Integer conversion without precision loss for literal function inputs

The function uint64 can take a literal input that is not representable by double and convert without precision loss, like

uint64(7725400999518902274)

ans = uint64
7725400999518902274

Unfortunately, this functionality does not seem to extend to a function with an argument block and type validation.

function test(a)
arguments
    a (1,1) uint64
end
disp(a)
end

test(7725400999518902274)

   7725400999518902272

I would have to do

test(uint64(7725400999518902274))

   7725400999518902274

Does anyone know if there is a trick to get this functionality or I am otherwise missing something?

[Chuguang Pan]

@AB. In the argument block, the variable a is a double number, it is converted to uint64.

uint64(double(7725400999518902274))

ans = uint64
7725400999518902272

uint64(7725400999518902274)

ans = uint64
7725400999518902274

[Matt J]

Unfortunately, this functionality does not seem to extend to a function with an argument block and type validation.

The argument block isn't the issue. Here's the same problem without it:

test(7725400999518902274)

   7725400999518902272

function test(a)

 a=uint64(a);

 disp(a)

end

[Dyuman Joshi]

"The function uint64 can take a literal input that is not representable by double and convert without precision loss"

That is not true, it can do that within [0, 2^64-1]

uint64(123456789012345678901)

ans = uint64
18446744073709551615

uint64(2^64-1)

ans = uint64
18446744073709551615

uint64(2^64+1)

ans = uint64
18446744073709551615

[AB]

Thanks everyone for the discussion, this might have been a bit of an unfair question without an answer beyond "no, there isn't".

To clarify, uint64 provides a special execution path that allows a literal that is not representable by a double but is representable by a uint64 to be created as a uint64 without precision loss, and I was hoping there might be a similar special handling that utilizes the arguments block.

Answer 1

[Matt J]

test 7725400999518902274

a = uint64
7725400999518902274

function test(a)
arguments
    a (1,:) string
end
 a=eval("uint64("+ a + ")"),
end

[Dyuman Joshi]

Matt - You are providing a string as an input, however OP is providing a double value.

The same problem resurfaces when providing the input as double -

test(7725400999518902274)

a = uint64
7725400999518902272

function test(a)
arguments
    a (1,:) string
end
a=eval("uint64("+ a + ")"),
end

[Matt J]

@Dyuman Joshi

No, the OP is providing the number as a literal. The Matlab parser clearly does not interpret literal input to uint64 as a double, since,

A=uint64(7725400999518902274);
B=uint64(double(7725400999518902274));

isequal(A,B)

ans = logical
   0

[AB]

This might actually be an ok solution for some use cases, just letting us specify a string with a literal. You could hit it with an sscanf though instead of the eval. I'll accept since this is probably the best we can do right now. Thanks!

[Dyuman Joshi]

@Matt J - Yes, you are right. That was an oversight by me.

[Dyuman Joshi]

@AB, using sscanf is what is suggested in the documentation as well - https://in.mathworks.com/help/matlab/ref/uint64.html#mw_73911831-6f61-4aae-89d5-5afb9ff9a902.

Answer 2

[Walter Roberson]

Does anyone know if there is a trick to get this functionality

There is no way of doing that.

Any way of doing that would have to affect the inputs at parse time. However, arguement blocks do not affect parse time. Arguement blocks apply conversions to whatever input was passed in. By the time the arguement block processing is applied, the parameter has already been parsed as double precision.