Ndtyjky

I'm preparing a simple program for my classes and found that scanf is not as strict as I expected:

int main()

{

  // Define variables

  float alpha;

  float b;

  float result;



  // Get user input

  printf("Type alpha and b. Space separated float numbers only.n");

  scanf("%f %f", &alpha, &b);

  printf("%f %fn", alpha, b);

  return 0;

}

Returns:

Type alpha and b. Space separated float numbers only.

a 0

-29768832.000000 0.000000

Where I can read about its behavior?

How can I make input strict with standard functions and throw an exception if there was neither ints nor floats?

P. S. I'm using gcc v5 in Ubuntu.

I'm preparing a simple program for my classes and found that scanf is
not as strict as I expected

No, you didn't.

Given this scanf call:

  float alpha;

  float b;



  scanf("%f %f", &alpha, &b);

... and this input:

a 0

... scanf has a matching failure when it encounters the 'a' while trying to scan a floating-point value and not having yet scanned any decimal digits. It stops at that point (and leaves the 'a' in the stream for a subsequent read), and, as always, it returns the number of input fields successfully scanned (0).

Where I can read about its behavior?

You could try running the command man scanf at the command line (Linux, OSX, ...), or entering the same as a Google search. Or google "scanf docs" or "scanf manual" or similar. Or look up the official manual for your particular C library. For standard library functions, the standard is an excellent reference, but your particular implementation may have extensions not documented there, or occasionally may even differ from the standard in minor ways.

How can I make input strict with standard functions and throw an
exception if there was neither ints nor floats?

C does not have exceptions. It tells you about errors via function return values, for the most part, though some third-party libraries may do it slightly differently. But you have to do your part, by checking the return values and responding appropriately. In this particular case, you must avoid reading the values of alpha and b. Because those variables were not initialized and did not subsequently have any values assigned to them, reading their values produces undefined behavior.

Others answers, especially @John Bollinger well explain why code fails to meet the goal.

Note input may converter to a int, float, neither, or both.

How can I make input strict with standard functions and throw an exception if there was neither ints nor floats?

The best is to use other tools like fgets() to read a line of input and form a string for later testing.

To test strictly for is a bit hard. *scanf("%d",...) has undefined behavior on overflow. *scanf("%f",...) has similar UB, especially when float does not support infinity/not-a-number.

To test if a string convert to a valid int

bool my_isint(const char *s) {

  // Add to dis-allow leading white-space

  if (isspace((unsigned char) *s)) return false;



  char *endptr; 

  base = 0; // use base = 0 to only allow base 10 input

  errno = 0;

  long val = strtol(s, &endptr, base); 

  if (s == endptr) return false; // no conversion



  if (errno == ERANGE) return false; // too big for long

  if (val > INT_MAX || val < INT_MIN) return false; // too big for int



  // Add to allow trailing white-space

  while (isspace((unsigned char) *endptr)) endptr++;



  // Add to dis-allow trailing junk

  while (*endptr) return false; // trailing junk



  return true;

}

To test if a string converts to a valid float is like testing for double when float supports infinity/non-a-number. More later- GTG.

fgets can be used to read a line. Parse the line with sscanf. sscanf will return the number of items successfully scanned. This also captures the character following the second float. If it is not a newline, the loop continues until two floats followed by a newline are scanned.

#include <stdio.h>

#include <stdlib.h>



int main( void) {

    char input[200] = "";

    char nl = 0;

    int items = 3;

    float alpha;

    float b;

    float result;



    do {

        printf("Type alpha and b. Space separated float numbers only.n");

        if ( fgets ( input, sizeof input, stdin)) {

            // Get user input

            items = sscanf ( input, "%f %f%c", &alpha, &b, &nl);

        }

        else {

            fprintf ( stderr, "fgets EOFn");

            return 0;

        }

    } while ( 3 != items || 'n' != nl);

    printf("%f %fn", alpha, b);

    return 0;

}

As @John Bollinger (answer) and @chux (answer) kindly explained the question is wrong by itself, actually scanf doesn't accept chars instead of floats but omitting it. To catch this behavior I've modified the code in the following way:

+ #include <errno.h> // Error codes constants



int main()

{

  // Define variables

  float alpha;

  float b;

+ int matches;

  float result;



  // Get user input

  printf("Type alpha and b. Space separated float numbers only.n");

± matches = scanf("%f %f", &alpha, &b);



+ if (matches != 2) {

+   printf("ERROR: Space separated float numbers only expected.n");

+   return EINVAL;

+ }



  printf("%f %fn", alpha, b);

  return 0;

}

Appropriate only for simple learning programs you're writing during your CS degree!

Otherwise, check @chux's answer for a more production-like solution.