Hi, Am Samstag, den 03.11.2012, 11:26 +0100 schrieb Peter Divianszky:
This can be implemented by replacing every record update
r' = r { x = y }
with
r' = r { x = unsafePerformIO (cond-update r r' (x r) y) }
where we keep the current record update mechanism and implement cond-update like
cond-update :: r -> r -> x -> IO x cond-update r_old r_new x_old x_new = do eval x_new b <- x_old === x_new when b (replace r_new r_old) return x_new
where (===) is pointer-equality and replace is a low-level function which replaces thunks in the heap.
I think one problem with this approach is that now, until "x r'" is evaluated, you keep both r and r' alive. If r was _not_ retained by anything else, then you are now preventing the GC from freeing it and hence increasing the memory footprint. Greetings, Joachim -- Joachim "nomeata" Breitner mail@joachim-breitner.de | nomeata@debian.org | GPG: 0x4743206C xmpp: nomeata@joachim-breitner.de | http://www.joachim-breitner.de/