String.getBytes()比Charset.encode更快？二者适用场景与性能疑问

Hey there, let's break down your code examples, test results, and the technical questions you have:

用法示例 #

• String.getBytes 用法：

  return "testingtestingtestingtesting".getBytes(StandardCharsets.UTF_8);
  

• Charset.encode 用法：

  Charset charset = Charset.forName(StandardCharsets.UTF_8.toString());
  return charset.encode("testingtestingtestingtestingtesting").array();
  

测试结果 #

From your test data: the getBytes() method took 1328ms, while charset.encode() took 1804ms. The quick test clearly shows the former is faster.

Technical Answers #

1. Why is String.getBytes() faster than Charset.encode? #

The key lies in the overhead differences of their underlying implementations:

• String.getBytes(Charset) leverages the JVM's optimized path for string encoding directly. It operates on the internal char array of the String, skipping the steps of creating intermediate objects like ByteBuffer and CharBuffer that Charset.encode requires, thus reducing extra overhead from object allocation and memory copying.
• Charset.encode is a universal entry point for charset encoding, which needs to handle more complex scenarios (such as supporting arbitrary CharBuffer input, buffer expansion during encoding, etc.). These general-purpose logics bring additional performance costs. Plus, calling array() to get the byte array from ByteBuffer in your example may trigger an extra memory copy (if the ByteBuffer isn't direct memory), further increasing the time consumption.

2. What are the decision factors for choosing between the two? #

You can judge from these dimensions:

• Performance priority: If you only need to get a byte array directly and have high performance requirements, String.getBytes(Charset) is the better choice.
• Data carrier needs: If you need the encoding result as a ByteBuffer (e.g., in NIO scenarios), Charset.encode is more suitable because it returns ByteBuffer directly, avoiding extra array conversion steps.
• Reusability: If you need to use the same charset for encoding multiple times, you can cache the Charset instance in advance (like in your example), but getBytes still has a performance advantage in single encoding.
• Exception handling: String.getBytes wraps encoding exceptions as UncheckedIOException, while Charset.encode doesn't throw checked exceptions. If you need more fine-grained exception control, you may need to consider the latter's handling method.
• Complex scenario support: If you need to handle non-String character inputs (such as CharBuffer or char arrays), Charset.encode is the only option.

内容的提问来源于stack exchange，提问作者M. Shashanka