Because it works by increasing size, yes, it doesn’t need guidance about the order. On the other hand, you’re exploring a different part of the space of possible inputs. There’s also Lazy SmallCheck, too.

Which is best? There’s no clear answer to this, but a reasonable principle is to try a bundle of approaches if you want to argue that you have used a limited amount of testing resource in as prudent as possible a way.

Simon

On 17 Jun 2018, at 11:28, Oliver Charles <ollie@ocharles.org.uk> wrote:

Is SmallCheck more principled in this regard, or would people consider that equally hacky?

On Sun, 17 Jun 2018, 10:18 am Petr Pudlák, <petr.mvd@gmail.com> wrote:
PS: Just to make clear, it's not that I have something against QuickCheck or similar libraries, on the contrary, they're great! I'm just playing the devil's advocate to analyze and understand the concept.

ne 17. 6. 2018 v 4:05 odesílatel Oleg Grenrus <oleg.grenrus@iki.fi> napsal:
Not only avoid extremely large trees, but in general guarantee termination of the generation process

Sent from my iPhone

On 15 Jun 2018, at 0.31, David Feuer <david.feuer@gmail.com> wrote:

data Foo a = Leaf a | Node [Foo a]

Without the size parameter, it's a bit tricky to control the distribution to avoid generating extremely large trees. I certainly agree, however, that the size parameter is an ugly and ill-specified hack.

On Thu, Jun 14, 2018, 4:20 PM Petr Pudlák <petr.mvd@gmail.com> wrote:
Hi everyone,

I'd like to better understand the principles behind the 'size' parameter. Looking at quickCheckWithResult [1], its computation seems to be somewhat non-trivial, or even arbitrary. As far as I understand it, the size is varied throughout tests, increasing from small to larger values. I see two main purposes:

- Test on smaller as well as larger values. But with generators having proper distribution of values, this should happen anyway, just as if we had a constant, larger 'size' parameter.
- Starting with smaller sizes allows to find smaller count-examples first. But with shrinking, it doesn't matter that much, big counter-examples are shrunk to smaller ones anyway in most cases.

So is this parameter actually necessary? Would anything change considerably if it was dropped?

Thanks,
Petr

_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.
_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.

Simon Thompson | Professor of Logic and Computation 
School of Computing | University of Kent | Canterbury, CT2 7NF, UK
s.j.thompson@kent.ac.uk | M +44 7986 085754 | W www.cs.kent.ac.uk/~sjt