Life in the Fast Lane (@nogc blog post)

[Adam D. Ruppe via Digitalmars-d-announce]

On Saturday, 17 June 2017 at 07:03:53 UTC, Petar Kirov 
[ZombineDev] wrote:
> 2. However, there's this long list of things that you have to 
> avoid.

There's like 10 things to avoid in the language itself: 
http://dlang.org/spec/garbage.html#op_involving_gc

and most of them are obviously array stuff so it is easy to 
remember - and easy to skip with a simple array helper... which 
is often a useful optimization too, make one that uses stack 
space! You can write one in about 15 minutes.


Now, @nogc is a bit more restrictive since it isn't just avoiding 
GC calls.... it avoids the *possibility* of GC calls. Even if it 
would only actually run one in a million times at which point you 
want the program to die anyway.... @nogc still bans it. Even if 
it would never actually run, but some leaf function somewhere 
wrote `int getLength() { return _length; }` instead of `int 
getLength() @nogc pure const @safe nothrow { return _length; 
}`... it obviously doesn't run the gc, but still fails @nogc for 
its entire call tree.


This is why I consider @nogc to be an *extremely* niche feature 
and generally harmful. It makes things look a lot harder than it 
really is - people confuse `@nogc` with "no gc". Instead, I 
suggest just minding the list and using runtime profiling and 
debugging to ensure your needs are met in actual practice.

> 3. So while this "@nogc" is technically possible you loose so 
> much of the language that you're better of with C++ (where e.g. 
> "cout" just works*).

of course you could always `printf` lol.

The reason writeln fails @nogc is that it *might* throw an 
exception with most args if stdout is closed or something.

Perfect example of an *extremely* rare case failing @nogc's 
ridiculously strict requirements.