Add inon, a version of non to be used for caching #916
Conversation
src/Control/Lens/Iso.hs
Outdated
| -- Using 'inon' instead of 'non' means that key will not be deleted | ||
| -- from the map: | ||
| -- | ||
| -- >>> fromList [("user1", "!!!")] & (at "user1" . inon def) .~ "" |
There was a problem hiding this comment.
Don't
fromList [("user1", "!!!")] & at "user1" ?~ ""has same effect?
And for caching, why the removal of defautl value would be bad. I don't understand that either.
There was a problem hiding this comment.
I'll answer the second question. The scenario is that your cache is a Map, and the set of keys is an important piece of information, you don't want to forget that a key was once an element of the map. Picture the user list of a web app, you don't want the user to disappear if they reset their user profile to the default.
There was a problem hiding this comment.
As for the first question, yes that has the same effect, but in my use case I am constructing the lens with non and then composing further lenses to look inside the value. This lens gets passed to a function, so we no longer have access to the lens that looks at the map itself. I guess I dropped that last bit out of my example.
There was a problem hiding this comment.
like this:
>>> (fromList [("user1", "axc")]) & (at "user1" . non "abc" . ix 1) .~ 'b'
fromList []
| -- >>> fromList [("user1", "axc")] & (at "user1" . inon "abc" . ix 1) .~ 'b' | ||
| -- fromlist [("user1", "abc")] | ||
| inon :: a -> Iso' (Maybe a) a | ||
| inon a = anon a (const False) |
There was a problem hiding this comment.
I have a gut feeling that you be ok with inon :: a -> Setter (Maybe a) a, which isn't as unlawful. (It still changes Nothing to Just a when you %~ id).
I'd ask you to make it a setter and move to Unsound module.
There was a problem hiding this comment.
Then I would have to pass both a getter and a setter to the function, which seems the opposite direction from where I am trying to go. I will give it some thought.
There was a problem hiding this comment.
I'll think more about this too.
There was a problem hiding this comment.
Ultimately I just started writing anon v (const False) to avoid this problem.
I had a long bewildering experience using regular non for maintaining a cache, so I thought this modification might be a useful addition to the library.