On Mon, 26 Apr 2021 11:24:23 GMT, Сергей Цыпанов
<github.com+10835776+stsypa...@openjdk.org> wrote:
>> That should be fine, the null check in Objects.equals is benign with these
>> usages.
>
> One more thing I'm thinking about (not to be done in this PR of course) is to
> move call to `String.intern()` from where it is now in
> `Class.getPackageName()`
>
> public String getPackageName() {
> String pn = this.packageName;
> if (pn == null) {
> Class<?> c = this;
> while (c.isArray()) {
> c = c.getComponentType();
> }
> if (c.isPrimitive()) {
> pn = "java.lang";
> } else {
> String cn = c.getName();
> int dot = cn.lastIndexOf('.');
> pn = (dot != -1) ? cn.substring(0, dot).intern() : ""; // <---
> }
> this.packageName = pn;
> }
> return pn;
> }
>
> to `packageName` field assignement like
>
> this.packageName = pn.intern();
>
> this would add two more Strings (`""` and `"java.lang"`) into string table
> and allow to avoid `String.equals()` in favour of `==` call when comparing
> package names. What do you think?
String literals are implicitly interned:
System.out.println("" == new String("")); // false
System.out.println("" == new String("").intern()); // true
System.out.println("java.lang" == new String("java.lang")); // false
System.out.println("java.lang" == new String("java.lang").intern()); // true
So we could already use `==` instead of `equals` when evaluating the result of
`getPackageName`, which might be a good optimization in some places
(`ReflectionFactory::isSamePackage` shows up in profiles of some
reflection-related benchmarks).
-------------
PR: https://git.openjdk.java.net/jdk/pull/3571
