Blog Post: Beating std::visit Without Really Trying

[Seb]

On Sunday, 6 October 2019 at 02:33:15 UTC, Walter Bright wrote:
> On 10/5/2019 6:58 AM, Seb wrote:
>> Phobos is essentially dead/frozen (feature-wise).
> I beg to disagree. A couple cases in point:
>
>     https://github.com/dlang/phobos/pull/7211
>
> which is a re-imagining, rethinking of hexString.

I'm sorry, my choice of words wasn't wise (and I certainly didn't 
mean it in a negative way). Phobos is amazing and stable, but 
exactly because of these attributes there isn't much active 
development happening. Stability implies no breaking changes 
which is **very very good** because it allows the ecosystem to 
thrive (and not break builds because Algebraic was swapped 
against SumType).

Below I have listed a summary of all announced changes in Phobos 
since the beginning of this year (six releases, 2.088 - 2.083). 
This lists only the announced changes (and there are more bug 
fixes), but in total we're looking at one new function 
(getAvailableDiskSpace), two new overloads (schwarzSort, 
unpredictableSeed) over these six releases. That's what I was 
referring to.
For comparison: in the meantime about 200 new Dub packages have 
been added [1, 2] and in total about 600 Dub packages have been 
updated [3].


2.088:
- std.array.Appender and RefAppender: use .opSlice() instead of 
data()
- ErrnoException.errno is now nothrow pure @nogc @safe
- Nullable alias get this has been deprecated
- Added the std.file.getAvailableDiskSpace functionality.
- Allow std.json to overlook trailing comma

2.087:
- Added a table of control characters in ASCII table
- Count processors via sched_getaffinity on Linux
- Add overload std.algorithm.sorting.schwartzSort!(alias 
transform, SwapStrategy ss, R)
- Phobos is now compiled with -preview=dip1000

2.086:
- Fixed comparison bug in 
std.algorithm.comparison.levenshteinDistance
- std.experimental.all has been moved to std

2.085: (no changes)

2.084:
- Add overload std.random.unpredictableSeed!UIntType

2.083: (no changes)

See more at [4].

[1] 
https://web.archive.org/web/20181229043818/http://code.dlang.org/
[2] https://code.dlang.org/
[3] 
https://code.dlang.org/?sort=updated&category=&skip=500&limit=100
[4] https://dlang.org/changelog/index.html

> and:
>
>     https://github.com/dlang/phobos/pull/7130
>     https://github.com/dlang/phobos/pull/7144
>
> both of which work to remove autodecode from Phobos. 7130 in 
> particular can use some help with anyone who wants to help 
> drive this forward.

I'm sorry and while I like this effort, I don't expect it to help 
because it does break the world.
AFAICT the only way to avoid breaking the world when removing 
autodecode is to ship a new/different version of Phobos 
_together_ with the current one, s.t. it can be opted-in by users.

Anyhow, I think Andrei summarized it better a few months ago:

> The time is ripe for std.v2 
> [https://forum.dlang.org/post/q7j3s0$15n7$1@digitalmars.com]

My earlier post tried to point out that SumType is an excellent 
candidate for v2.