Why is MATLAB rounding this?

I have a part of my code:

eps = 10^(log10(eps) + 0.125);

The program will round it off. For instance, if the original value of eps is 2, the calculated value is 133. This is rounded. If I do the same code in the command window, it gives me the correct value, which is a decimal.

If I run the code:

eps = 10^(log10(eps) - 0.125);

and the original value of eps is 0, it returns 0, and the while loop I have it in blows up.

I would like to stop it from rounding. These take place in a nested if statement, within a while loop.

[Oleg Komarov]

Are you sure it's not just a visualization problem?

And btw don't use eps as a variable since it's a builtin matlab fcn

[Sam Falzone]

I'm not actually using eps. Not sure about a visualization error, but I'm pretty sure it is not just a visualization. I am displaying the variable with:

disp(['eps value = ' int2str(eps)])

[Bruno Luong]

Please write what do you expect and you get with a concrete value for eps (and btw change the name eps to something else as Oleg suggested).

[Paulo Silva]

So you want matlab to give 10^-Inf instead of zero?

Answer 1

[Bruno Luong]

a = 2
fprintf('%1.16e', 10^(log10(a) + 0.125))

2.6670428643266479e+000

What's your problem???

[Paulo Silva]

his problem seems to be with a=0

[Sam Falzone]

Not using eps. That's just what I used to describe the problem.

The problem is that 10^(log10(100) + 0.125) displays as 133, not the full decimal.

10^(log10(1) - 0.125) is not equal to 10^-INF. It's 10^-0.125.

[Oleg Komarov]

It doesn't give me any problem. What you do before that in your code? Do you have a variable log10 defined, or a self-written log10 fcn?

[Paulo Silva]

I wonder why is Sam messing around with matlab built in functions?!

Answer 2

[Jan]

If you use INT2STR to display the result, the printed number is rounded - read the help text of INT2STR! Use NUM2STR or SPRINTF instead to get the decimals.

Try Bruno's example and post, if it runs as wanted.

Answer 3

[Walter Roberson]

Double-check to see which "format" you have in effect. And if you are using a variable for the "100" part, check the data class of the variable: if it happened to be one of the integer types then the answer would be integer.