Hi all,
This is mainly a question for the Clojure core dev team. I'm trying
really hard to understand the thinking behind the new IAtom2 interface.
My train of thought is detailed below, but you don't have to read the
full thing. The core question is "How come the existing IAtom didn't
grow, given the fact that you guys *own* the interface?"
OK so here is the full story:
I recently found out about https://dev.clojure.org/jira/browse/CLJ-1454,
and it's true that often you want a version of `swap!` that returns the
value that was swapped *out*. As indicated on the ticket various people
have worked around the lack of such a fn in various ways. The solution
that I'm using/maintaining looks very much like the approach taken with
IAtom2, only it's goes through a protocol and it's written in clojure.
Basically I've defined a new abstraction and extending it to Atom.
(defprotocol IAtomic
... ;; more methods
(trade! [this f]
[this f x]
[this f x y]
[this f x y more]
"Like `clojure.core/swap!`, but returns the value that was swapped out."))
(defmacro ^:private trade*
[ref f & args]
`(let [validate# (.getValidator ~ref)];; extract the validator-fn once (loop
[]
(let [oldv# (.deref ~ref)
newv# (~f oldv# ~@args)]
(try (when-not (validate# newv#)
(throw (IllegalStateException. "Invalid reference state")))
(catch Exception e#
(IllegalStateException. "Invalid reference state" ) e#))
(if (.compareAndSet ~ref oldv# newv#)
(do (.notifyWatches ~ref oldv# newv#)
oldv#)
(recur))))))
(extend-protocol IAtomic
... ;; more types
Atom
(trade! ([this f]
(trade* this f))
([this f x]
(trade* this f x))
([this f x y]
(trade* this f x y))
([this f x y more]
(trade* this #(apply f % x y more)))
)
)
Ok, so at this point I want to stress out that the way I see it this
solution seems to me like the second best option. I say 'second',
because IMO the best option would be to add `trade!` (or however you
want to call it) to the original IAtom.java and implement it straight in
Atom.java. But obviously I can't do that - only clojure.core can. Since
clojure.core didn't, I'd say that it's safe to assume that either a)
growing IAtom is not desirable, and/or b) having IAtom2 is a superior
solution . Personally, I wouldn't be able to explain/defend any of those
in a conversation. I'm probably missing something here, and that's
exactly why I'm sending this email. Enlighten me please... :) what am i
missing? How come IAtom was, in some sense, cloned, rather than grown?
Many thanks in advance...
Jim
ps: oh and btw congrats on the 1.9 release :)
--
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.
