Unless I'm missing something subtle, all of your points would hold if you
removed the & in your argument vector to turn your kwargs into an explicit
map, wouldn't they? One advantage is you'd be able to (apply f [m]), but
I'm not sure the :or logic would be any less troublesome.
On Wed, Apr 30, 2014 at 8:06 PM, Mark Engelberg <mark.engelb...@gmail.com>wrote:
> Here's the thing I can't stand about keyword args:
>
> Let's start off with a simple function that looks for keys x and y, and if
> either is missing,
> replaces the value with 1 or 2 respectively.
>
> (defn f [& {:keys [x y] :or {x 1 y 2}}]
> [x y])
>
> => (f :x 10)
> [10 2]
>
> So far, so good.
>
> Now, let's do an extremely simple test of composability. Let's define a
> function g that destructures the keyword args, and if a certain keyword
> :call-f is set, then we're just going to turn around and call f, passing
> all the keyword args along to f.
>
> (defn g [& {call-f :call-f :as m}]
> (when call-f
> (apply f m)))
>
> => (g :call-f true :x 10)
> [1 2]
>
> What? Oh right, you can't apply the function f to the map m. This
> doesn't work. If we want to "apply" f, we somehow need to apply it to a
> sequence of alternating keys and values, not a map.
>
> Take 2:
>
> (defn g [& {:keys [call-f x y] :as m}]
> (when call-f
> (f :x x :y y)))
>
> OK, so this time we try to workaround things by explicitly calling out the
> names of all the keywords we want to capture and pass along. It's ugly,
> and doesn't seem to scale well to situations where you have an unknown but
> at first glance, it seems to work:
>
> => (g :call-f true :x 80 :y 20)
> [80 20]
>
> Or does it?
>
> => (g :call-f true :x 10)
> [10 nil]
>
> What is going on here? Why is the answer coming out that :y is nil, when
> function f explicitly uses :or to have :y default to 2?
>
> The answer is that :or doesn't do what you think it does. The word "or"
> implies that it substitutes the default value of :y any time the
> destructured :y is nil or false. But that's not how it really works. It
> doesn't destructure and then test against nil; instead the :or map only
> kicks in when :y is actually missing as a key of the map.
>
> This means that in g, when we actively destructured :y, it got set to a
> nil, and then that got passed along to f. f's :or map didn't kick in
> because :y was set to nil, not absent.
>
> This is awful. You can't pass through keyword arguments to other
> functions without explicitly destructuring them, and if you destructure
> them and pass them along explicitly, nil values aren't picked up as absent
> values, so the :or default maps don't work properly.
>
> To put it simply, keyword args are bad news for composability.
>
> It's a shame, and I'd love to see this improved (rather than just having
> the community give up on keyword arguments).
>
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clojure@googlegroups.com
> Note that posts from new members are moderated - please be patient with
> your first post.
> To unsubscribe from this group, send email to
> clojure+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "Clojure" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to clojure+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
---
You received this message because you are subscribed to the Google Groups
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to clojure+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
