Allow trailing comments about doctest

[sol]

@kazu-yamamoto I agree that True :: Bool is not important. But I'm not yet sure if we can exclude such cases in general.

Say we have a type class Foo.

class Foo where

instance Foo Int where

And a property p.

p :: Foo a => a -> Bool

Then the user might want to write.

prop> p :: Int -> Bool

Any thoughts?

[kazu-yamamoto]

If this is important, we can write:

prop> (p :: Int -> Bool)

I think that we should more flexible one to users. I guess that "$" is more flexible than "()" at this moment.
What do you think?

[sol]

The user would still need to know about this. How would you document it and make sure that the user actually reads it?

Saying something like "Trailing comments do not work!" is relatively clear.

An other approach would be to strip trailing comments; but we may run into other corner cases here. I'm really not sure what to do here.

[kazu-yamamoto]

Just for image:

With trailing comments:
http://mew.org/~kazu/tmp/Data-Set-Splay.html#v:delete

Without trailing comments:
http://mew.org/~kazu/tmp/Data-Set-Splay2.html#v:delete

Which do yo prefer?

BTW, I guess dropping trailing comments is possible but its implementation would be troublesome.

[sol]

Personally I prefer the second.

In the first, the comments look too much like being part of the property. I think trailing comments could be useful, but only if they get a different layout (e.g. normal font weight, italic, and gray text color). Something similar to this.

prop> xs == ys -- xs always equals ys

That would again require parsing.

I agree that striping trailing comments could be troublesome. The longer I think about it, the less I like it.

So what to do? I'd tend to just document that trailing comments are not supported. What do you think?

[kazu-yamamoto]

OK. Let's not allow trailing comments at this moment. We can get back this issue if users actually require.