How to eliminate the error caused by converting double to uint16

How to eliminate the error caused by converting double to uint16

[Dyuman Joshi]

Could you specify which error?

Are you trying to convert a non-integer double value to uint16?

Or are you trying to assign numbers outside of the range of uint16 i.e. (0:1:2^16-1)

[Walter Roberson]

It would be natural to speculate that there might be some double precision values that cannot be converted, such as maybe inf or nan. However it turns out that every double precision number can be converted to some uint16 without error message.

So we have to guess that maybe you assigned a value to a variable named "uint16", or else that the number of input double precision values is not the same as the number of locations available to be assigned to.

[Dyuman Joshi]

@Walter Roberson

"However it turns out that every double precision number can be converted to some uint16 without error message."

Would using "mapped" be more appropriate instead of "converted"?

[Walter Roberson]

@Dyuman Joshi "mapped" is not a bad word there, but see that the help documentation uses "convert"

help uint16

 UINT16 Convert to unsigned 16-bit integer.
    I = UINT16(X) converts the elements of the array X into unsigned 16-bit
    integers. X can be any numeric object, such as a DOUBLE. The values
    of a UINT16 range from 0 to 65535, or INTMIN('uint16') to INTMAX('uint16').
    Values outside this range saturate on overflow, namely they are mapped
    to 0 or 65535 if they are outside the range. If X is already an unsigned
    16-bit integer array, then UINT16 has no effect. DOUBLE and SINGLE values
    are rounded to the nearest UINT16 value on conversion.
 
    Some arithmetic operations are defined for UINT16 on interaction with
    other UINT16 arrays. For example, +, -, .*, ./, .\ and .^.
    If at least one operand is scalar, *, /, \ and ^ are also defined.
    UINT16 arrays may also interact with scalar DOUBLE variables, including
    constants, and the result of the operation is UINT16.
    UINT16 arrays saturate on overflow in arithmetic.
 
    You can define or overload your own methods for the UINT16 class (as you
    can for any object) by placing the appropriately named method in an
    @uint16 directory within a directory on your path.
    Type HELP DATATYPES for the names of the methods you can overload.
 
    A particularly efficient way to initialize a large UINT16 arrays is: 
 
       I = zeros(1000,1000,'uint16')
 
    which creates a 1000x1000 element UINT16 array, all of whose entries are
    zero. You can also use ONES and EYE in a similar manner.
 
    Example:
       X = 17 * ones(5,6,'uint16')
 
    See also DOUBLE, SINGLE, DATATYPES, ISINTEGER, UINT8, UINT32, UINT64, INT8,
    INT16, INT32, INT64, INTMIN, INTMAX, EYE, ONES, ZEROS.

    Documentation for uint16
       doc uint16

    Other uses of uint16

       categorical/uint16      fixedpoint/uint16    opc.hda.Data/uint16
       codistributed/uint16    gpuArray/uint16

[Dyuman Joshi]

I see, "convert" is used generally, where as "map" is used particularly when trying to convert numbers outside the range of an integer class.

[Walter Roberson]

Consider, though, that (for example) π is within the numeric range 0 to 65535 so if you were using "convert" for in-range but "map" for out of range, you would have to use "convert" for π . uint16(pi) is of course uint16(round(pi)) which is 3 and there are a lot of other nearby values like that also get changed to 3. So even within the range 0 to 65535 the mapping is not "one to one and onto". It would not be unreasonable to argue that "convert" is possibly the wrong verb when the conversion is not bijective, that "map" might be a more appropriate verb -- but that "map" should be used throughout, not just for values outside of the range 0 to 65535.