How do I round off exponential numbers?

I have a code for which I get a long exponential number eg. a = 2.572920056e-3 as the output. I need to round off only the non-exponential part i.e I want the output to be something like 2.573e-3.

I know how to round off integers using ceil, round etc, but how can one round off only the number outside the exponential like mentioned above. Also, the output varies so I cannot use a command with a fixed number like

 (ceil(a*10^6)/10^6).

(The fixed number being '6' in this case).

Answer 1

[Walter Roberson]

ex = 10^(3-floor(log10(a)));
ar = round(a * ex) / ex;

Caution: might not work for 0 or infinities

[Nitin Samuel]

Thank you so much Mr.Roberson..works fine !.just one more thing..I actually want it to be in the form x.xx*e(x)...Whereas now its in the form x.xxxx*e(x)...its a GUI and the user will not want to see too many figures in an edit box... Can something be done about that? Thanks a lot again...

[Nitin Samuel]

even if 3 is changed to 2 in your code, it adds a 0 making it in the form x.xxxx again..

[Walter Roberson]

How are you doing the formatting for the edit box?

Note: if all of this is just for display purposes, there is no need to do the computation I showed here.

sprintf('%.2e', a)

[Nitin Samuel]

This works fine! This code I shall use for display purposes and the one you gave earlier shall be used for report generation for extra accuracy.Thanks a lot Mr. Roberson..!

[Mikhail Lisakov]

Solution is great but fails for powers > 0 (though I've used it to round significand to integer).

To account for possible powers greater than 0 there should be a condition like this :

if(ex < 0)
ar = round(a * ex) / ex;
} else {
ar = round(a / ex) * ex;
}

[Walter Roberson]

Different requirement, Mikhail. That formula would be for rounding decimals after the decimal point in fixed point format, but the original question was for rounding in engineering format.