Last call for AliasSeq

[deadalnix via Digitalmars-d]

On Monday, 27 July 2015 at 09:01:33 UTC, Marc Schütz wrote:
> On Monday, 27 July 2015 at 02:14:57 UTC, Jonathan M Davis wrote:
>> AliasTuple in particular has serious issues with it from the 
>> perspective of teaching people what it is an how to use it, 
>> because it has Tuple in its name,
>
> People keep claiming that, but have never posted any evidence. 
> We know that _TypeTuple_ had issues, but for all we know, the 
> problem was the "Type" part, not the "Tuple" part.
>

We have various reports that are consistent and confirm this is 
an issue. At this point, this is a repeatable experiment, not an 
anecdote anymore. Ignoring repeatable experiment puts you in the 
tinfoil hat section of the population. You don't want to be there.

>> and the construct in question is not actually a tuple
>
> And that's IMO another really bizarre claim. Of course it's a 
> tuple.
>

Technically you are right. But in fact, this is NOT what people 
EXPECT a tuple to be.

>> (in addition to being easily confused with 
>> std.typecons.Tuple). This has been shown time and time again 
>> with TypeTuple.
>>
>> On technical merit, AliasSeq is one of the better choices; it 
>> was what TypeTuple had been changed to prior to the recent, 
>> large discussion on it; and none of the new suggestions are 
>> better enough to win any kind of consensus.
>
> If ease of learning / avoidance of misunderstandings is a goal, 
> it's clear that AliasSeq is one of the worst choices. The term 
> "sequence" - in all other related areas of science or 
> engineering - invokes certain associations that don't fit this 
> thing at all. This has been brought up by several people, but 
> it's just getting ignored by the proponents.
>

It is not ignored. It is simply that alternative proposal also 
have issues, and many of them have issues that are worse. No 
proposal was significantly better so that it reached any kind of 
consensus.

No your personal favorite did not either.

> As for Andrei and Walter: From what I've seen, they, too, agree 
> that AliasSeq is bad. Andrei's critique even started that other 
> thread.

Yet, nobody came with anything significantly better, so here it 
is.