hypot unexpected behavior

[mattsignorelli]

Suppose you have two ComplexTPSs (only scalars set)

ct3[0] = 3 + 3im
ct4[0] = 4 + 4im

The function
hypot(ct3, ct4) gives 5 + 5im. So it appears it treats the real and imaginary parts separately

However, suppose you have

t4[0] = 4    # (no 4im)

Now, hypot(ct3, t4) gives 4.4767839548935457 +im*2.0103717513913431
It seems mad_ctpsa_hypot! is computing something different than Julia hypot, so this will need to be resolved. For now, they will be excluded from testing but still exported.

[DavidSagan]

Julia hypot gives a real scalar. I think it will be less confusing to have GTPSA follow Julia.

[mattsignorelli]

@DavidSagan I agree. I will implement hypot for ComplexTPS to do the same as for TPS, which is hypot(t1,t2) = sqrt(abs(t1)^2+abs(t2)^2) . I am wondering though what exactly the complex hypot in gtpsa is calculating and which applications it would be useful

[DavidSagan]

I am perplexed here as well. You can ask Laurent.