Jascha Wetzel wrote:
> ok, the const part is missing, but the performance part can be dealt
> with. maybe "const inout" is a worthwhile feature for future D versions.
> it should have a less contradictory name, though.

Hey, no fair correcting yourself before I finish my post! ;)

Seriously: as I mentioned in my post there was some discussion a while back about adding this sort of stuff, but so far nothing concrete has shown up.

;)

maybe we could have an additional "inref" modifier. on the other hand
"out" doesn't force write-only either, atm.
therefore i think having a way to return references/lvalues is more
important on that front.
i'd really like to do matrix[row,col] += bla;

Frits van Bommel wrote:
> Jascha Wetzel wrote:
>> ok, the const part is missing, but the performance part can be dealt with. maybe "const inout" is a worthwhile feature for future D versions. it should have a less contradictory name, though.
> 
> Hey, no fair correcting yourself before I finish my post! ;)
> 
> Seriously: as I mentioned in my post there was some discussion a while back about adding this sort of stuff, but so far nothing concrete has shown up.

[corrected upside-down post]

Jascha Wetzel wrote:
> Frits van Bommel wrote:
>> Jascha Wetzel wrote:
>>> ok, the const part is missing, but the performance part can be dealt
>>> with. maybe "const inout" is a worthwhile feature for future D versions.
>>> it should have a less contradictory name, though.
>> Hey, no fair correcting yourself before I finish my post! ;)
>>
>> Seriously: as I mentioned in my post there was some discussion a while
>> back about adding this sort of stuff, but so far nothing concrete has
>> shown up.
>
> ;)
>
> maybe we could have an additional "inref" modifier. on the other hand
> "out" doesn't force write-only either, atm.

'out' does force initialization though. Before anything can be read from or written to the parameter, it's automatically set to the initial value of its type. So you can read it, but it won't have the value it had in the context of the caller (unless, of course, that value was the initial value :) ).

> therefore i think having a way to return references/lvalues is more
> important on that front.

This was IIRC also a topic in the discussion I mentioned.

> i'd really like to do matrix[row,col] += bla;

So would I. Though I think that syntax is currently possible with proxy objects, that's not the most elegant solution.