doing..? hacking cell phone software in Lua? I'd love to have at some embedded stuff with a non-c language. So much less to worry about screwing up. Although a lot of embedded stuff is hardware level stuff that has to be done in c anyway, so no lua there.
yes. Maybe we will release the interpreter part open source., but it is not done yet. Basically a J2ME and C++ (for brew and palm), interpreter for lua.
The whole idea, is that you can build an app once in Lua, and run it everywhere (for reals). This is a company thing, (we are not a small startup), so I am not sure how much of it will be realeased open source.
Still, how do you get around the fundamental problem of symbol lookup? Maybe sometimes you can infer the type, but I doubt that it is possible in general.
If the symbols are "hard linked", it is not dynamic anymore.
I wonder what strategy the LISP dialects out there typically use? I was impressed how easy it is to create an object system in a LISP, but somewhat dismayed because all examples just used (cond) for symbol lookup. Maybe because they were just academic examples, or perhaps they really don't care about performance at all?
> What about the things you pass into the function? You can't know their properties in advance, in general, or can you?
You can know that, e.g. a function passed in as an argument can be dereferenced as a pointer stored in a specific offset on the stack.
If your typechecks are compiler-generated statements inserted before a function call, then if you know at compile-time (via declarations or type inference, or type propagation) that a certain variable (to be passed as an argument) will always be of the correct type T, then you can omit the typecheck, as well as boxing/tagging.
You can get really close to the metal this way, and still have your language be as dynamic as can be.
Does anyone here have a single example of a performance problem caused by using a 'dynamic' language rather than a 'static' language? The only performance problems I've ever encountered 'in the wild' have been down to things like database access, I/O, bad sorting algorithms, string concatenation, etc.
Fundamental limitations of Python are responsible for most of the memory footprint and performance difference between the old official Python implementation of BitTorrent and C++ implementations like libtorrent. The kind of obsessive optimization that made uTorrent so tiny and fast is just not achievable in Python, due to things like per-object memory overhead and reference counting.
I've also done audio processing projects in Python, and ended up switching to D because Python couldn't do floating point additions and multiplications across large lists fast enough.
Yes. You are thinking of dynamic langauges merely on the function of being used on web apps. Embeded programming, programs in mobile applications, Javascript, Actionscript in flash.... you name it. Performance is important, especially on applications that are intesive (computing wise), and not i/o wise.
Eg. Games: A lot of games use something like Lua for scripting their UI, but for the hardcore game engine they use C or C++
Yes. A couple of years ago, I was implementing the Quadratic Sieve integer factorization algorithm in Maple (technically, I was implementing PP-SIQS). Taking a routine which used the Pollard rho algorithm to factor 12-16 digit values from MAPLE (a dynamically typed language) and implementing it in C -- using calls to exactly the same GMP primitive functions as Maple was using -- resulted in a tenfold speedup; this routine was used commonly enough that the overall routine was sped up by roughly a factor of two.
Thanks for the link, that is an interesting language. It seems like it's mostly a lot of C# features / programming style with python syntax and optional dynamic typing. Generally a good idea, though the code samples didn't feel right to me - lack of punctuation, perhaps? Maybe I've just gotten used to symbol-filled code.
