ERROR: DomainError, Complex numbers?

New to Julia
1

Hi all!

I have a very simple expression similar to:

a = b ^ c

with b and c as positive floating point numbers.

The expression above is part of nested functions (outer and inner functions) with each function defined within the scope of its parent function.

If I run my these nested functions from a Main.jl file, I get the following error (despite a and b being positive Float64):

ERROR: DomainError with -5.991263951486869e-16:
Exponentiation yielding a complex result requires a complex argument.
Replace x^y with (x+0im)^y, Complex(x)^y, or similar.

I tried to sort this error out with the function real() but didn’t work.

Interestingly, if I run my functions line by line (instead of calling the Main.jl file), I can compute the expression above without the error (giving me as a result a floating point number).

I would appreciate any advice/comment/help on this because I am out of ideas!

Thanks

2

put a print statement in. the base of the exponent is negative.

1 Like
3

Thanks! I am 100% sure that the base of the exponent is a positive floating number. When I print b in my example, I get: 21.778641870591873, and the value of the expression a is printed as: 1.2954354359893232

4

As @Oscar_Smith suggested, you should evaluate the expression, assign it to a variable a, and print it and b just before you call a^b. You are 100% sure a is positive, but the error message tells us that a is negative. You need to figure out where/what the discrepancy is.

If you’re 100% sure a is positive because the expression is provably positive, be aware that floating point numbers don’t obey the same rules as the reals. Your value of a is negative but very small, suggesting that it should be zero or a small positive, but instead is negative due to rounding errors.

1 Like
5

Apologies, @Oscar_Smith, I didn’t get your comment right away but with the clarification of @John_Gibson, now I see what you meant.

Printing the exponent before the expression gave me some insights. I don’t know how but, indeed, there were some very small negative numbers in there.

For the moment, I am taking the absolute value of the base of the exponent to avoid the error:

a = abs(b) ^ c

Thanks for the help!

6

That number looks to be on the order of eps(Float64), so it could easily come about in floating point calculations, even with something simple like a-b where a and b seem equal.