single character input

[Peter J Desnoyers]

In article <3423 at drivax.UUCP> braun at drivax.UUCP (Kral) writes:
>... what I consider to be rather odd
>behaviour on the part of scanf.  For instance, if I do the following:
>	char	c ;
>	puts(prompt) ;
>	scanf("%c ", c) ;
>	switch(c) {
>	 ... }
>The program ends up reading one line behind what is input; in other words, I
>have to enter something (anything!) the first time, then the second line
>entered triggers the program reading the first line, etc.  Not at all what I
>expected, and it seems not too useful to me.  I got so hung up on scanf that
>I couldn't think of anything else.  I wonder why scanf does this?  Am I missing
>something equally obvious here?

Someone here made the mistake of changing one of the intro programming
courses at MIT from pascal (I think) to C, and at least one person I know
got caught by this same bug. They wrote:

  for (...) { printf( "y/n:"); scanf( "%c", &answer); ...}

Unfortunately, what they got was:
    value read:		input typed:
       'y'		    'y\n'	(note - scanf will not return until
       '\n'		    nothing      you hit return)

and so forth. Since they were testing only against 'y', they never noticed
that the non-y value wasn't a 'n', and got seriously confused.

Moral - the interaction of character-at-a-time reads through scanf or
whatever and buffered IO in the TTY driver is non-intuitive. Remember
that you are going to get every character typed at the program, but a
line at a time. 

				Peter Desnoyers
				peter at athena.mit.edu

Please, no flames about C not being appropriate to an intro programming
course. (It was 1.00 or 2.10 - taught by either civ. e. or mech. e.)
You don't have a good idea of how the computer (not the language, the
computer) works until the end of a course - C requires this knowledge
in the beginning. Besides, if you know {C, Pascal} you know {Pascal, C}.
They're all Algol derivatives, anyway.