I'd like to know how to limit an input value to signed decimals using std::cin
.
6 Answers
double i;
//Reading the value
cin >> i;
//Numeric input validation
if(!cin.eof())
{
peeked = cin.peek();
if(peeked == 10 && cin.good())
{
//Good!
count << "i is a decimal";
}
else
{
count << "i is not a decimal";
cin.clear();
cin >> discard;
}
}
This also gives an error message with the input -1a2.0 avoiding the assignation of just -1 to i.

- 12,061
- 13
- 49
- 64
-
1#1: What is `discard`? #2: Why `10`? What about Mac? #3: How do you form a loop for retrying? When you wrote it try the input `"ddd"`. You will get your errors many times. If not, then I'm really curious about your solution. – Notinlist Oct 10 '11 at 08:25
-
Is that supposed to be `int peeked`? https://en.cppreference.com/w/cpp/io/basic_istream/peek says it returns a character or `Traits:eof()`. – Peter Cordes Sep 16 '21 at 10:54
If the backing variable of the cin
is a number, and the string provided is not a number, the return value is false, so you need a loop:
int someVal;
while(!(cin >> someVal)) {
cin.reset();
cout << "Invalid value, try again.";
}

- 13,998
- 1
- 57
- 77
-
14
-
8And also - if I replace the reset() with clear(), which you meant or not - it leads to infinite loop when a non-numeric input is given. – Notinlist Oct 10 '11 at 07:46
Combining the techniques from the top answer here and this website, I get
input.h
#include <ios> // Provides ios_base::failure
#include <iostream> // Provides cin
template <typename T>
T getValidatedInput()
{
// Get input of type T
T result;
cin >> result;
// Check if the failbit has been set, meaning the beginning of the input
// was not type T. Also make sure the result is the only thing in the input
// stream, otherwise things like 2b would be a valid int.
if (cin.fail() || cin.get() != '\n')
{
// Set the error state flag back to goodbit. If you need to get the input
// again (e.g. this is in a while loop), this is essential. Otherwise, the
// failbit will stay set.
cin.clear();
// Clear the input stream using and empty while loop.
while (cin.get() != '\n')
;
// Throw an exception. Allows the caller to handle it any way you see fit
// (exit, ask for input again, etc.)
throw ios_base::failure("Invalid input.");
}
return result;
}
Usage
inputtest.cpp
#include <cstdlib> // Provides EXIT_SUCCESS
#include <iostream> // Provides cout, cerr, endl
#include "input.h" // Provides getValidatedInput<T>()
int main()
{
using namespace std;
int input;
while (true)
{
cout << "Enter an integer: ";
try
{
input = getValidatedInput<int>();
}
catch (exception e)
{
cerr << e.what() << endl;
continue;
}
break;
}
cout << "You entered: " << input << endl;
return EXIT_SUCCESS;
}
Sample run
Enter an integer: a
Invalid input.
Enter an integer: 2b
Invalid input.
Enter an integer: 3
You entered: 3.
cin's >> operator works by reading one character at a time until it hits whitespace. That will slurp the whole string -1a2.0
, which is obviously not a number so the operation fails. It looks like you actually have three fields there, -1, a, and 2.0. If you separate the data by whitespace, cin will be able to read each one without problem. Just remember to read a char
for the second field.

- 34,290
- 15
- 75
- 125
I tried many techniques for reading integer input from the user using the >>
operator, but in a way or another all my experiments have failed.
Now I think that getline()
function (not the method with the same name on std::istream
) and the strtol()
function from the include cstdlib
is the only predictable consistent solution for this problem. I would appreciate if someone proved me wrong. Here is something like the one I use:
#include <iostream>
#include <cstdlib>
// @arg prompt The question to ask. Will be used again on failure.
int GetInt(const char* prompt = "? ")
{
using namespace std; // *1
while(true)
{
cout << prompt;
string s;
getline(cin,s);
char *endp = 0;
int ret = strtol(s.c_str(),&endp,10);
if(endp!=s.c_str() && !*endp)
return ret;
}
}
- *1: Placing
using namespace whatever;
to the global scope may lead to broken "unity builds" (google!) on larger projects, so should be avoided. Practice to not use that way, even on smaller projects! - Reading integers from files is a very different matter. Raúl Roa's approach can be good for that if properly worked out. I also suggest that wrong input files should not be tolerated, but it really depends on the application.
- Be warned that using
>>
andgetline()
in the same program oncin
will lead to some problems. Use one of them only, or google to know how to handle the issue (not too hard).

- 16,144
- 10
- 57
- 99
Something like:
double a;
cin >> a;
Should read your signed "decimal" fine.
You'll need a loop and some code to make sure it handles invalid input in a sensible way.
Good luck!

- 1,883
- 5
- 18
- 20