consequences of removing semicolons in D like in Python

[cym13 via Digitalmars-d]

On Saturday, 17 September 2016 at 11:03:14 UTC, Jonathan M Davis 
wrote:
> On Saturday, September 17, 2016 10:15:57 eugene via 
> Digitalmars-d wrote:
>> On Saturday, 17 September 2016 at 02:21:54 UTC, Chris M. wrote:
>> > Pointless and not worth breaking everyone's code over
>>
>> why?
>
> Having semicolons actually serves as a synchronization point 
> for the compiler, which helps it give better error messages. 
> Sure, trying to use newlines that way can work, but it's more 
> problematic - especially when you take stuff like multiline 
> statements into account. D - like C/C++/Java/C# - is not 
> whitespace sensitive, and like those languages, it uses 
> semicolons. There really is no technical reason why getting rid 
> of semicolons improves things, and it does make them at least 
> slightly worse as far as the compiler goes. It's just cleaner 
> to have semicolons, and most D programmers have no problem with 
> that - even prefer it.

There are also language constructs with semicolons inside, I'm 
not sure it would add that much complexity, but there are some 
cases that would be ambiguous:

     void filterEven() {
         print("Nothing here")
     }

     auto filterEven(int[] arr) {
         import std.algorithm : filter
         return arr.filter!(x => x%2)
     }

     void main() {
          auto i = [1, 2, 3]
                           .filterEven()
     }

Note how a leading dot means “global scope” but a dot after 
something means UFCS or method/attribute. What should this 
program do? If it is akin to “auto i = [1, 2, 3]; .filterEven();” 
then i is an int[] and the program prints “Nothing here”. But if 
it is understood as “auto i = [1, 2, 3].filterEven();” then i is 
whatever type filter returned and nothing is printed.

> In addition, needlessly altering D's syntax like that would 
> make porting C/C++ code to it much harder, which would be a 
> significant problem for a decent chunk of our user base. D's 
> syntax tends to stick with C/C++'s syntax unless it has a 
> compelling reason to do otherwise.

That's the really good reason. Semicolons are a cultural choice 
more than a technical one but once we've made that choice there's 
no turning back without loosing nearly all users.

Besides the ability to port code easily from one language to 
another is important. I have in mind this recent article “Why I'm 
dropping Rust” 
(https://hackernoon.com/why-im-dropping-rust-fd1c32986c88). The 
author tried to port a C++ library but as language constructs are 
different he wasn't able to do so without having to rethink the 
architecture from scratch. We want it to be easy to port any 
project to D.

> And regardless of whether not having semicolons would be better 
> or not, it would be a huge breaking change to get rid of them, 
> which isn't even vaguely worth it. Sure, D continues to 
> involve, but it's reasonably stable, and we want it to become 
> _more_ stable, not less. It's _way_ past the point that we 
> would consider making a change like this. Doing something like 
> removing semicolons from the language would alienate a huge 
> portion of our user base. Most would consider the change to be 
> objectively worse, and even if they didn't care about whether D 
> had semicolons or not, they would _not_ be pleased to have 
> their code break.
>
> - Jonathan M Davis