This Week in D #19: Reggae, ranges, and DConf Wed. Afternoon summary
Meta via Digitalmars-d-announce
digitalmars-d-announce at puremagic.com
Mon Jun 8 19:27:25 PDT 2015
On Monday, 8 June 2015 at 19:21:10 UTC, Adam D. Ruppe wrote:
> http://arsdnet.net/this-week-in-d/jun-07.html
>
> These dconf articles are taking a long time to write, hence the
> lateness again, but here's the rest of Wednesday and some
> roundup of changes from the forum and pull requests.
>
> I was a bit sloppy on the Q&A, I just copy/pasted the notes I
> took live and formatted them a little. In a few cases, I wasn't
> sure who the speaker was, so they are labeled 'idk', and my
> personal shorthand isn't literally what people said, but
> hopefully this will give people an idea of what was discussed
> without having to have been there.
>
> If someone else wants to post to reddit I'd appreciate it, I
> get trouble if I post my own thing too often.
>
>
> Over this next week, my life is starting to return to normality
> so I plan to write up Thursday and Friday for next week, then
> we have interviews queuing up for the following week! The next
> few weeks should be exciting issues.
It seems like every year, tuple unpacking comes up. The
functionality has been done for awhile now, so what is the main
blocker? It's fine to talk about getting things done and not
letting good work sit around, but we have multiple DMD pull
requests over 3 years old, and a large number over 2 years old
(https://github.com/D-Programming-Language/dmd/pulls?page=4&q=is%3Apr+is%3Aopen).
Andrei muses as to whether tuple unpacking is all we want, or if
there is some more general and powerful feature. There certainly
is, and most newer languages have it: pattern matching +
destructuring. The number of times that it's been mentioned on
Reddit alone should at least flag it for consideration. Neither
of these are critical to D, but first-class tuples and pattern
matching / destructuring are very important quality of life
issues and a great selling point for D. Even going with what we
currently have for template type-matching, but for values, would
be a great step in this direction.
More information about the Digitalmars-d-announce
mailing list