Gary Willoughby Gary Willoughby - 4 months ago 19
Brainfuck Question

Creating a Brainfuck parser, whats the best method of parsing loop operators?

I'm creating a Brainfuck parser (in a BASIC dialect) ultimately to create an interpreter but i've realise it's not as straight forward as i first thought. My problem is that i need a way to accurately parse the matching loop operators within a Brainfuck program. This is an example program:

,>,>++++++++[<------<------>>-]
<<[>[>+>+<<-]>>[<<+>>-]<<<-]
>>>++++++[<++++++++>-],<.>.


'[' = start of loop

']' = end of loop

I need to record the start and end point of each matching loop operator so i can jump around the source as needed. Some loops are alone, some are nested.

What would be the best way to parse this? I was thinking maybe move through the source file creating a 2D array (or such like) recording the start and end positions of each matching operator, but this seems like a lot of 'to'ing and fro'ing' through the source. Is this the best way to do it?

More info: Brainfuck homepage

EDIT: Sample code in any language greatly appreciated.

Answer

Have you considered using a Stack data structure to record "jump points" (i.e. the location of the instruction pointer).

So basically, every time you encounter a "[" you push the current location of the instruction pointer on this stack. Whenever you encounter a "]" you reset the instruction pointer to the value that's currently on the top of the stack. When a loop is complete, you pop it off the stack.

Here is an example in C++ with 100 memory cells. The code handles nested loops recursively and although it is not refined it should illustrate the concepts..

char cells[100] = {0};   // define 100 memory cells
char* cell = cells;      // set memory pointer to first cell
char* ip = 0;            // define variable used as "instruction pointer"

void interpret(static char* program, int* stack, int sp)
{
    int tmp;
    if(ip == 0)              // if the instruction pointer hasn't been initialized
        ip = program;        //  now would be a good time

    while(*ip)               // this runs for as long as there is valid brainF**k 'code'
    {
        if(*ip == ',')
            *cell = getch();
        else if(*ip == '.')
            putch(*cell);
        else if(*ip == '>')
            cell++;
        else if(*ip == '<')
            cell--;
        else if(*ip == '+')
            *cell = *cell + 1;
        else if(*ip == '-')
            *cell = *cell - 1;
        else if(*ip == '[')
        {   		
            stack[sp+1] = ip - program;
            *ip++;
            while(*cell != 0)
            {
                interpret(program, stack, sp + 1);
            }
            tmp = sp + 1;
            while((tmp >= (sp + 1)) || *ip != ']')
            {
                *ip++;
                if(*ip == '[')
                    stack[++tmp] = ip - program;
                else if(*ip == ']')
                    tmp--;
            }   		
        }
        else if(*ip == ']')
        {
            ip = program + stack[sp] + 1;
            break;
        }
        *ip++;       // advance instruction
    }
}

int _tmain(int argc, _TCHAR* argv[])
{   
    int stack[100] = {0};  // use a stack of 100 levels, modeled using a simple array
    interpret(",>,>++++++++[<------<------>>-]<<[>[>+>+<<-]>>[<<+>>-]<<<-]>>>++++++[<++++++++>-],<.>.", stack, 0);
    return 0;
}

EDIT I just went over the code again and I realized there was a bug in the while loop that would 'skip' parsed loops if the value of the pointer is 0. This is where I made the change:

while((tmp >= (sp + 1)) || *ip != ']')     // the bug was tmp > (sp + 1)
{
ip++;
if(*ip == '[')
    stack[++tmp] = ip - program;
else if(*ip == ']')
    tmp--;
}

Below is an implementation of the same parser but without using recursion:

char cells[100] = {0};
void interpret(static char* program)
{
    int cnt;               // cnt is a counter that is going to be used
                           //     only when parsing 0-loops
    int stack[100] = {0};  // create a stack, 100 levels deep - modeled
                           //     using a simple array - and initialized to 0
    int sp = 0;            // sp is going to be used as a 'stack pointer'
    char* ip = program;    // ip is going to be used as instruction pointer
                           //    and it is initialized at the beginning or program
    char* cell = cells;    // cell is the pointer to the 'current' memory cell
                           //      and as such, it is initialized to the first
                           //      memory cell

    while(*ip)             // as long as ip point to 'valid code' keep going
    {
        if(*ip == ',')
            *cell = getch();
        else if(*ip == '.')
            putch(*cell);
        else if(*ip == '>')
            cell++;
        else if(*ip == '<')
            cell--;
        else if(*ip == '+')
            *cell = *cell + 1;
        else if(*ip == '-')
            *cell = *cell - 1;
        else if(*ip == '[')
        {   		
            if(stack[sp] != ip - program)
                stack[++sp] = ip - program;

            *ip++;

            if(*cell != 0)
                continue;
            else
            {   				
                cnt = 1;
                while((cnt > 0) || *ip != ']')
                {
                    *ip++;
                    if(*ip == '[')
                    cnt++;
                    else if(*ip == ']')
                    cnt--;
                }
                sp--;
            }
        }else if(*ip == ']')
        {   			
            ip = program + stack[sp];
            continue;
        }
        *ip++;
    }
}

int _tmain(int argc, _TCHAR* argv[])
{   
    // define our program code here..
    char *prg = ",>++++++[<-------->-],[<+>-]<.";

    interpret(prg);
    return 0;
}