Should unit tolerance be a % rather than fixed at 1e-5?

[tmilnthorp]

I noticed that when adding the Femto and Peta prefixes to PowerDensity that the tests fail.

[1 W/m³].As( PowerDensityUnit.FemtowattPerCubicMeter ) is calculating 1 / 1e-15 = 999999999999999.87

Obviously this isn't within 1e-5 of 1e15 (actual difference is 0.125).

For a number this large that difference is 0.00000000000001.25% :)

I know that I could override the tolerance in the tests, but perhaps we should think about changing the tolerance across the board? For very small numbers, the opposite is true: 1e-5 may be too large.

[gojanpaolo]

See Relative tolerance instead of absolute tolerance #391

[tmilnthorp]

Ah thank you - I looked through the list but must have missed this. I will close this and reference that one.