Programming language where every expression makes sense

Question Detail: 

Per recommendation I am reposting this from Stack Overflow.

Recently I have been thinking about following issue.

Consider the code for a standard "Hello world!" program:

main() {     printf("Hello World");  } 

Now almost any change in this code will make it completely useless, in fact almost every change will prevent the code from compiling. For example:

main(5 {     printf("Hello World");  } 

Now to the actual question. Is there a programming language where every possible combination of symbols - that is, every expression - makes sense? I tried thinking about some sort of solution and came up with two:

1. Postfix with a limited number of variables. Essentially all variables are already defined before you write any code and you have to work just with them. Theoretically you can than perform an arbitrary number of operations by forming a chain of many simple programs, each one of them feeding results to others. Code could be written as a series of characters in postfix notation;

2. "Postfix" with a stack of variables. Variables are stored on a stack; every operation takes two variables from the top and puts the result in their place. The program ends when it reaches the last operation or variable.

Personally I hate both of these. Not only are they limited, they are inelegant. They are not even real solutions, more like workarounds, essentially "offshoring" some work to an external process.

Does anyone have any other idea how to solve this problem?

Asked By : user1561358

Best Answer from StackOverflow

Question Source : http://cs.stackexchange.com/questions/48574

Answered By : Cort Ammon

Redcode, the assembly language behind codewars, was explicitly written to have very few halting instructions, because the code often gets mangled before it finally gives out, and the more opportunities it has to halt, the less interesting the game is.

You see very few such languages in practice because we don't just want a program to run, we want it to run in the way we expect. If you can make a typo and change the way the program ran, it must be acceptably close to the original expected behavior, or the programmers with seethe in frustration.

There is some precedence for such things by using natural languages rather than formal languages, but it's not what I would call a large field when you compare it to the use of formal languages. If you're interested in such programming languages, the natural language processing community is where I'd look.

Another field you could look at is genetics. There are remarkably few genetic sequences which are simply invalid. Plenty of them that aren't very effective at reproductions, but very few invalid ones.