Simple haskell lens from tutorial violates lens law

Question

I'm reading this tutorial:

http://blog.jakubarnold.cz/2014/08/06/lens-tutorial-stab-traversal-part-2.html

and my code looks like this:

import Control.Applicative
import Data.Traversable
import Control.Lens

data User = User String [Post] deriving Show
data Post = Post String deriving Show

posts :: Lens' User [Post]
posts f (User n p) = fmap (User n) (f p)

users :: [User]
users = [User "john" [Post "hello", Post "world"], User "bob" [Post "foobar"]]

tp :: (Traversable t, Applicative f) => ([Post] -> f [Post]) -> t User -> f (t User)
tp = traverse . posts

Now following the blog post, here are some common lens computations:

*Main> view tp users
[Post "hello",Post "world",Post "foobar"]

*Main> set tp [Post "x",Post "y"] users
[User "john" [Post "x",Post "y"],User "bob" [Post "x",Post "y"]]

*Main> view tp (set tp [Post "x",Post "y"] users)
[Post "x",Post "y",Post "x",Post "y"]

The last evaluation has left me confused. Isn't the following lens law supposed to hold?

view l (set l v s) = v

(The law is taken from http://artyom.me/lens-over-tea-2.)

Answer 1

This is a lens law, and tp is a traversal, so it doesn't have to abide by this law.

Strictly speaking, view shouldn't work with traversals at all (it works anyway, but it does so by combining the gathered results monoidally).