setdiff

Arguments

a

a set of elements of type double, boolean, string, sparse, duration, datetime, table and timeseries

b

elements to remove from a.

orien

oriented processing:

• "r": rows of a are searched among b ones.
• "c": columns of a are searched among b ones.
• no orien: elements of a are searched among b ones.

If the option orien="r" is used, a and b must have the same number of columns. If the orien="c" is used, they must have the same number of rows.

This option is not supported for table and timeseries.

v

depending on type of a and b:

• sorted vector of a's components that are not in b.
• orien="r": matrix of rows of a that are not in b, sorted in lexicographic order.
• orien="c": matrix of columns of a that are not in b, sorted in lexicographic order.
• table or timeseries containing the rows of a that are not in b

ka

vector of linear indices of selected a's components, rows, or columns, such that

• v = a(ka), or
• v = a(ka,:) (orien="r"), or
• v = a(:,ka) (orien="c")

Description

setdiff(a, b,..) computes and returns the elements or rows or columns of a that are not in b.

All duplicates (elements or rows or columns) are removed from a and from b before processing.

Both a and b must have the same data type (and integer type), but may mix dense and sparse encoding. It can be vectors, matrices or hypermatrices of real or complex numbers, encoded integers, booleans, strings, duration or datetime. Sparse-encoded inputs are also accepted, as well as table and timeseries.

If a is an hypermatrix and the "r" option is used, a is replaced with the matrix of all its rows over all its higher dimensions, before processing. Same thing if b is an hypermatrix. If the "c" option is used, a or/and "b" are replaced with the respective matrices of all their columns.

If a and b are table or timeseries, setdiff(a, b) returns the sorted unique rows contained in a that are not in b. a and b must have the same variable names. For tables, row names are ignored unlike timeseries where row times are taken into account.

The format of results is presented in the following table, according to the shape of a and the orien option. In all cases, if all entities of a are in b, [] is returned for v as for ka:

v and ka become empty if a is empty (whatever is b), or if all a elements are in b.

For booleans, setdiff(…) is useful mainly with the "r" or "c" options.

For table and timeseries, the orien option is ignored (a warning is displayed).

Examples

Example #1:

a = grand(1, 10,"uin", 0, 9)
b = grand(2, 4, "uin", 0, 9)
[d, k] = setdiff(a, b);
d, k

--> a = grand(1, 10,"uin", 0, 9)
 a  =
   2.   2.   4.   5.   4.   1.   9.   5.   8.   3.

--> b = grand(2, 4, "uin", 0, 9)
 b  =
   5.   0.   9.   9.
   5.   6.   0.   4.

--> [d, k] = setdiff(a, b);
--> d, k
 d  =
   1.   2.   3.   8.

 k  =
   6.   1.   10.   9.

Example #2: column-wise processing

a = grand(2, 7,"uin", 0, 3)
b = grand(2, 10, "uin", 0, 3)
[d, k] = setdiff(a, b, "c");
d, k

--> a = grand(2, 7,"uin", 0, 3)
 a  =
   0.   1.   0.   2.   3.   0.   2.
   2.   2.   2.   1.   0.   1.   2.

--> b = grand(2, 10, "uin", 0, 3)
 b  =
   1.   1.   3.   1.   1.   1.   3.   0.   2.   0.
   3.   3.   2.   2.   0.   0.   1.   0.   1.   0.

--> [d, k] = setdiff(a, b, "c");
--> d, k
 d  =
   0.   0.   2.   3.
   1.   2.   2.   0.

 k  =
   6.   1.   7.   5.

Example #3: with some text

v1 = tokens("ab  ca  ba  bb  ca  cb  ba  aa  cc  bc  ac  aa")'
v2 = tokens("cc  ac  ca  bb  ac  bc  ab")'
[r, k] = setdiff(v1, v2);
r, k

--> v1 = tokens("ab  ca  ba  bb  ca  cb  ba  aa  cc  bc  ac  aa")'
 v1  =
!ab  ca  ac  bb  ca  cb  ba  aa  cc  bc  ac  aa  !

--> v2 = tokens("cc  ac  ca  bb  ac  bc  ab")'
 v2  =
!cc  ac  ca  bb  ac  bc  ab  !

