Lisp

An argument for Lisp from the founder of Viaweb, what eventually became Yahoo! Stores. He says that when you're looking up the power continuum of programming languages you tend to see those more powerful languages as weird. But he's making the assumption that Lisp is the most powerful of programming languages. What if it's not? What if he's just viewing those "other" languages as weird? He also attributes a lot of his success to Lisp when maybe it was more a combination of luck and talent. via Slashdot

That was interesting

I read the article and then I read the responses (3+) on Slashdot.

My first response, about 20% into the article, was the obvious problem of finding Lisp programmers. Assuming Lisp is better than everything else people have to get proficient at it. There are probably 20 to 1 or worse good C/C++ programmers as there are Lisp programmers so finding your staff for your startup is going to be 20 times harder.

That's a good arguement for using C/C++ I think. Part of making a startup work is finding the people. That he didn't need to do that just means his startup was small enough, and he was lucky enough, that just the people that started the company were enough. Had he needed to search for employees I think he would have quickly found this the biggest weak point.

I will say that having learned a little bit about lisp from working with something similar at Naughty Dog for 6 months was very informative. The macros do in some ways kick big time ass as they let you build the language to suit your needs.

The problem with that is that it can also mean the language quickly gets complicated because instead of having just say C/C++ and a bunch of functions you now have actually new language structures that have to be learned by all the other programmers on your team. Worse, if there is no standard design to these language additions it becomes very confusing. Is is "A=B+C" or "B+C=A" for example. Or maybe sometimes it's "AequalsB+C" because some other programmer didn't know that there was already and "=" language addition.

I suppose that some of that is true in any language. In C you could have both strdup and dupstr and you could have strcpy (dst,str) and cpystr(src,dst) but my experience was that lisp tends to make that much worse. for loops are macros. if you aren't careful, you might have a macro

(for i 1 10 (code))

which in C would be

for (i = 1; i <= 10; i++) code;

then, somebody makes a new for like macro like this

(forstep 2 0 20 i (code))

which in C would be

for (i = 0; i <= 20; i += 2) code;

That is avoided in C because nobody would likely make a new "forstep" macro.

Clearly though for Andy (who admitedly can keep the entire state of the program (or at least the parts he wrote) in his head it's very easy for him to see how it's all working and how to add a feature quickly. I wonder though if Lisp has the problem of not being very multi-programmer friendly or if that's just my experience. The point being that using some API in C seems easier to grok than learning the new language features that were added this week.

That was interesting was interesting

I haven't read the Slashdot stuff, but agree that a core issue in language choice is familiarity. And another core issue is efficiency. If you sacrifice these for language power you might get a bit more productivity from experts that already know the language but I think greater value comes from being able to hire inspired junior programmers and level script designers that can immediately start adding cool stuff in C or C++ or Python.

In games, a really big part of the problems to solve is control over time issues. This includes attaching reactions and content state changes to game objects based on time parameters or other triggers. I don't know what Lisp offers above other languages for this kind of code. I'm not sure I know of any language that offers advantages for this kind of code.

Just thought I'd join the Ex-ND party here. How are you guys doing? Send me an email...