www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Expression-bodied functions

reply Gabriel Garcia <garciat live.com> writes:
Hello all (: This is my first post here.

I don't know if this has been proposed before. Probably, it has 
been. Either way...

https://github.com/dotnet/roslyn/wiki/New-Language-Features-in-C%23-6#expression-bodied-function-members

Given D's tendency toward expression-oriented (or functional, 
whatever) programming, this seems like a logical addition to the 
language to me.

Cheers.
Dec 31 2015
next sibling parent reply Timon Gehr <timon.gehr gmx.ch> writes:
On 12/31/2015 06:15 PM, Gabriel Garcia wrote:

 Hello all (: This is my first post here.

 I don't know if this has been proposed before. Probably, it has been.
 Either way...

 https://github.com/dotnet/roslyn/wiki/New-Language-Features-in-C%23-6#expression-bodied-function-members


 Given D's tendency toward expression-oriented (or functional, whatever)
 programming, this seems like a logical addition to the language to me.

 Cheers.


https://issues.dlang.org/show_bug.cgi?id=7176
Dec 31 2015
parent reply Gabriel Garcia <garciat live.com> writes:
On Thursday, 31 December 2015 at 17:46:16 UTC, Timon Gehr wrote:

 On 12/31/2015 06:15 PM, Gabriel Garcia wrote:

 Hello all (: This is my first post here.

 I don't know if this has been proposed before. Probably, it 
 has been.
 Either way...

 https://github.com/dotnet/roslyn/wiki/New-Language-Features-in-C%23-6#expression-bodied-function-members


 Given D's tendency toward expression-oriented (or functional, 
 whatever)
 programming, this seems like a logical addition to the 
 language to me.

 Cheers.


 https://issues.dlang.org/show_bug.cgi?id=7176


I would really like to see this moving forward. What would it 
take? A PR on GitHub? The attached Proof of Concept seems like a 
decent start. I am not familiar at all with the codebase, though 
the patch is quite small and understandable. If nobody with 
actual credentials or experience steps up in the coming 
days/weeks, I suppose I could give it a try.

FWIW, I've given the issue 20 votes.
Dec 31 2015
parent Jack Stouffer <jack jackstouffer.com> writes:
On Thursday, 31 December 2015 at 19:52:25 UTC, Gabriel Garcia 
wrote:

 I would really like to see this moving forward. What would it 
 take?


First off you would have to convince the language maintainers 
that it's a good idea. I see two major contributors, Kenji Hara 
and Jonathan M Davis, are against it in that thread.
Dec 31 2015
prev sibling parent reply ZombineDev <valid_email he.re> writes:
On Thursday, 31 December 2015 at 17:15:16 UTC, Gabriel Garcia 
wrote:

 Hello all (: This is my first post here.

 I don't know if this has been proposed before. Probably, it has 
 been. Either way...

 https://github.com/dotnet/roslyn/wiki/New-Language-Features-in-C%23-6#expression-bodied-function-members

 Given D's tendency toward expression-oriented (or functional, 
 whatever) programming, this seems like a logical addition to 
 the language to me.

 Cheers.


By the way, a very similar language feature was approved and 
implemented for the upcomming DMD 2.070 release. It allows you to 
use a short syntax for function templates:

alias add = (a, b) => a + b;
// Shorthand for:
// auto add(T1, T2)(T1 a, T2 b) { return a + b; }

IFTI makes it look like a regular function call:
writeln(add(3, 4.5));

It works for free functions, but I'm not sure if it would also 
work for member-functions (methods).

https://issues.dlang.org/show_bug.cgi?id=12421

http://dlang.org/spec/declaration.html#AliasDeclaration
(The relevant part is:
AliasDeclarationY:
     Identifier TemplateParametersopt = FunctionLiteral)

https://github.com/D-Programming-Language/dmd/commit/dbba59cadafe540c0c98d69ac867173335e145cd
Dec 31 2015
parent reply Enamex <enamex+d outlook.com> writes:
On Friday, 1 January 2016 at 00:27:49 UTC, ZombineDev wrote:

 By the way, a very similar language feature was approved and 
 implemented for the upcomming DMD 2.070 release. It allows you 
 to use a short syntax for function templates:

 alias add = (a, b) => a + b;
 // Shorthand for:
 // auto add(T1, T2)(T1 a, T2 b) { return a + b; }

 IFTI makes it look like a regular function call:
 writeln(add(3, 4.5));

 It works for free functions, but I'm not sure if it would also 
 work for member-functions (methods).

 https://issues.dlang.org/show_bug.cgi?id=12421


IMHO, anything like that is just a tangential 'fix' to the fact 
that D's design rests on commands like 'return' to leave scope 
and return values.

Something like

auto add(auto x, auto y) { x + y }

would be much more wholesome IMO. (The 'funcName(auto arg)' 
syntax is just a convenience suggestions; the important part is 
'{ x + y }').

That said, it does fix something that 'alias' should've been able 
to do from the beginning. I'm not sure what else but I think 
there other places where alias needs '= I!(stuff)' identity 
template to accept the 'stuff'.
Jan 02 2016
parent ZombineDev <valid_email he.re> writes:
On Saturday, 2 January 2016 at 15:04:15 UTC, Enamex wrote:

 On Friday, 1 January 2016 at 00:27:49 UTC, ZombineDev wrote:

 By the way, a very similar language feature was approved and 
 implemented for the upcomming DMD 2.070 release. It allows you 
 to use a short syntax for function templates:

 alias add = (a, b) => a + b;
 // Shorthand for:
 // auto add(T1, T2)(T1 a, T2 b) { return a + b; }

 IFTI makes it look like a regular function call:
 writeln(add(3, 4.5));

 It works for free functions, but I'm not sure if it would also 
 work for member-functions (methods).

 https://issues.dlang.org/show_bug.cgi?id=12421


 IMHO, anything like that is just a tangential 'fix' to the fact 
 that D's design rests on commands like 'return' to leave scope 
 and return values.

 Something like

 auto add(auto x, auto y) { x + y }

 would be much more wholesome IMO. (The 'funcName(auto arg)' 
 syntax is just a convenience suggestions; the important part is 
 '{ x + y }').


Personally, I prefer the new alias syntax, (which sort of reminds 
me of Haskell), because it looks much cleaner than this mashup of 
Swift and C++14.


 That said, it does fix something that 'alias' should've been 
 able to do from the beginning. I'm not sure what else but I 
 think there other places where alias needs '= I!(stuff)' 
 identity template to accept the 'stuff'.


The most obvious thing that comes to my mind are values (that 
enum accepts):
enum number = 3;
alias aliasedNumber = I!(3);

However wouldn't this blur the line between enum and alias too 
much?
Right now enum can refer only to values computed at compile-time, 
while alias can only refer to symbols*, which I think is a good 
separation.

* Alias to a function literal refers to the _symbol_ of the 
anonymous function template.

Maybe a good direction forward would be to allow alias to refer 
to expressions (i.e. similar to the AST macros idea), which you 
can introspect at CT (like with CT type reflection).

auto query = db.orders
     .where! {price > 200}
     .select!(order => order.name);

 {...} looks like q{...} but would capture the AST of the 
expession between the braces (which must be valid D code).

auto where(alias predExpr, R)(R range)
     if (isInputRange!R &&
         isAstCapture!predExpr)
{
     // http://dlang.org/spec/expression.html#OrOrExpression
     static assert(__traits(isExpression, OrOrExpression, 
predExpr);

     // prepare where clause of SQL statement...
}
Jan 02 2016