Input other than negative numbers that may produce warning "invalid value" with numpy.sqrt?

Question

I had a np.sqrt() snippet that didn't check for negative values, so I understandably got warnings. I thought I had fixed it, but I got one again:

RuntimeWarning: invalid value encountered in sqrt
  data_std = np.sqrt(np.absolute(data_std))

Are there any invalid numbers other than negative ones for np.sqrt?

@Dux `np.sqrt(np.nan)` returns `nan` and `np.sqrt(np.infty)` returns `inf` — iacob, Jun 04 '18 at 14:16
Can you reliably reproduce the warning? If so, a call to `print(np.absolute(data_std))` would give you the answer. — Warren Weckesser, Jun 04 '18 at 14:43
Does this answer your question? [I am getting a warning ](https://stackoverflow.com/questions/39123766/i-am-getting-a-warning-runtimewarning-invalid-value-encountered-in-sqrt) — iacob, Mar 28 '21 at 12:12

iacob · Answer 1 · 2021-03-28T12:25:13.577

Square roots are not defined for negative real numbers, and numpy will produce nan for negative inputs of "real" dtype int, float, -np.inf (and nan itself).

However, square roots are defined for all complex numbers:

dtype	example `x`	`np.sqrt(x)`	`RuntimeWarning`
Positive float	`1.`	`1.`
Positive int	`1`	`1`
Positive complex	`1+0J`	`1`
Negative float	`-1.`	`nan`	⚠️
Negative int	`-1`	`nan`	⚠️
Negative complex	`-1+0j`	`1j`
	`np.inf`	`np.inf`
	`np.inf`	`nan`	⚠️
	`np.nan`	`nan`

Input other than negative numbers that may produce warning "invalid value" with numpy.sqrt?

1 Answers1