Calculation accuracies
Opened this issue · 2 comments
NicoLaval commented
@noahboerger reported:
- In trevas the result of an operation is kept in double accuracy (e.g. numeric/power_1)
- User tests assume sometimes less accurate results (e.g. user test: 8.006001 trevas: 8.006000999999998)
I think every language will give this kind of result. To be sure of the result, if a particular precision is needed, user can use round
operator. @hadrienk, can you confirm?
noahboerger commented
I fully agree on this. This is not an issue on trevas, but more on the testcases we have been using. I think these should be adjusted to use the rounding operation where necessary.
hadrienk commented
@NicoLaval yes, rounding is the solution. This is a case of the good old 0.1 + 0.2 != 0.3
.
https://0.30000000000000004.com/