resizeable arrays: T[new]

[Lionello Lunesu]

Manfred Nowak wrote:
> Walter Bright wrote
> 
>> The attractiveness of this 
>> solution is it nearly completely solves the problem with zero
>> runtime cost. 
> 
> Great thought.
> If incorporating this new type, please address also the complication 
> of having two views on the storage allocated for resizable arrays:
> - the "potential" capacity, to which the `length' property currently 
> can grow without the need of allocating more consecutive space
> - the "actual" capacity, which is currently set/get by the `length' 
> property
> 
> It is a pity, that the documentation, 
> http://www.digitalmars.com/d/arrays.html#resize,
> illustrates a pattern that can be easily incorporated into the 
> language as a default behaviour:
> - introduce new `length.pot' r/w-property
> - whenever the `length' property exceeds the `length.pot' property 
> double the `length.pot' property and allocate appropriate space
> - issue an error when `length.pot' is decreased below the value of 
> `length'

I agree. Although I'd keep the syntax less freaky by adding a property 
directly to arrays, array.capacity makes the most sense IMHO. It can be 
write-only for all I care.

Honestly, the array.length=12; array.length=0; -pattern (if you can call 
it that) is horrendous and should have been taken care of before v1.0 
was released, in my not so humble opinion.

L.