# Unique changing order(unique output values are reshuffled)

10 views (last 30 days)
shaz on 18 Dec 2012
Commented: Stephen23 on 17 Jan 2018
i have cell array with data like '2/6/2009','2/6/2009','2/6/2009','2/6/2009','2/7/2009',,'2/7/2009','2/7/2009','2/8/2009','2/8/2009','2/8/2009'
if i make use of unique to get only unique values the order is getting reshuffled
getting output like : '2/8/2009','2/6/2009','2/7/2009'
Desired output format: '2/6/2009','2/7/2009','2/8/2009'
Stephen23 on 17 Jan 2018
Simply using my FEX submission natsort:
>> C = {'2/6/2009','2/6/2009','2/6/2009','2/6/2009','2/7/2009','2/7/2009','2/7/2009','2/8/2009','2/8/2009','2/8/2009'};
>> D = natsort(unique(C));
>> D{:}
ans = 2/6/2009
ans = 2/7/2009
ans = 2/8/2009
But the best solution is to change the date format for an ISO 8601 date format, which sort correctly into chronological order when you do a character sort. Once you start using ISO 8601 date formats you simply avoid all of these trivial problems.

Muruganandham Subramanian on 18 Dec 2012
b={'2/6/2099','2/6/2099','2/6/2099','2/6/2099','2/7/2099','2/7/2099', '2/7/2099','2/8/2099','2/8/2099','2/8/2099','2/10/2099','2/10/2099', '2/12/2099'};
c=datevec(b);
c(:,4:6)=[];
c(:,4)=c(:,1);
c(:,1)=[];
d=num2str(c);
d=cellstr(d);
d=unique(d)
Check this above code. But it's not effective way to do..here In unique(), it's not chosing in random way, it considers the data of first,which is sorted minimum(i.e. '10' is first than '6').
shaz on 31 Dec 2012
thanks a lot

Muruganandham Subramanian on 18 Dec 2012
Edited: Muruganandham Subramanian on 18 Dec 2012
>>b={'2/6/2009','2/6/2009','2/6/2009','2/6/2009','2/7/2009','2/7/2009','2/7/2009',
'2/8/2009','2/8/2009','2/8/2009'};
>> c=unique(b)
>> c =
'2/6/2009' '2/7/2009' '2/8/2009'
Are you expecting this?
shaz on 18 Dec 2012
b={'2/6/2099','2/6/2099','2/6/2099','2/6/2099','2/7/2099','2/7/2099', '2/7/2099','2/8/2099','2/8/2099','2/8/2099','2/10/2099','2/10/2099', '2/12/2099'};
c=unique(b)
the output is
c= '2/10/2099' '2/12/2099' '2/6/2099' '2/7/2099' '2/8/2099'
which is in a random way

Jan on 18 Dec 2012
With a modern Matlab version you can use:
c = unique(b, 'first');
Stephen23 on 17 Jan 2018
Or option 'stable'.

Sebastian on 17 Jan 2018
I know it's a bit late for an answer, but maybe it helps others. If you want to preserve the order of the input of 'unique', you can use the second return parameter. Like that it works for arbitrary data (i.e. strings and numerals, basically every data type 'unique' supports):
a = {'2/6/2009' '2/6/2009' '2/6/2009' '2/7/2009' '2/7/2009' '2/7/2009' '2/8/2009' '2/8/2009' '2/8/2009'};
[~, uIdx] = unique(a);
a(sort(uIdx))
This snippet does the following: Store the indices of the unique elements of 'a' in 'uIdx'. Then return the elements of 'a' from the first to the last using 'sort'.
Jan on 17 Jan 2018
The problem of shaz was, that he wanted a specific numerical order, in which '6' appears before '10'. Therefore the data must be converted from string to double, such that the sorting works.

Steven Lord on 17 Jan 2018
Convert your date data into a datetime array, then call unique (with or without the 'stable' flag) on the datetime array.
C = {'2/6/2009','2/6/2009','2/6/2009','2/6/2009','2/7/2009','2/7/2009', ...
'2/7/2009','2/10/2009','2/8/2009','2/8/2009'};
D = datetime(C, 'InputFormat', 'MM/dd/uuuu')
unique(D)
unique(D, 'stable')
For purposes of this example I assumed your dates were days in February 2009. If they were the second of each month from June through October (without September) of 2009 swap the MM and dd sections of the InputFormat parameter.
Stephen23 on 17 Jan 2018
"For purposes of this example I assumed your dates were days in February 2009. If they were the second of each month from June through October (without September) of 2009 swap the MM and dd sections of the InputFormat parameter."
Or avoid this pointless and confusing ambiguity entirely by using ISO 8601 dates.