsl.Cached;
+import com.oracle.truffle.api.interop.InteropLibrary;
+import com.oracle.truffle.api.interop.InvalidArrayIndexException;
+import com.oracle.truffle.api.interop.TruffleObject;
+import com.oracle.truffle.api.library.ExportLibrary;
+import com.oracle.truffle.api.library.ExportMessage;
+import com.oracle.truffle.api.profiles.BranchProfile;
 
 @ValueType
-public final class Bytes {
+@ExportLibrary(InteropLibrary.class)
+public final class Bytes implements TruffleObject {
     public final byte[] array;
     public final int offset;
     public final int length;
@@ -70,4 +78,32 @@ public Bytes clampedRange(int start, int end) {
     public byte get(int i) {
         return array[offset + i];
     }
+
+    // region Array messages for TRegex
+    @ExportMessage
+    public boolean hasArrayElements() {
+        return true;
+    }
+
+    @ExportMessage
+    public long getArraySize() {
+        return length;
+    }
+
+    @ExportMessage
+    public Object readArrayElement(long index,
+            @Cached BranchProfile errorProfile) throws InvalidArrayIndexException {
+        if (isArrayElementReadable(index)) {
+            return get((int) index);
+        } else {
+            errorProfile.enter();
+            throw InvalidArrayIndexException.create(index);
+        }
+    }
+
+    @ExportMessage
+    public boolean isArrayElementReadable(long index) {
+        return index >= 0 && index < length;
+    }
+    // endregion
 }