Question

I need to validate that a text file is in CSV format (i.e. that each digit is separated by a comma).

From reading online, it seems that people have conflicting views about it – but is Strtok() the best way to do this?

Any help would be great.

Answer 1

Your input seems so easy that I would probably just use a loop around fgetc(3); I’ll sketch some pseudo-code here:

fd = fopen("file", "r");
int c;
while((c=fgetc(fd)) != EOF) {
    switch(c) {
        case '0':
        case '1':
        /* so on */
        case '9':
            handle_digit(c);
            break;
        case ',':
            handle_comma();
            break;
        case '\n':
            handle_newline();
            break;
        default:
            fprintf(stderr, "mistaken input %c\n", c);
            break;
    }
}
fclose(fd);

You’ll have to manage the input in the functions in a manner that may be a bit awkward if you’re used to higher-level languages such as Ruby or Python where you’d just run line.split(',') to get a list of numbers, but that is pretty idiomatic C.

Of course, if this were a real problem, I’d probably prefer flex and bison, and write a tiny lexer and grammar, mostly because it would be a lot easier to extend in the future as needs change.

---

Update

With some additional criteria to check, the handle_{digit,comma,newline}() routines are easier to sketch. I’ll sketch using global variables, but you could just as easily stuff these into a struct and pass them around from function to function:

enum seen {
    NEWLINE,
    COMMA,
    DIGIT,
};

enum seen last_seen = NEWLINE;

handle_digit(int c) {
    if (last_seen == DIGIT) {
        /* error if numbers cannot have multiple digits
           or construct a larger number if numbers can have
           multiple digits */
    } else if (last_seen == COMMA || last_seen == NEWLINE) {
        /* start a new entry */
    }
    last_seen = DIGIT;
}

handle_comma() {
    if (last_seen == COMMA) {
        /* error */
    } else if (last_seen == NEWLINE) {
        /* error */
    } else if (last_seen == DIGIT) {
        /* end previous field */
    }
    last_seen = COMMA;
}

handle_newline() {
    if (last_seen == NEWLINE) {
        /* error */
    } else if (last_seen == COMMA) {
        /* error */
    } else if (last_seen == DIGIT) {
        /* end previous field */
    }
    last_seen = NEWLINE;
}

Add whichever checks you need to validate the contents according to whichever rules you have. You might wish to standardize the order and contents of the tests to ensure that you never forget one, even if it means you write a /* nop */ comment once or twice to remind yourself that something is fine.