comparing two tables or datasets

The question relates to datasets as well as tables. It is stated in the following for datasets only. Assume you have two datasets dsA and dsB, which are of equal dimension with identical variables but are unequal (i.e.

isequal(dsA, dsB)

evaluates "false".

Is there a better way to find the actual row(s) and column(s) (or variable(s)) which are different, than the obvious and tedious one? (Obvious and tedious is to check dimensions, if equal check each variable, apply proper "eq" to each variable)

Answer 1

[alexandre iolov]

Look into setdiff(dsA, dsB),

'''If A and B are tables, then setdiff returns the rows from A that are not in B, with repetitions removed. The rows of table C are in sorted order.'''

however, beware that nan~=nan and similarly for entries of categorical arrays which are undefined. undefined ~= undefined

There might be other gotchas that I don't know of, please share.

Answer 2

[Azzi Abdelmalek]

You can access your data then compare them

dsA={1;2;3};
dsA=mat2dataset(dsA);
dsB={1;2;3};
dsB=mat2dataset(dsB)

%compare

sign(dsA.dsA1-dsB.dsB1)

[gg]

I do not think this works for variables where "-" is not defined.

[Azzi Abdelmalek]

If you have numeric data, you can use

sign(cell2mat(dsA.dsA1)-cell2mat(dsB.dsB1))

Answer 3

[Sean de Wolski]

Use a for-loop to loop over the variables and test equality:

LastName = {'Smith';'Johnson';'Williams';'Jones';'Brown'};
Age = [38;43;38;40;49];
Height = [71;69;64;67;64];
Weight = [176;163;131;133;119];
BloodPressure = [124 93; 109 77; 125 83; 117 75; 122 80];

% Two tables, Height is different
T1 = table(Age,Height,Weight,BloodPressure,...
  'RowNames',LastName);

Height = flipud(Height);
T2 = table(Age,Height,Weight,BloodPressure,...
  'RowNames',LastName);
props = T2.Properties.VariableNames;
for ii = numel(props):-1:1
  eqvars(ii) = isequal(T1.(props{ii}),T2.(props{ii}));
end

Then look at

eqvars

Which will be true everywhere the variables are equal.

[gg]

This seems to give me only the variable and not the actual observation(s) creating the difference.

[Sean de Wolski]

Then extend it to do that :)

Or use the logical index to extract them after:

T1{:,~eqvars}-T2{:,~eqvars}

[gg]

This is (part of) what I mentioned in my question as the "obvious but tedious" solution. I was surprised that objects like table or dataset do not come with an "isequal" method able to handle this. I just wanted to make sure I haven't overlooked something.

[Sean de Wolski]

My interpretation is that you want varfun to work like cellfun in that cellfun can work on two cell arrays at once:

C1 = {1, pi, magic(3)}
C2 = {2, pi, magic(3)}
cellfun(@(x,y)x-y,C1,C2,'UniformOutput',false)

Which seems reasonable enough if two tables have all of the same variables..

Answer 4

[Markus Leuthold]

table1=table(....)
table2=table(....)

%comparison
s1=table2struct(table1)
s2=table2struct(table2)
if isequal(s1,s2)
   disp('table1 and table2 are equal')
end

However, I agree there should be an overloaded function isequal in the table class. As a previous commenter already said: If you expect NaN in your table, you have to use

    isequaln

instead of

    isequal

[Peter Perkins]

The isequal function already works on tables, so you don't need to convert to structs:

>> t1 = array2table(randn(5,3));
>> t2 = t1;
>> isequal(t2,t1)
ans =
     1
>> t2.Var1(1) = 0;
>> isequal(t2,t1)
ans =
     0

The original question was, "how to find what part of them is not equal?", for which setdiff/xsetxor, or some of the other posted solutions, will work.

[Al in St. Louis]

This solution does not work for me. isequaln(t1,t2) yields 0 when the tables are identical without being two pointers to the same table as in your example.

[Guillaume]

Can you provide a recipe for that. It seems to work fine for me:

>>t1 = table([1;2;3]);
>>t2 = table([1;2;3]);
>>isequaln(t1, t2)
ans = 
    1

[Peter Perkins]

There are no pointers involved in my code, t1 and t2 are completely different instances. Perhaps you have missing values in your tables, in which case you'll probably want to use isequaln.

Answer 5

[Dan]

isequaln(table1 ,table2) possibly your error derives from the fact that if there are nans isequal will show false.