Combining Unlike Integer Types


If you combine different integer types in a matrix (e.g., signed with unsigned, or 8-bit integers with 16-bit integers), MATLAB® returns a matrix in which all elements are of one common type. MATLAB sets all elements of the resulting matrix to the data type of the left-most element in the input matrix. For example, the result of the following concatenation is a vector of three 16-bit signed integers:

A = [int16(450) uint8(250) int32(1000000)]

Example of Combining Unlike Integer Sizes

After disabling the integer concatenation warnings as shown above, concatenate the following two numbers once, and then switch their order. The return value depends on the order in which the integers are concatenated. The left-most type determines the data type for all elements in the vector:

A = [int16(5000) int8(50)]
A =
   5000   50

B = [int8(50) int16(5000)]
B =
   50   127

The first operation returns a vector of 16-bit integers. The second returns a vector of 8-bit integers. The element int16(5000) is set to 127, the maximum value for an 8-bit signed integer.

The same rules apply to vertical concatenation:

C = [int8(50); int16(5000)]
C =


You can find the maximum or minimum values for any MATLAB integer type using the intmax and intmin functions. For floating-point types, use realmax and realmin.

Example of Combining Signed with Unsigned

Now do the same exercise with signed and unsigned integers. Again, the left-most element determines the data type for all elements in the resulting matrix:

A = [int8(-100) uint8(100)]
A =
   -100   100

B = [uint8(100) int8(-100)]
B =
   100   0

The element int8(-100) is set to zero because it is no longer signed.

MATLAB evaluates each element prior to concatenating them into a combined array. In other words, the following statement evaluates to an 8-bit signed integer (equal to 50) and an 8-bit unsigned integer (unsigned -50 is set to zero) before the two elements are combined. Following the concatenation, the second element retains its zero value but takes on the unsigned int8 type:

A = [int8(50), uint8(-50)]
A =
  50    0