Add [without_shrinker] to [QCheck2.Gen]

[vch9]

Closes #232.

I don't know if that's the most convenient solution, but it seems to work. If you have any suggestion to improve this function, I'd be happy to implement it.

[jmid]

Thanks for this! :pray:

In QCheck(1) there's:

val set_shrink : 'a Shrink.t -> 'a arbitrary -> 'a arbitrary

meaning you can achieve something like this with set_shrink Shrink.nil some_gen.

Perhaps we should rename and generalize slightly, so that the QCheck2 operation takes a shrinker function as an argument?

  val set_shrink: ('a -> 'a Seq.t) -> 'a Gen.t -> 'a Gen.t

That way

• we continue to keep the interfaces close, in an effort to make QCheck->QCheck2 porting easy
• you should be able to achieve the same functionality with set_shrink (fun -> Seq.empty) some_gen (we could also include a binding Shrink.nil for the function argument)

[vch9]

Thanks for this! pray

In QCheck(1) there's:

val set_shrink : 'a Shrink.t -> 'a arbitrary -> 'a arbitrary

meaning you can achieve something like this with set_shrink Shrink.nil some_gen.

Perhaps we should rename and generalize slightly, so that the QCheck2 operation takes a shrinker function as an argument?

  val set_shrink: ('a -> 'a Seq.t) -> 'a Gen.t -> 'a Gen.t

That way

* we continue to keep the interfaces close, in an effort to make QCheck->QCheck2 porting easy

* you should be able to achieve the same functionality with `set_shrink (fun -> Seq.empty) some_gen` (we could also include a binding `Shrink.nil` for the function argument)

Yep, that's a good idea. Pushed a new commit with set_shrink.

[jmid]

Looks good!

Three minor/bike-shed'dy things:

• ~~We should probably add a small test of set_shrink too?~~
• Regarding documentation a reader could misunderstand this comment to be side-effecting (if they do not look at the type signature). One possibility is to adapt it to say returns a generator using [gen] but with shrinking disabled (or something like that):

  val without_shrinker : 'a t -> 'a t
  (** [without_shrinker gen] removes the shrinker from [gen]. *)
  

• I was thinking about the naming. Looking at names related to shrinking, the last one stands out a bit (only entry using shrinker). Furthermore it is a short-hand for something that can be achieved with set_shrink - but it is a relatively long identifier. One option could be to use no_shrink instead (same form of 'shrink' + short) - I'm open to other proposals though

  val add_shrink_invariant : ('a -> bool) -> 'a t -> 'a t
  val set_shrink : ('a -> 'a Seq.t) -> 'a t -> 'a t
  val without_shrinker : 'a t -> 'a t
  

[jmid]

Sorry, I take back the suggestion of adding a test of set_shrink, since you're already using it to implement without_shrinker/no_shrink and hence testing it as part of that... (I blame the work injury from having written too many tests of QCheck previously... :sweat_smile:)

[vch9]

Hi! Sorry for the huge delay.

Pushed a rebased branch:

1. Used your suggestion for the documentation
2. Went fo no_shrink, I like the name, explicit enough and fits well with set_shrink
3. Fixed the changelog I think? The # 0.20 was missing. I don't know if that was intended

[jmid]

Looks good to me, thanks!