left outerjoin without changing initial order in the output

i want to make outerjoin in such way to have output C keeping the same order as the key variable, precisely my key variable is dates when i apply outerjoin i see that the output C is sorted ascending way because the dates are detected as numbers example

   dates=[05012000 02032000 31012000 15092007]';
    ret1=[100 55 66 23]';
  tab1=table(dates,ret1);
  clear dates
  %========
  dates=[05012000 08012000 31122000]';
  ret2=[2 8 7]';
  tab2=table(dates,ret2);
  %====
  [C] = outerjoin(tab1,tab2,'Type','left');
  %result
   dates_tab1          dates_tab2
 *2032000*  55  NaN  NaN
  5012000  100  5012000  2
 15092007  23  NaN  NaN
*31012000*   66  NaN  NaN

you can see that in the output we have 31012000 in the bottom wich is not convenient cause i want to have the same sorting dates us in my initial vector dates such that

 dates_tab1          dates_tab2
  5012000  100  5012000  2 
 *31012000*   66  NaN  NaN
 *2032000*  55  NaN  NaN
 15092007  23  NaN  NaN

Answer 1

[Guillaume]

The problem is your definition of the order of numbers is not the mathematical standard one. Matlab does not let you redefine the order of numbers 3101 is always greater than 1509, so the proper solution is not to use numbers to store your date but a proper date type such as datetime

%in R2016b:
tab1.dates = datetime(compose('%08d', tab1.dates), 'InputFormat', 'ddMMyyyy');
tab2.dates = datetime(compose('%08d', tab2.dates), 'InputFormat', 'ddMMyyyy');
C = outerjoin(tab1, tab2, 'Type', 'left')

Or

%any version with outerjoin
todatetime = @(d) datetime(mod(d, 1e4), floor(mod(d, 1e6)/1e4), floor(d/1e6));
tab1.dates = todatetime(tab1.dates)
tab2.dates = todatetime(tab2.dates)
C = outerjoin(tab1, tab2, 'Type', 'left')

[jean claude]

good answer Guillaume, i use matlab R2015b the second alternativ is right for me !

Thanks very very much

[wgourlay]

I'm also trying to use outerjoin without changing the initial order of my data, however my keys are alphanumeric strings. If I modify the original example as follows:

keys1={'b1' 'b2' 'c1' 'a1'}';
keys2={'d1' 'b1' 'a1'}';

tab1=table(keys1, 'VariableNames', {'Key1'});
tab2=table(keys2, 'VariableNames', {'Key1'});

[C, ia, ib] = outerjoin(tab1,tab2)

What I get is the following:

C = 

      Key1_tab1    Key1_tab2
      _________    _________

      'a1'         'a1'     
      'b1'         'b1'     
      'b2'         ''       
      'c1'         ''       
      ''           'd1'

What I would like to see instead is the following (note these rows occur in the same order that they do in the original data sets):

    C = 

        Key1_tab1    Key1_tab2
        _________    _________

        ''           'd1'    
        'b1'         'b1'     
        'b2'         ''       
        'c1'         ''       
        'a1'         'a1'

I've managed to create a complex block of code that reorders things again based on the ia and ib indexes that are returned from the outerjoin command, but it requires a for loop, two more rounds of sorting, and a few calls to find. I'm hoping there is a simpler solution, something like the 'stable' option used for the unique function, i.e. outerjoin(tab1,tab2,'stable'). Is anyone aware of such a thing?

[Arek Majka]

C.sortvar = ia;

C = sortrows(C,'sortvar');

C.sortvar = [];