From a4b5b1169aaf145996dc2a8248a10b81e45d57b7 Mon Sep 17 00:00:00 2001
From: Peter Shipton <Peter_Shipton@ca.ibm.com>
Date: Mon, 27 May 2024 15:11:05 -0400
Subject: [PATCH] jdk11 create compact Strings when possible from split(),
 substring()

Issue https://github.com/eclipse-openj9/openj9/issues/19543

Signed-off-by: Peter Shipton <Peter_Shipton@ca.ibm.com>
---
 .../share/classes/com/ibm/jit/JITHelpers.java | 16 +++++
 .../share/classes/java/lang/String.java       | 62 +++++++++++++------
 .../openj9/test/java/lang/Test_String.java    |  6 ++
 3 files changed, 64 insertions(+), 20 deletions(-)

diff --git a/jcl/src/java.base/share/classes/com/ibm/jit/JITHelpers.java b/jcl/src/java.base/share/classes/com/ibm/jit/JITHelpers.java
index 6c36e77528a..0528c3e96db 100644
--- a/jcl/src/java.base/share/classes/com/ibm/jit/JITHelpers.java
+++ b/jcl/src/java.base/share/classes/com/ibm/jit/JITHelpers.java
@@ -566,6 +566,22 @@ public char getCharFromArrayByIndex(Object obj, int index) {
 		}
 	}
 
+	public boolean canEncodeAsLatin1(byte[] array, int start, int length) {
+		int count = 0;
+		int index = start << 1;
+		if (!IS_BIG_ENDIAN) {
+			index += 1;
+		}
+		while (count < length) {
+			if (array[index] != 0) {
+				return false;
+			}
+			index += 2;
+			count += 1;
+		}
+		return true;
+	}
+
 	/**
 	 * Returns the first index of the target character array within the source character array starting from the specified
 	 * offset.
diff --git a/jcl/src/java.base/share/classes/java/lang/String.java b/jcl/src/java.base/share/classes/java/lang/String.java
index b5b04794883..c4da934a70f 100644
--- a/jcl/src/java.base/share/classes/java/lang/String.java
+++ b/jcl/src/java.base/share/classes/java/lang/String.java
@@ -824,19 +824,25 @@ public String(char[] data, int start, int length) {
 			} else {
 				char theChar = helpers.getCharFromArrayByIndex(data, start);
 
-				if (theChar <= 255) {
-					value = decompressedAsciiTable[theChar];
+				if (COMPACT_STRINGS && (theChar <= 255)) {
+					value = compressedAsciiTable[theChar];
+					coder = LATIN1;
+					hash = theChar;
 				} else {
-					value = new byte[2];
-
-					helpers.putCharInArrayByIndex(value, 0, theChar);
-				}
+					if (theChar <= 255) {
+						value = decompressedAsciiTable[theChar];
+					} else {
+						value = new byte[2];
+	
+						helpers.putCharInArrayByIndex(value, 0, theChar);
+					}
 
-				coder = UTF16;
-				hash = theChar;
+					coder = UTF16;
+					hash = theChar;
 
-				if (COMPACT_STRINGS) {
-					initCompressionFlag();
+					if (COMPACT_STRINGS) {
+						initCompressionFlag();
+					}
 				}
 			}
 		} else {
@@ -850,6 +856,11 @@ public String(char[] data, int start, int length) {
 				}
 
 				coder = LATIN1;
+			} else if (COMPACT_STRINGS && ((start != 0) || (data.length != length * 2)) && helpers.canEncodeAsLatin1(data, start, length)) {
+				value = new byte[length];
+				coder = LATIN1;
+
+				compress(data, start, value, 0, length);				
 			} else {
 				if (start == 0 && data.length == length * 2) {
 					value = data;
@@ -887,19 +898,25 @@ public String(char[] data, int start, int length) {
 			} else {
 				char theChar = helpers.getCharFromArrayByIndex(data, start);
 
-				if (theChar <= 255) {
-					value = decompressedAsciiTable[theChar];
+				if (COMPACT_STRINGS && (theChar <= 255)) {
+					value = compressedAsciiTable[theChar];
+					coder = LATIN1;
+					hash = theChar;
 				} else {
-					value = new byte[2];
-
-					helpers.putCharInArrayByIndex(value, 0, theChar);
-				}
+					if (theChar <= 255) {
+						value = decompressedAsciiTable[theChar];
+					} else {
+						value = new byte[2];
+	
+						helpers.putCharInArrayByIndex(value, 0, theChar);
+					}
 
-				coder = UTF16;
-				hash = theChar;
+					coder = UTF16;
+					hash = theChar;
 
-				if (COMPACT_STRINGS) {
-					initCompressionFlag();
+					if (COMPACT_STRINGS) {
+						initCompressionFlag();
+					}
 				}
 			}
 		} else {
@@ -913,6 +930,11 @@ public String(char[] data, int start, int length) {
 				}
 
 				coder = LATIN1;
+			} else if (COMPACT_STRINGS && ((start != 0) || (data.length != length * 2)) && helpers.canEncodeAsLatin1(data, start, length)) {
+				value = new byte[length];
+				coder = LATIN1;
+
+				compress(data, start, value, 0, length);				
 			} else {
 				if (sharingIsAllowed && start == 0 && data.length == length * 2) {
 					value = data;
diff --git a/test/functional/Java8andUp/src/org/openj9/test/java/lang/Test_String.java b/test/functional/Java8andUp/src/org/openj9/test/java/lang/Test_String.java
index f74132a0104..931634814c4 100644
--- a/test/functional/Java8andUp/src/org/openj9/test/java/lang/Test_String.java
+++ b/test/functional/Java8andUp/src/org/openj9/test/java/lang/Test_String.java
@@ -937,6 +937,12 @@ public void test_indexOf3() {
 		String needle = "\u00b0\u00b1";
 		String hay = new StringBuilder("a").append(needle).toString();
 		AssertJUnit.assertEquals("Failed to find string 3", 1, hay.indexOf(needle));
+
+		String multi = "\u0100:abc";
+		String[] splits = multi.split(":");
+		String sub = multi.substring(2);
+		AssertJUnit.assertEquals("Failed to find string 4", 3, "123abc".indexOf(splits[1]));
+		AssertJUnit.assertEquals("Failed to find string 5", 3, "123abc".indexOf(sub));
 	}
 
 	/**