scikit-hep / iminuit / 34246684507

100%

push

github

fix: `log_or_zero` helper mistakenly downcasts integers to floats when the input is of integer dtype (#1191)

Fixes https://github.com/scikit-hep/iminuit/issues/1190

The line `r[ma] = np.log(x[ma])` effectively down casts floats to ints
if `r` is of integer dtype meaning that `r` should be constructed with a
floating dtype but it accidentally inherited `x`'s dtype which might be
integer as it was constructed with `np.zeros_like(x)`. We now just
construct it as `np.zeros_like(x * 1.0)`. This was the simplest
solution. See the comment in the code for the reason.

1 of 1 new or added line in 1 file covered. (100.0%)

3944 of 3957 relevant lines covered (99.67%)

1.0 hits per line