[06] Create benchmark for String::split

Nicolai Parlog · Nicolai Parlog · commit 8105b4cb3760 · 2019-05-24T18:36:30.000+02:00
diff --git a/src/main/java/org/codefx/demo/effective_java/_06_unnecessary_objects/StringSplit.java b/src/main/java/org/codefx/demo/effective_java/_06_unnecessary_objects/StringSplit.java
@@ -0,0 +1,71 @@
+package org.codefx.demo.effective_java._06_unnecessary_objects;
+
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.Param;
+
+import java.util.ArrayList;
+
+public class StringSplit extends ObjectCreationBenchmarks {
+
+	@Param({ "foobar", "foo:bar", "foo::bar" })
+	private String string = "foobar";
+
+	@Benchmark
+	public String[] singleColonSplit() {
+		// NOTE unnecessary object:
+		//   For a single-character, `split` employs an optimization
+		//   and does not create a reg ex pattern
+		return string.split(":");
+	}
+
+	@Benchmark
+	public String[] singleColonCode() {
+		// this code is unlikely to be 100% correct (it is basically untested),
+		// so I assume a proper implementation would potentially be a little slower,
+		// but surely in the same order of magnitude
+		ArrayList<String> matches = new ArrayList<>();
+		int lastIndex = 0;
+		int nextIndex = string.indexOf(':');
+		while (nextIndex > -1) {
+			if (nextIndex > lastIndex + 1)
+				matches.add(string.substring(lastIndex, nextIndex));
+			lastIndex = nextIndex + 1;
+			nextIndex = string.indexOf(':', nextIndex + 1);
+		}
+		if (string.length() > lastIndex)
+			matches.add(string.substring(lastIndex));
+
+		return matches.toArray(String[]::new);
+	}
+
+	@Benchmark
+	public String[] doubleColonSplit() {
+		// NOTE unnecessary object:
+		//   Beyond a single character, `split` will create a create a reg ex pattern,
+		//   which makes it rather slow
+		return string.split("::");
+	}
+
+	@Benchmark
+	public String[] doubleColonCode() {
+		// this code is unlikely to be 100% correct (it is basically untested),
+		// so I assume a proper implementation would potentially be a little slower,
+		// but surely in the same order of magnitude
+		ArrayList<String> matches = new ArrayList<>();
+		int lastIndex = 0;
+		int nextIndex = string.indexOf("::");
+		while (nextIndex > -1) {
+			if (nextIndex > lastIndex + 2)
+				matches.add(string.substring(lastIndex, nextIndex));
+			lastIndex = nextIndex + 2;
+			nextIndex = string.indexOf("::", nextIndex + 2);
+		}
+		if (string.length() > lastIndex)
+			matches.add(string.substring(lastIndex));
+
+		return matches.toArray(String[]::new);
+	}
+
+
+
+}
