Java Performance, 2nd Edition

Chapter 12. Java SE API Tips

This chapter covers areas of the Java SE API that have implementation quirks affecting their performance. Many such implementation details exist throughout the JDK; these are the areas where I consistently uncover performance issues (even in my own code). This chapter includes details on the best way to handle strings (and especially duplicate strings); ways to properly buffer I/O; classloading and ways to improve startup of applications that use a lot of classes; proper use of collections; and JDK 8 features like lambdas and streams.

Strings

Strings are (unsurprisingly) the most common Java object. In this section, we’ll look at a variety of ways to handle all the memory consumed by string objects; these techniques can often significantly reduce the amount of heap your program requires. We’ll also cover a new JDK 11 feature of strings involving concatenation.

Compact Strings

In Java 8, all strings are encoded as arrays of 16-bit characters, regardless of the encoding of the string. This is wasteful: most Western locales can encode strings into 8-bit byte arrays, and even in a locale that requires 16 bits for all characters, strings like program constants often can be encoded as 8-bit bytes.

In Java 11, strings are encoded as arrays of 8-bit bytes unless they explicitly need 16-bit characters; these strings are known as compact strings. A similar (experimental) feature in Java 6 was known as compressed strings; compact strings are conceptually the ...