Too long to tweet, too short to blog.
Oct 08
-
<cdleary-lappy>
Waldo: what's the canonical way of detecting whether you're being called from a constructor?
-
<Waldo>
um
-
<cdleary-lappy>
s/from a/as a
-
<Waldo>
cdleary-lappy: what's regexp_call do?
-
<Waldo>
(it calls a common helper as I recall)
-
<cdleary-lappy>
I meant in JS itself -- is there an idiom for that?
-
<Waldo>
you can't
-
<Waldo>
you can sometimes know you *weren't* called that way
-
<Waldo>
but even then not always
-
<cdleary-lappy>
what's the "not always" case?
-
<cdleary-lappy>
(probably a discussion for #jslang, but w/e)
-
<Waldo>
function C() { if (!(this instanceof C)) throw "not new"; }; var c = new C; C.call(c);
-
<cdleary-lappy>
ah, tricky
-
<Waldo>
function C() { if (!(this instanceof C) || this.constructed) throw "not new"; Object.defineProperty(this, "constructed", { value: true }); }; var c = new C; C.call(c) /* *will* throw */;
-
<Waldo>
that's the only hackaround I know
-
<cdleary-lappy>
Waldo: seems like reasonable usage will work with the simple instanceof test, though
-
<Waldo>
yes
-
<cdleary-lappy>
constructing an invoking separately was something I wasn't even aware of! :-)
-
<Waldo>
it takes some work to get this in one's head
-
<cdleary-lappy>
what's annoying about it, though, is that you need the constructor binding hack to get arguments in the constructor after the check
-
<cdleary-lappy>
eh, I guess not, you can just make the two forms expect different arguments form, but it's still quite ugly
