You can see some margin notes I added this morning about stuff I
was thinking about saying to fill time!
For example, being willing to experiment helps with speed, like
Walter said. It also applies here: disassemblies might be scary
the first time, but if you dive into it and see about toying with
it, just tinker, guess and check your way to better benchmarks...
you'll probably find success before too long.
And the key point with the "just try it gnashing of teeth" note
is that you really don't have much to lose! Except time perhaps,
but once you've done it a few times you'll be able to find
shortcuts. And besides put a time limit on yourself to make sure
you don't waste your whole life and you'll be ok.
LOL up top I called myself a "web/dynamic god". See what I said
about my ego? I praise me even in notes to self!
Up top, I made a note to self after Don's talk, where he talked
about getting serious with industrial use. Well, I say back to
fun... but at the same time, knowing the whole library is useful.
Not only is it good to know the underlying stack when
abstractions start leaking, but there's actually practical
applications to playing with the runtime:
1) Special performance or size requirements (Sociomantic, I think
Mike Parker did that for some games too, or embedded stuff)
2) Organization-wide lint kind of stuff. You can use a custom
object.d for RTInfo for example to do checks and not worry about
third-party integration so much because you are only interested
in "works for me" without the hassle of redistribution.
3) Running D in places like a platform that only has a C compiler
was dtojs forked off Daniel Murphy's D->C converter that first
got me playing without the runtime) might find use of these
functions being implemented differently or ignored.
So while I still tend to say half the stuff I mentioned there is
useless info, the general idea of being willing to play with it
might not be - you never know when one of these random ideas will
* * *
Two other margin notes that came up during the conference: Walter
mentioned UFCS with the first argument as the "this" thing....
but when you do a method on a struct, the this reference is
actually passed as a (hidden) last argument! So I just lol'd a
bit at it not being quite so uniform.
The other one is moving the language to object.d. For example,
some of the chatter after Don's talk was about size_t being
unsigned. So annoying! I think it would be kinda nice if that
could actually be redefined somehow. I've also talked before
about how it might be cool to to "alias int = __int;" to use
builtin ints or "int = RangedChecked!(0, 100);" to make the int
type behave differently in the case of that custom runtime.
To my surprise though, I ran out of time and even had to cut some
questions short :( Oh well, y'all can always find me here.
PS sorry I had to run out early today during the lightning talks!
My ride back to the airport was already there, they wanted to
avoid rush hour traffic.