I'm very well aware of this and I totally agree that you cannot do this
for all "constants".
I thought it obvious and didn't mention that constants has to be side
effect free.
On 2014-10-03 04:32, Richard A. O'Keefe wrote:
On 3/10/2014, at 4:03 AM, Björn-Egil Dahlberg wrote:
Is there any other potential literals that are needed. I don't think so,
perhaps only solve this for fun exports and not the general case .. If it needs
solving that is.
Just because a value *could* be explicitly written as
a literal doesn't mean it is *easy* to do so.
Don't make the mistake of thinking that "is a literal" and
"is evaluated only once" are the same. They aren't.
Suppose you want to have a list of references that is
going to be requested often.
You cannot write a reference as a literal at all.
You *can* fake it by doing
B = term_to_binary([make_ref() || I <- lists:seq(1, 100)])
then pasting that into the skeleton
refs() -> binary_to_term(
% paste here
).
But it would be better if you could write
-once.
refs() ->
[make_ref() || I <- lists:seq(1, 100)].
I note that if something is a literal in Erlang I can use
it in a pattern. This is not true of fun mod:name/arity.
With the patch we are discussing to make these things that
can be stored at load time, they still are not *LITERALS*.
Think of the possibilities of tree-structured sets and
dictionaries; you *can* break encapsulation and write those
as explicit ground terms, but you would be mad to do so.
Personally, I've wanted to do that several times.
_______________________________________________
eeps mailing list
[email protected]
http://erlang.org/mailman/listinfo/eeps
