make all semicolons optional, there is no need in them

[antonkulaga]

Modern languages like Scala and Groovy do not enforce users to write semicolons. I think it is a right decision, because in fact semicolons are nothing more than just a redundant typing.
I've been a scala developer for many years and I only needed semicolons when putting several lines of code in one line to fit the code in the slide=)

[Gankra]

This was actually implemented last month! 🎆

https://github.com/mystor/slag

[antonkulaga]

@gankro thank you! I did not know about that, RustDTcomplains about missing semicolons and I did not look at nighty builds before opening this issue

[makoConstruct]

😞 those fat arrows.. It seems to me that it'd be possible to do a syntax that is very very similar to Rust's primary syntax but where all braces and semicolons are optional(or, only used when needed, EG, when indentation isn't sufficient)

One of the reasons the current situation makes this difficult:
A naive solution would parse

fn do_things(a:u32 b:u32)-> u32 { a*b + a }
fn main()
    do_things(1 2)
    do_things(3 4)

into

fn do_things(a:u32, b:u32)-> u32 { a*b + a }
fn main(){
    do_things(1, 2);
    do_things(3, 4)
}

See how it leaves the last statement of a block without a semicolon to make sure it gets returned. Unfortuantely, there are many cases where we do want to return something and many cases where we don't. This is a case where we don't, and because do_things(3,4) returns a non-unit value we get a compilation error, main wants a unit at the end, do_things didn't give it one without the semicolon.

I read somewhere that allowing all types to coerce to unit had been considered(tried, even?), but it was rejected. Having difficulty finding the discussion. I can't see why it's considered useful to tell the programmer off for leaving the semicolon off of the last (non-unit)expression in a block that returns unit. For every case where a programmer wrongly expects the last expression to transmit its value out of the block, there is a receiver. If the transmission is interrupted by a coercion to unit, the error will be pointed out at the receiver's end, when it gets unit instead of whatever it was expecting. Requiring it to be pointed out at the transmitter's doesn't seem helpful.

[makoConstruct]

@apasel422 sorry for the confusion, I accidentally hit ctrl+enter before I'd finished writing it. Should be comprehensible now.