Bugs with `complex` math
Opened this issue · 1 comments
bkDJ commented
Some examples of surprising results. Not sure which should be ignored due to floating point imprecision, and which are actual issues.
ⁿ1/2 ℂ0 ¯1 # NaN
=i √ ℂ0 ¯1 # 0
=ℂ∞ 0 ×i∞ # 0
=ℂ2 0 √ℂ0 ¯4 # 0
kaikalii commented
Lines 2 and 4 were not properly displaying the error bound. I've fixed that in 89f4faf.
I'm not sure what to do about lines 1 and 3.
Line 1 has to do with the fact that for doubles, ⁿ1/2 ¯1
is NaN
.
Line 3 has to do with the fact that ×0 ∞
is NaN
.
While these could be fixed with checks in the interpreter, it would slow down complex arithmetic, so I'm not sure it's something I want to do.