Is it a good idea to introduce in D a "fat delegate" that has a variable length (but its length is known at compile-time), able to contain two or more context pointers and/or one or more values or references, avoiding both some heap allocations of the closure and allowing coding patterns that require two or more context pointers?

Bye,
bearophile

On Tuesday, 16 December 2014 at 13:26:56 UTC, bearophile wrote:
> Is it a good idea to introduce in D a "fat delegate" that has a variable length (but its length is known at compile-time), able to contain two or more context pointers and/or one or more values or references, avoiding both some heap allocations of the closure and allowing coding patterns that require two or more context pointers?
>
> Bye,
> bearophile

http://wiki.dlang.org/DIP30

deadalnix:

> http://wiki.dlang.org/DIP30

It seems DIP30 proposes a struct with a pointer as second field. Are the delegates of DIP30 usually @nogc? They don't seem variable-length structs.

Bye,
bearophile

On Tuesday, 16 December 2014 at 14:13:23 UTC, bearophile wrote:
> deadalnix:
>
>> http://wiki.dlang.org/DIP30
>
> It seems DIP30 proposes a struct with a pointer as second field. Are the delegates of DIP30 usually @nogc? They don't seem variable-length structs.
>
> Bye,
> bearophile

The first element can be a struct with 2 pointers.