Scanf() stores a newline when I type characters 'i' or 'n'-CodePudding

I'm trying to have my program end if the user types 'end' and while brainstorming for an implementation, one of the solutions I came up with has an interesting behavior.

In the code below, numbers are stored in firstArray.content[index]; and letters/characters get stored in end[whileCounter]. Everything works as intended, except when end gets an 'i' or an 'n'. When that happens, a new line character '\n' gets stored instead.

I searched online and in the debugger for an explanation of this interesting result but couldn't find a reason. Does anyone know why it happens?

Edit: It doesn't seem to be scanf()'s existing trailing issue as this only happens with specific characters.

Example: If I type 4lol everything works as expected but if I type 4null only ull gets stored in the character array.

I recreated this behavior in the code below:

#define _CRT_SECURE_NO_WARNINGS
#include <stdio.h>
#include <stdlib.h>

struct HeapArrRealNums {
    size_t size;
    size_t nextIndex;
    size_t currentIndex;
    double* content;
};

int main() {

    struct HeapArrRealNums firstArray = { 0 };
    firstArray.size = 4;
    firstArray.nextIndex = 1;
    firstArray.currentIndex = 0;
    firstArray.content = malloc(firstArray.size * sizeof(double));

    #define NOT_A_NUMBER  0
    int scanCheck;
    size_t index = firstArray.currentIndex;
    char end[6] = { 'c' };
    size_t whileCounter = 0;

    while ((scanCheck = scanf("%lf", &firstArray.content[index])) != 1) {
        if (scanCheck == EOF) {
            printf("\n----EOF. Program terminated----");
            firstArray.currentIndex = index;
            //return true;
        }
        if (scanCheck == NOT_A_NUMBER) {

            (void)scanf("%c", &end[whileCounter]);
            whileCounter  ;
            printf("-%s-", end);
            //continue;
        }
    }
}

CodePudding user response：

i is the first character in inf and n is the first character in nan. Those are legitimate inputs for the %f format conversion, so scanf will look at the next character. When it turns out the the next character is \n, scanf puts that character back into the input stream and returns a conversion failure.

But it doesn't put the i/n back, because the standard says that only one character can be returned to the input stream. The restriction comes from ungetc, which is documented as possibly failing if it is called more than once without a getc in between. scanf uses ungetc to return the character, at least in principle, and it's not allowed to try something which might fail.

Really, scanf is not a great interface for reading data more complicated than a sequence of numbers. But it can be made to work if you are aware of its limitations. When it's possible that either a number or a word is the next thing, try the word first:

// This is one of a zillion variants on this theme. You need to be
// clear about your precise requirements.
unsigned char c;
int status;
while (1) {
  // Skip spaces but not newlines
  while ((status = scanf("%c", &c)) == 1
         && isspace(c) && c != '\n') { }
  if (status == EOF) break;
  if (c == '\n') {
    // Handle end of line
  }
  ungetc(c, stdin);
  if (isalpha(c)) {
    status = scanf("%5[a-zA-Z]", end);
    // do something with the word
  }
  else {
    status = scanf("%lf", &datum);
    if (status == 0) {
      // Handle conversion error
      // Note: the invalid character is back in the input.
    }
    // Handle datum
  } 
  // Continue looping
}