Thread: is this simple program correct? K&R 1-12

Originally Posted by fsx

Is this solution acceptable? If not, why?

Since your program is not a correct solution, it is especially not acceptable. Re-read the exercise: "Write a program that prints its input one word per line." Your program occasionally prints empty lines, i.e. lines that don't contain a word (and certainly K&R weren't thinking about the empty word here).

Try to get rid of these empty lines, e.g. by introducing two states "last character was a whitespace" and "last character wasn't". Also have a look at the isspace() function/macro.

mine seems a lot easier than the other, even if sometimes it prints too many newlines.

If you allow your program to occasionally produce incorrect results, here's an even simpler version:

Code:

int main() {}

;-),
Philip

Originally Posted by fsx

Code:

	while((c = getchar()) != EOF)
	{
		if((c == ' ') || (c == '\t'))
			putchar('\n');

What you really want is a "nop" (no-op, no operation) and to continue --

Code:

		if((c == ' ') || (c == '\t'))
			continue;

Except you still need a line break! So maybe

Code:

#include <stdio.h>

int main(void)
{
	int c;
	char flag=0;

	while((c = getchar()) != EOF)
	{
		if((c == ' ') || (c == '\t'))
			if (flag) continue;
			else { putchar('\n');
				flag=1; }
		else {	putchar(c);
			flag=0; }
	}
	return 0;
}

If you are just on chapter 1 you may not know about "continue" (it just starts the loop again without doing anything else).

The use of a flag is a very handy, basic, chapter 1 kind of thing, so if you don't understand what that's about, try to, or ask and I'll explain.

Basically now you could type:

Code:

this       is     a   sentance   with  extra      spaces
Output:
this
is
a
sentence
with
extra
spaces

That was the idea that Snafuist introduced at the beginning. I think you could come up with a better name than "flag" for that variable, though. Maybe something like "saw_whitespace". There's also a logical bug in the program (which was present in my first version too): since the flag is initially zero, there will be a blank line at the beginning of the output if the first character is whitespace. Also, continue isn't really necessary.

Anyway, here's my shot at it.

Code:

#include <stdio.h>
#include <ctype.h>

int main() {
    int c, between_words = 1;

    while((c = getchar()) != EOF) {
        if(isspace(c)) {
            if(!between_words) {
                putchar('\n');
                between_words = 1;
            }
        }
        else {
            putchar(c);
            between_words = 0;
        }
    }

    return 0;
}

Note the use of isspace() from ctype.h. isspace [C++ Reference]

What is not working?

I would expect that you will need a nested if-statement:

Code:

   if (c == ' ' || c == '\t')
   {
     if (pc != c)
       putchar('\n');
   }
   else 
       putchar(c);

This will prevent multiple spaces from giving multiple newlines.

--
Mats

Because you do not want the ELSE part to be executed if the second if is false - only if the first if is false.

You could of course add a "if (pc != c)" inside the else-branch instead, but that's less logical.

--
Mats

Originally Posted by dwks

That was the idea that Snafuist introduced at the beginning. I think you could come up with a better name than "flag" for that variable, though.

I called it "flag" because it is a (boolean) flag and I wanted the OP (fsx) to pick up on that (Snafuist refers to this as "two states", which is not quite as specific -- I guess you could see a boolean flag as an implementation of a two-state concept).

You're right about the continue being unnecessary and the initial whitespace issue.

Here's a related question: I usually use a char for boolean values cause it's smaller, but would it be true that an int is actually faster for the processor because it has a four byte boundary? Sort of nit-picking, but this could have a lot of significance in the right place...

Originally Posted by MK27

Here's a related question: I usually use a char for boolean values cause it's smaller, but would it be true that an int is actually faster for the processor because it has a four byte boundary? Sort of nit-picking, but this could have a lot of significance in the right place...

yes it could. you need profiling to prove it or opposite (when the cache size is critical issue)

Originally Posted by MK27

Here's a related question: I usually use a char for boolean values cause it's smaller, but would it be true that an int is actually faster for the processor because it has a four byte boundary? Sort of nit-picking, but this could have a lot of significance in the right place...

Can't bit fields be used for this (making the int size smaller) ? Or is the struct making it larger again?

Code:

#include<stdio.h>

struct foo
{
	int flag:2;
} bar;

int main(void)
{
	bar.flag = 1;
	printf("%d", bar.flag);
	
	return 0;
}

Bitfields won't actually shrink the structure smaller than the actual type. That is to say, even if you set up a bit field of :2, it's still going to be at least an int in size.

Quzah.