Parsing an equation

[pheonixkid]

I'm trying to parse equations like these which only has two values or the square root of a certain value from a text file:
100+100
-100-100
-(100)+(-100)
sqrt(100)
by the minues signs, parenthesis and the operator symbol in the middle and the square root, and i have no idea how to start off...i've got the file part done and the simple calculation parts except that i couldnt get my program to solve the equations in the above.

char *tok = buff;
        sscanf(tok,"%lf%c%lf",&num1,&sym,&num2);

that is what have i written for the other basic operations without parenthesis and the minus sign for the second value and it works great for the simple ones. i'm using a switch method to calculate the add,sub,mul and divide but i'm not sure how to properly use the sscanf function (if i am not using it properly) or if there is another way using a function like strtok to properly parse the parenthesis and the minus signs. any kind help?

[David W]

100+100
-100-100
-(100)+(-100)
sqrt(100)

The last line above could be tested / handled by:
if "sqrt(" 
    then if  int  
         then if ")" 
            then process ...


The next to last line could be tested / handled by:
else if "+(" or "-("
  then if int   
       then if ")"    
           then if valid_binary_op
              then if int or (int)
                 then process

The first two lines above both could be handled as:
else if int 
   then if valid_binary_op 
      then if int
         then process ...

else ... data input error, that input format not handled here

[L7Sqr]

Generally, when you have to parse input that must adhere to some constraints (a language, for example) you use a grammar. The grammar can be built up in a variety of ways: using tools like yacc/bison or in code ala llvm. In the former, you usually have a parser generator that will create parser code for you from the grammar (lex/flex provides this role for yacc).

In this case, where the input is simple enough, I'd go with an in-code model. Write a routine that takes a line and returns the tokens of the line:

int tokenize (const char * line, 
              std::vector< std::pair< const char *, token_type > > &tokens) {
   // walk the input line here looking for tokens
   // for example... line = "(123)"
   tokens.push_back (std::make_pair (&line[0], TOKEN_LPAREN))
   tokens.push_back (std::make_pair (&line[1], TOKEN_NUMBER))
   tokens.push_back (std::make_pair (&line[4], TOKEN_RPAREN))
   return 3;
}

Of course, that is incomplete: you would need to do the string walking and provide the token enum. But one you have that you can easily iterate the token vector and do your processing.

[pheonixkid]

thanks! but i'm not familiar with C++ language. i'm guessing that i have to use 'strtok' in C using delimiters of the parenthesis. but the number values that come could be any length meaning that i would have to find the length of each of the value first for such method?

[ddanbe]

See what you can do with the isdigit() function.

[L7Sqr]

In addition to the isdigit check, you can also use strtol which will read as many digits as are valid and return a pointer to the next character in the string.

I provided a Code Snippet that uses this exact technique (although doesn't account for things like sqrt). Look at the ShuntingYard::convert method. It's written in C++ and uses strtod but the parsing loop could be used as C code and strtol could be used instead.