Generating expr

[phadej]

This is try to fix #154

We could rewrite all [Key] or [(Key, Int)] based properties into Expr Key () or Expr Key Int based ones, respectively. I think including union is worth it, as it is has non-trivial implementation.

First commit is from #156.

[treeowl]

I'm not quite convinced this fully solves the problem. We want to guarantee (as best we can) that every valid HashMap tree is available to the Arbitrary instance. Do we have good reasons to believe that all such trees can be generated by this language?

An alternative, more complicated, approach (but quite possibly the right one) would be to use an expression language that includes all primitive operations, interpreting it using HashMap or HashSet and also (say) Data.Map or Data.Set.

One of the key ideas of the containers test suite is that all the invariants for each data structure are expressed in a single valid function. In this context, that would allow your Expr interpretation to find the smallest invalid subexpression, rather than waiting for that invariant failure to show up as a hard-to-debug semantic error.

[treeowl]

Why div and not quot?

[phadej]

Prelude Test.QuickCheck> 3 `div` 2 + 3 `quot` 2
2

[phadej]

I'm not quite convinced this fully solves the problem. We want to guarantee (as best we can) that every valid HashMap tree is available to the Arbitrary instance. Do we have good reasons to believe that all such trees can be generated by this language?

I don't know, I looked at non-trivial -> HashMap k v functions in Base.hs module.

An alternative, more complicated, approach (but quite possibly the right one) would be to use an expression language that includes all primitive operations, interpreting it using HashMap or HashSet and also (say) Data.Map or Data.Set.

AFAICS the insert, delete and union are a good start.

One of the key ideas of the containers test suite is that all the invariants for each data structure are expressed in a single valid function. In this context, that would allow your Expr interpretation to find the smallest invalid subexpression, rather than waiting for that invariant failure to show up as a hard-to-debug semantic error.

That's a good idea, but it can be added independently. I'm not 100% sure what are the invariants.

[treeowl]

Can you rebase this, please?

[phadej]

rebased

[phadej]

ping @treeowl, is there something I can do to get this off my "open PR" -list? :)

[treeowl]

Uh, I seem to have forgotten about this. I'll try to get to it in the next couple days.

[phadej]

This seems to be still forgotten...

[treeowl]

I think we've removed the need for it by removing the representation redundancy. See the changes in Eq implementation.

…

On Mon, Apr 15, 2019, 4:08 PM Oleg Grenrus ***@***.***> wrote: This seems to be still forgotten... — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#157 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABzi_a16EfMStl1eIeoxWmncYbDLEowPks5vhNwrgaJpZM4N4WVP> .