Thread: Enumerators vs Magic no.s

I understand that it is not is good idea to use magic no.s ..but just take a look at the situation..
I need to describe a token..
A

Code:

int type[5]

would be sufficient for containing all the data..whether it is a number...or an operator....what type of operator...no of operands...etc etc....but it would scatter seemingly meaningless

Code:

if((x[0]==3)&&(x[2]!=0)) {;}

^ like statements through the whole code..
So, is it a better Idea to use 4 or 5 different enumerators?..
&& Also ..is there a hidden difference in performance..?..although I know that enums are basically numbers...still there could be something..!

Sorry, I can't think of a good way to implement the struct...
Could you show a little example..?

Do you not know what a struct is? It's a record type. You would learn the syntax for using struct and then make all these things into fields:

a number...or an operator....what type of operator...no of operands

Cprogramming.com Tutorial: Structures

Further, that struct may or may not need to be part of a larger class type in your C++ program, so read

Cprogramming.com Tutorial: Classes

also.

..I know very well what classes and structures are..but what difference does using them here make?
Placing an object of

Code:

class Type
{
     int n;
     int n_o;
     int pre;
     //&& some more ints..
};

inside the token class..
brings the same problem with magic no.s...as with arrays....only somewhat better in specifying where to get what data...but the data is still in numbers...
Is there a better way to use them here?

No, now the data is in names. Instead of talking about foo[2], which means who knows what, now you have foo.number_of_operands and everybody is a lot happier to see that sort of thing in code.

...consider that each mathematical function would still have a corresponding number(say 'foo.f_no') ...Is there a way to put names to those...?..and call the functions directly from the names?

...I didn't think of that... I could easily put a std::string inside the Type class to denote the name of the functions...and '==' them when required..!

we're storing function objects

Please elaborate...I tried to implement it in that way(some months ago)..and after that 'pair classes'...but was totally pwned by the complexity .

Well, I had thought that there was an overarching class above unary_function and binary_function in the <functional> header, but I guess not. Still you can run with those, especially if you have a majority of one or the other.

Either way, you can always roll your own with something like

Code:

struct Functor {
    int no_of_arguments;
};

and then you inherit dozens of subclasses off of this one, each one with its own operator().

If you're still doing arithmetic from the other thread, then one advantage of the <functional> struct is that plus, multiply, etc., are already there for the taking.