--> [r, k] = setdiff(v1, v2);
--> r, k
 r  =
!aa  ba  cb  !

 k  =
   8.   3.   6.

setdiff() for duration and datetime:

A = hours([1 7 7 4 3 6 6]);
B = hours([7 5 3 0]);
[v, ka] = setdiff(A, B)

A = datetime("today") + A;
B = datetime("today") + B;
[v, ka] = setdiff(A, B)

 
--> A = hours([1 7 7 4 3 6 6]);
--> B = hours([7 5 3 0]);
--> [v, ka] = setdiff(A, B)
v = [1x3 duration]

01:00:00   04:00:00   06:00:00

ka = [1x3 double]

1.   4.   6.

--> A = datetime("today") + A;
--> B = datetime("today") + B;
--> [v, ka] = setdiff(A, B)

v = [1x3 datetime]

2025-08-08 01:00:00   2025-08-08 04:00:00   2025-08-08 06:00:00

ka = [1x3 double]

1.   4.   6.
        

setdiff() for table:

A = table([1 7 7 4 3 6 6]', [1:7]')
B = table([7 5 3 0]', [2;2;5;5])
[v, kab] = setdiff(A, B)

A = table([1;3;4;2], ["d";"c";"f";"h"], [%f;%t;%t; %f], "VariableNames", ["double", "string", "boolean"])
B = table([2;3;4;1], [%t;%t;%t;%t], ["d";"c";"f";"h"], "VariableNames", ["double", "boolean", "string"])
[v, ka] = setdiff(A, B)

 
--> A = table([1 7 7 4 3 6 6]', [1:7]')
 A = [7x2 table]

   Var1   Var2
   ____   ____
              
   1      1   
   7      2   
   7      3   
   4      4   
   3      5   
   6      6   
   6      7   

--> B = table([7 5 3 0]', [2;2;5;5])
 B = [4x2 table]

   Var1   Var2
   ____   ____
              
   7      2   
   5      2   
   3      5   
   0      5   

--> [v, kab] = setdiff(A, B)

 v = [5x2 table]

   Var1   Var2
   ____   ____
              
   1      1   
   4      4   
   6      6   
   6      7   
   7      3   

 kab = [5x1 double]

   1.
   4.
   6.
   7.
   3.

--> A = table([1;3;4;2], ["d";"c";"f";"h"], [%f;%t;%t; %f], "VariableNames", ["double", "string", "boolean"])

 A = [4x3 table]

   double   string   boolean
   ______   ______   _______
                            
   1        d        F      
   3        c        T      
   4        f        T      
   2        h        F      

--> B = table([2;3;4;1], [%t;%t;%t;%t], ["d";"c";"f";"h"], "VariableNames", ["double", "boolean", "string"])

 B = [4x3 table]

   double   boolean   string
   ______   _______   ______
                            
   2        T         d     
   3        T         c     
   4        T         f     
   1        T         h     

--> [v, ka] = setdiff(A, B)

 v = [2x3 table]

   double   string   boolean
   ______   ______   _______
                            
   1        d        F      
   2        h        F      

 ka = [2x1 double]

   1.
   4.
        

setdiff() for timeseries:

A = timeseries(hours([1:10])', [2 7 0 3 6 6 8 6 8 0]')
B = timeseries(hours(1:2:15)', [2 9 6 8 8 3 0 0]')
[v, ka] = setdiff(A, B)

 
--> A = timeseries(hours([1:10])', [2 7 0 3 6 6 8 6 8 0]')
 
 A = [10x1 timeseries]

     Time     Var1
   ________   ____
                  
   01:00:00   2   
   02:00:00   7   
   03:00:00   0   
   04:00:00   3   
   05:00:00   6   
   06:00:00   6   
   07:00:00   8   
   08:00:00   6   
   09:00:00   8   
   10:00:00   0   
--> B = timeseries(hours(1:2:15)', [2 9 6 8 8 3 0 0]')

 B = [8x1 timeseries]

     Time     Var1
   ________   ____
                  
   01:00:00   2   
   03:00:00   9   
   05:00:00   6   
   07:00:00   8   
   09:00:00   8   
   11:00:00   3   
   13:00:00   0   
   15:00:00   0   

--> [v, ka] = setdiff(A, B)

v = [6x1 timeseries]

     Time     Var1
   ________   ____
                  
   02:00:00   7   
   03:00:00   0   
   04:00:00   3   
   06:00:00   6   
   08:00:00   6   
   10:00:00   0   

 ka = [6x1 double]

   2.
   3.
   4.
   6.
   8.
   10.

See also

• unique — ベクトルまたは行列のユニークなな要素を展開
• union — Set of all elements, rows, or columns of two arrays, without duplicates
• members — 配列の各要素または他の配列の行または列を数える(及び位置を調べる)
• vectorfind — locates occurences of a (wildcarded) vector in a matrix or hypermatrix

History