It's flattering but not very accurate, I'm afraid. I'd say it's still clearly an interpreted/scripting language which uses machine code as an optimization among several others. Afterall, the machine code is essentially only a list of calls to opcode functions and not inlined code that does the operations themselves (with a few exceptions for really trivial opcodes).

The runtime typing will keep it that way. It's no use inlining the opcode functions when each has to do complete type checking on every argument - it'd only be large code duplication to save a function call and maybe some stack handling for the arguments. In theory it's possible to do better in areas where the programmer has used strong compile-time typing (no mixed, no int|string etc) but it'd be a lot of work to implement it. It'd also require that we introduce some sort of smallint type since a normal int can be a native integer or a bignum object which requires runtime type checking in itself.

Doesn't Java try to fill that gap? It's strongly typed and designed to be possible to optimize to machine code. Are there still no tools that can do that successfully?

/ Martin Stjernholm, Roxen IS

Previous text:

2003-10-25 00:55: Subject: pike fills a gap?

---

hi,

recently when describing pike i wrote the following:

pike is a classical all purpose language, which contrary to php or perl was designed from the beginning to write large complex applications. pike is rather fast because it is optimized strongly for the individual machines. (the produced bytecode is machine dependant, (contrary to eg python and java which produce machine independant bytecode, and are therefore slower))

(i hope that is correct)

interresting was a reply that said:

ah. this language closes the gap between compilerlanguages and unoptimized interpreted languages.

i have never looked at pike from this perspective sounds like an interresting observation.

what do you think?

greetings, martin.

/ Brevbäraren