Formal review of std.buffer.scopebuffer

[Walter Bright]

On 3/17/2014 11:10 AM, Dicebot wrote:
> 1)
> Walter has been pushing for getting this through the review queue to the point
> where I needed to ask Brian to delay voting for his module and switch to
> proceeding with Walter's. It didn't do any harm this time as Brian got busy
> anyway but I am very unhappy that I even had to do it.
>
> Now it suddenly gets cancelled and merged, internal or not (the very existence
> of std.internal rings a bell but it is a different story). Why bother me and
> push on Brian if you are just going to hurry merge it?

The ongoing threads about it made it clear that it was never going to happen as 
std.buffer.scopebuffer. I had assumed you were monitoring that, and I shouldn't 
have assumed so. I apologize.

The point of moving it to std.internal was so it would not be a documented 
feature of Phobos. It does take care to use successfully, and the move was in 
response to concerns that this would make it not in the spirit of Phobos.


> 2)
> There has been several very important concerns raised by monarch_dodra about how
> this specific implementation fits into D type system. He still finds absolutely
> horrible lines of code in that PR thread right here and now. I am absolutely
> ashamed of the fact that we have now non-legacy code in Phobos that breaks the
> immutable/const system (most recent finding).

The objective technical issues raised were all addressed, and the 
immutable/const one was corrected and unittests added for it before it was pulled.


> Some of such concerns has been straight rejected with appeal to authority and
> those who asked have been treated as if it is their guilt. You can't both try to
> sell D as community project and practice such workflow.

Some of the issues were subjective, where there are no clearly right or wrong 
answers, and a decision needs to be made at some point.


> 3)
> I have been asking in that PR why this proposal is even considered urgent when
> it looks like unexpected emergency focus put in completely wrong moment, before
> addressing basic issue of same domain. It wasn't only my question but also
> seconded by Martin Nowak. There is only one answer from Walter which does not
> actually answer any of those questions but continues that ridiculous
> "performance-performance-performance" main line.
>
> This looks terribly like panic-driven development.

ScopeBuffer has been there and commented on for about 2 months now. At last 
count it had over 4 comments per line of code. It is probably the most reviewed 
PR ever.

It is necessary to resolve a serious issue we have with Phobos that comes up 
constantly in public discussions about D. At some point we've got to move on and 
resolve this.

For reference, the two threads about it are:

https://github.com/D-Programming-Language/phobos/pull/1911

https://github.com/D-Programming-Language/druntime/pull/739