Please add two more testcases for exposure bias:

1. negative bias
2. abs(bias) > 1

Rational, without 'e'

This will give a wrong result (1/2) for 0.4 seconds.

I changed the threshold to be lower than that as a workaround.
My impression was that only 1 over something rationals are used, this computation will of course fail for all other cases...

For a chart of typical exposure times, see https://www.borrowlenses.com/blog/shutter-speed-chart/

I can confirm for my Canon SLR there is at one more value which would give a deviating result, 0.3 seconds (!= 1/3s). 0.25 seconds is shown as 1/4s.

Thanks for the information. From what I've read, for the exposure time the right threshold is 0.3, and the value is always a rational of 1/x.
The story seems to be different for the bias, though, so I've excluded it here.

Computation now also falls back to decimal representation if it cannot be represented with a 1/x rational