Jan Midtgaard
Jan Midtgaard
On the one hand I'm supportive of this feature. I've used `set_shrink Shrink.nil some_gen` in similar situations for "QCheck1" in the past. One could go for a similar design for...
Would it be a viable strategy to shrink the 'mantissa' and 'exponent' separately? I suppose halving (or subtracting) could work for the exponent at least. E.g. 1.323 * 10^18 -->...
As I wrote elsewhere I think only #188 is required before a release (from my POV)
Closing as we already had a first release of `ppx_deriving_qcheck`
I agree such generators would be a nice addition (also for consistency). It should be relatively straightforward to add them if you study the existing `int` implementations in, e.g., https://github.com/c-cube/qcheck/blob/master/src/core/QCheck.ml#L158...
For, e.g., `i64_bound` and `i64_range` it should be possible to mimic the `int` versions, shouldn't it? I can see `pint` uses `RS.bits` by glueing chunks of 30 random bits together....
@Sir4ur0n Surprising indeed! Also see #90 - I can see I'm assigned but I unfortunately have my hands full in a new job.
Closing this as completed since OCaml5 comes with a new, splittable LXM PRNG (ocaml/ocaml#10742)
Dang! I just realized this was missing only to realize I already reported it 2.5 years ago :sweat_smile:
Getting an idea of our coverage would be nice! A few thoughts: - Have you tried running #206 to see how bad it is? 😅 - I was wondering how...