Round to max N decimals

Question

I know how to print a rounded number to N decimals like this:

using namespace std;

cout << setprecision(N) << fixed;
float k = 1.23441;
cout << k;

Let's use for example N = 2.

This will output 1.23, which is good, but if k is 1.0012, it will output 1.00, and I want it to output 1. If k is 1.1045, it will output 1.10, and I want it to output 1.1, so I want to round to 2 decimals and remove the unnecesary zeroes.

How can I do so?

Answer 1

Actually, the problem is not setprecision, it's fixed. With fixed you explicitly asked to have a fixed number of digits!

If you use setprecision without fixed, it'll do exactly what you are asking for:

#include <iostream>
#include <iomanip>

int main() {
    double const number = 123.456789;
    for (int i = 0; i != 15; ++i) { std::cout << i << ": " << std::setprecision(i) << number << "\n"; }
    return 0;
}

Will output:

0: 1e+02
1: 1e+02
2: 1.2e+02
3: 123
4: 123.5
5: 123.46
6: 123.457
7: 123.4568
8: 123.45679
9: 123.456789
10: 123.456789
11: 123.456789
12: 123.456789
13: 123.456789
14: 123.456789

You can find documentation on the different modes here:

• fixed
• scientific
• hexfloat
• defaultfloat

And what you want is defaultfloat, which as the name implies is the default.

Answer 2

The appearance of (seemingly) redundant zeros at the end should not be suppressed. Consider

1.2
1.23
1.234
1.2000
1.2300
1.2340

The first number says that it is 1.2 +/- 0.05, the next one says it's 1.23 +/- 0.005 and the third one, 1.234 +/- 0.0005.

In constrast, all of the following numbers indicate a precision that isn't farther off than 0.00005.

So, removing trailing zeros in a decimal fraction is not a way to present date more readable; usually the lack of trailing digits is understood to indicate a lack of precision in the computation.

Later

The removal of characters from a string resulting from a formatted conversion must be coded explicitly if the conversion does not meet your requirements. Therefore, convert to a string using the format that's closest to your requirements, and then post-process the string.

If the string is the result of some fixed-point formatted conversion, according to the pattern \d+\.\d+, you can reverse-scan the string from the end (hint: string::find_last_not_of). Do you want to truncate 1.000 to 1.0 or 1. or 1? The last form will even insinuate that the value is an integer, which it isn't.