The Case Against Autodecode

[Andrei Alexandrescu via Digitalmars-d]

On 06/02/2016 09:55 AM, cym13 wrote:
> On Thursday, 2 June 2016 at 13:06:44 UTC, Andrei Alexandrescu wrote:
>> Your claim was obliterated, and now you continue arguing it by
>> adjusting term definitions on the fly, while at the same time
>> awesomely claiming to choose the high road by not wasting time to
>> argue it. I should remember the trick :o). Stand with the points that
>> stand, own those that don't.
>
>> Definitely. It's a fine line to walk; this particular decision is not
>> that much on the edge at all. We must stay with autodecoding.
>
> If you are to stay with autodecoding (and I hope you won't) then
> please, *please*, at least make it decode to graphemes so that it
> decodes to something that actually have some kind of meaning of its
> own.

That's not going to work. A false impression created in this thread has 
been that code points are useless and graphemes are da bomb. That's not 
the case even if we ignore the overwhelming issue of changing semantics 
of existing code.

> I think the real reason about why this isn't mentioned in the
> critics you mention is that people don't know about it. Most people
> don't even imagine it can be as broken as it is.

This should be taken at face value - rampant speculation. From my 
experience that's not how these things work.

> Heck, it even
> took Walter by surprise after years! This thread is the first real
> discussion we've had about it with proper deconstruction and
> very reasonnable arguments against it. The only unreasonnable thing
> here has been your own arguments. I'd like not to point a finger at
> you but the fact is that you are the only single one defending
> autodecoding and not with good arguments.

Fair enough. I accept continuous scrutiny of my competency - it comes 
with the territory.

> Currently autodecoding relies on chance only. (Yes, I call “hoping
> the text we're manipulating can be represented by dchars” chance.)
> This cannot be anymore.

The real ticket out of this is RCStr. It solves a major problem in the 
language (compulsive GC) and also a minor occasional annoyance 
(autodecoding).


Andrei