grabbing number after decimal

Hi I have a dataset which has column 'specialNumber'. It has data like '1234.5, 9087.3...' and so on, I want to extract number after decimal. so for 1234.5 = 5 for example. How do i do that

[Bart McCoy]

The "fix" function is useful here, since it rounds toward 0. This is important when extracting the fractional part of a NEGATIVE number:

If value = pi, then "value - fix(value)" = 0.141592653589793

If value = -pi, then "value - fix(value)" = -0.141592653589793

Answer 1

[Star Strider]

If you know you will only have one number after the decimal, just use the rem function:

x = 1234.5;
d = 10*rem(x,1);

produces:

d =

       5

[Elena]

how would you do this to for a number like 0.05120? I would only like to get the '05' value from this. (last two numbers after decimal)

Answer 2

[David Sanchez]

N = 1234.5;
R = 10*(N-floor(N))

R =

       5

[Patrik Ek]

You may want to add a floor( R ) after here. Also the, number of digits is selected with 10*.., which in this case means 1 digit. This number may be a variable

N = 1234.537;
nDigits = 2;
R = 10^nDigits * (N-floor(N));
R = floor(R)

Otherwise, nice solution. I will see my comment as a generalized modification rather then a new solution. All cred to you.

[Hikaru]

@Patrik. This does not always work. If N = 1234.5, then your code will return R = 50

Plus, using round is better compared to floor because of the floating points.

[Patrik Ek]

@Hikaru It is true that R becomes 50 when there is only 1 decimal digit in the number. This is not a bug though, since the second digit indeed would be 0 if N is exactly 1234.5. If the accuracy of N is one digit (or if you only want the first decimal digit), it is still not a problem. Then you explicitely need to state that you do only want one digit and not 2 by setting nDigits=1. This is not an unreasonable approach, since this decision should really be the user's.

And about using round instead of floor I do not buy it. You solve a problem that is unlikely to happen, by creating a problem that will occur in 50% of all the cases. Assume that I would use round in my case; This would give the result R = 54, since round(53.7) is 54. This could possibly be a common case. However, since you do the operation N-floor(N), you are unlikely to get floating point errors. Focus on the real problem instead of obscure errors that will probably not occur anyway. An alternative could be you uploading a solution that is safe from floating point errors. To use round is not really a solution to the problem. It is just a workaround and in this case a workaround that does not work as intended.

Answer 3

[Image Analyst]

neesha, what if there are 3 or 4 numbers after the decimal point? Do you want only the first one, or all of them? What if you did

fractions = t.specialNumber - floor(t.specialNumber);

where t is the table and specialNumber is the column in the table. So for example 73.8234 would give 0.8234 as a result. Would that work for you? Or do you only want the 8 as 8 and not as 0.8 or 0.8234?

Answer 4

[Hikaru]

One way to do this

specialNumber = [1234.5; 9087.3];
d = abs(specialNumber - fix(specialNumber));
answer = d*10; %assuming that your data only has one decimal place
final = round(answer)

Note: This is not really the best solution since subtracting floating points can introduce error.

[Patrik Ek]

Well I guess that you are quite safe from floating point errors here, since you you use the same variable. However, I think that a completely stable solution would cost more than you would gain. Floating point errors in operations like this is uncommon in matlab.

Answer 5

[Patrik Ek]

Since there have been a discussion about floating point errors here (which IMO seems to be overkill to care about for this operation) I will provide an integer casting solution. This should be free from floating point issues since no addition or subtraction is done with floating points.

N = 1234.537;
nDigits = 2;
R = cast(floor(10^nDigits*N),'int32') - cast(10^nDigits*floor(N),'int32');
R = cast(R,'double') % If double is important then cast it back

The floor in the first term is there because of matlabs solution of casting, which rounds the double to closest int.

Answer 6

[Neesha]

Hello everybody, thanks a lot for fantatic solutions. While all of your solution seem to work, since i have only one decimal after the number , following works out just fine

x = 1234.5; d = 10*rem(x,1); produces:

d = 5

thanks a bunch again

[Star Strider]

My pleasure!