Read on Twitter
A neat JVM trick:
By the spec, dereferencing null has to throw a NullPointerException. So the JVM has to explicitly check for null on every access, right?
It doesn't, in a pretty amazing way.
By the spec, dereferencing null has to throw a NullPointerException. So the JVM has to explicitly check for null on every access, right?
It doesn't, in a pretty amazing way.
When the JVM compiles the byte code to machine code, it elides this check. So when the code actually tries to dereference a null object, it will trigger a SIGSEGV, just like a C program would do.
Now unlike a C program, the JVM will catch this SIGSEGV, look at the instruction that caused the segfault, restore control flow, and generate and throw a NullPointerException.
So this makes the happy path fast. But segfaults are incredibly slow, and if this null access is frequent, it'd a lot slower than the explicit null checks, wouldn't it?
Yes. And if this happens, the JVM will recompile the corresponding method at runtime.
Yes. And if this happens, the JVM will recompile the corresponding method at runtime.
You don't have to like #Java; but I think seriously believe the JVM is one of the finest pieces of software engineering.
