Tango vs Phobos

[bearophile]

Sean Kelly:
> It's just a matter of opinion, I suppose.  As D is a systems language, I don't
> agree with many arguments about safety when that safety conflicts with my
> ability to control what the language is doing behind the scenes.

I agree that:
- a system language has to be efficient;
- there must be a short, readable and nice way to disable heap allocation of closures (that I think has to happen by default);
- that sometimes safety has to be put aside (that's why I have said "practically possible": if a safe feature slows down code too much it's not much fit for a system language, especially when compiling in release mode).
- I'm using D instead of Python because sometimes I need a modern language that runs my programs faster (this isn't a system purpose).

Despite that, there are tons of situations where a "fast language" can help avoid bugs with little or no performance penalty (and adding a statement to disable the heap allocation of closures leads to no performance penalty).

Examples of other possible improvements:
foreach (i, x in array)
instead of:
foreach (i, x; array)
(But people here have explained me that despite this being better for the programmer it's worse for the compiler, making it more complex etc, and in the end a worse compiler is bad for the programmer too).

Or a switch() statement with a semantic safer than then C switch, or various other things that help you decrease bug count and on average produce a correctly running code in less time :-)


> Portability between versions is also an issue--code that is correctly designed for D 1.0 may be unusable on D 2.0 because the default behavior for certain language features is different, although the syntax is still completely legal.<

I have written a library that despite being surely tiny compared to Tango, is currently already 817 KB of code, with no redundancy inside, and it's written for D 1.x, so I understand your pain of porting things to D 2.x.

Despite this, D is a young language still, and it's developing quickly still, so it's too much early now to impose too many backward constraints... D 1.x may die and be forgotten, and maybe D 2.x too (Walter already says that AST macros are for D 3.x) so every current big lib for D will probably need one or two major rewrites. In 5 years we may have CPUs with several cores, that the currently mostly-sequential code used in my libs and Tango will be partially obsolete, that may require large changes in the libs and maybe in the language too...

Bye,
bearophile