编码和解码后字符串有所不同

Question

I stumbled across weird behaviour of encoding/decoding string. Have a look at an example:

@Test
public void testEncoding() {
    String str = "\uDD71"; // {56689}
    byte[] utf16 = str.getBytes(StandardCharsets.UTF_16); // {-2, -1, -1, -3}
    String utf16String = new String(utf16, StandardCharsets.UTF_16); // {65533}
    assertEquals(str, utf16String);
}

I would assume this test will pass, but it is not the case. Could someone explain why the encoded and decoded string is not equal to the original one?

Answer 1

U+DD71 is not a valid codepoint, as U+D800..U+DFFF are reserved by Unicode so as not to cause confusion with UTF-16. As such, these codepoints should never appear as valid character data. From the Unicode standard:

Isolated surrogate code points have no interpretation; consequently, no character code charts or names lists are provided for this range.

This works, though:

@Test
public void testEncoding() {
    String str = "\u0040";
    byte[] utf16 = str.getBytes(StandardCharsets.UTF_16);
    String utf16String = new String(utf16, StandardCharsets.UTF_16);
    assertEquals(str, utf16String);
}

So, it's not your code at fault, but that you're trying to use a codepoint that isn't valid.