#(f %)
is expanded by the reader into (fn [%] (f %)
. Likewise, #({:key %})
is expanded into (fn [%] ({:key %})
. The python equivalent of this would be lambda v: {'key': v}()
, which has the exact same problem as the Clojure version.
What you are looking for is something equivalent to (fn [v] {:key v})
. If you really want to use #(...)
notation, you could use #(do {:key %})
.
Incidentally, I personally never use #(...)
. I think it’s more difficult to grok (as examples such as this evidence), and is only very slightly more compact than an equivalent fn
form. Then there’s also the limitation that #(...)
forms can not be nested.