www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - alias two froms

reply "eles" <eles215 gzk.dot> writes:
See this:

http://forum.dlang.org/post/kfdkkwikrfvaukhctrmq forum.dlang.org

"alias" supports two syntaxes, one of them specifically to 
address writing things like alias A this.

That's inconsistent. I agree not the most urgent thing in the 
world, but while the fixing things happens (see the  property), 
why not address this too?

So? Deprecate the old syntax?
Sep 11 2014
next sibling parent reply "Peter Alexander" <peter.alexander.au gmail.com> writes:
On Thursday, 11 September 2014 at 19:08:27 UTC, eles wrote:

 See this:

 http://forum.dlang.org/post/kfdkkwikrfvaukhctrmq forum.dlang.org

 "alias" supports two syntaxes, one of them specifically to 
 address writing things like alias A this.

 That's inconsistent. I agree not the most urgent thing in the 
 world, but while the fixing things happens (see the  property), 
 why not address this too?

 So? Deprecate the old syntax?


This was discussed recently.

The problem is that the new syntax is only a few versions old, so 
deprecating the old syntax means breaking all D code that's more 
than a year old.

If D had no existing customers then yeah, we'd remove it, but I 
think it's too early to start deprecation.
Sep 11 2014
next sibling parent ketmar via Digitalmars-d <digitalmars-d puremagic.com> writes:
On Thu, 11 Sep 2014 20:06:15 +0000
Peter Alexander via Digitalmars-d <digitalmars-d puremagic.com> wrote:


 If D had no existing customers then yeah, we'd remove it, but I=20
 think it's too early to start deprecation.


so people can write even more old-styled code...
Sep 11 2014
prev sibling next sibling parent "eles" <eles215 gzk.dot> writes:
On Thursday, 11 September 2014 at 20:06:17 UTC, Peter Alexander 
wrote:

 On Thursday, 11 September 2014 at 19:08:27 UTC, eles wrote:

 See this:





 so deprecating the old syntax means breaking all D code that's


Breaking? :-o Why?

IIRC, D uses as first stage a strange "scheduled for deprecation" 
that won't break anything, except will print a warning.

And, when would you plan to start deprecation? Next year? After 
the GC is precised? Why keep things in limbo for so long? I see 
no practical reasons to mark them as "scheduled for deprecation".
Sep 11 2014
prev sibling parent "Mike" <none none.com> writes:
On Thursday, 11 September 2014 at 20:06:17 UTC, Peter Alexander 
wrote:

 The problem is that the new syntax is only a few versions old, 
 so deprecating the old syntax means breaking all D code that's 
 more than a year old.

 If D had no existing customers then yeah, we'd remove it, but I 
 think it's too early to start deprecation.


If the desire is to deprecate something in the language, it is 
best to start early to prevent the problem from getting worse.

I believe this could be done gradually with little, or no 
disruption.

1. Create a dfix utility to automate updating the alias syntax,
2. Update documentation to fix all examples using the old syntax
3. Update documentation stating that while both syntaxes are 
allowed the latter is preferred
4. Using the dfix utility, update phobos, druntime, and 
potentially other projects in the D ecosystem
5. Give users a few versions to let it sink in
6. Mark old syntax as deprecated so users get a warning
7. Give users a few versions to adjust
8. Finally deprecate old syntax

This recipe could potentially be used to fix many annoyances in 
the language besides the one that started this thread.  But it 
first requires someone (like me) to first create the dfix 
utility.  The existence of such a tool could potentially change 
D's culture as well.  Automating the upgrade of the alias syntax 
would set an important precedent.

Mike
Sep 11 2014
prev sibling next sibling parent reply "Brian Schott" <briancschott gmail.com> writes:
On Thursday, 11 September 2014 at 19:08:27 UTC, eles wrote:

 So? Deprecate the old syntax?


I tried to get people to agree to deprecate an old syntax that 
was worse. I also wrote an automated upgrade tool that would fix 
people's code.

I couldn't get it approved. As far as I can tell, the official 
position is that parser bugs and language inconsistencies are 
permanent.
Sep 11 2014
parent reply "Mike" <none none.com> writes:
On Thursday, 11 September 2014 at 20:14:47 UTC, Brian Schott 
wrote:

 On Thursday, 11 September 2014 at 19:08:27 UTC, eles wrote:

 So? Deprecate the old syntax?


 I tried to get people to agree to deprecate an old syntax that 
 was worse. I also wrote an automated upgrade tool that would 
 fix people's code.

 I couldn't get it approved. As far as I can tell, the official 
 position is that parser bugs and language inconsistencies are 
 permanent.


I share your frustration.

I saw your code to automate a fix in support of DIP65, but I 
haven't seen any code to automate updating the alias syntax.  If 
you have this on hand would be willing to share it?

Mike
Sep 11 2014
parent "Brian Schott" <briancschott gmail.com> writes:
On Thursday, 11 September 2014 at 22:54:49 UTC, Mike wrote:

 I saw your code to automate a fix in support of DIP65, but I 
 haven't seen any code to automate updating the alias syntax.  
 If you have this on hand would be willing to share it?

 Mike


dfix will remain unwritten until Walter is willing to make the 
decisions that would require it. I have several other things to 
work on in my spare time, and those other things will actually be 
used.

In the meantime you can use D-Scanner to find instances of old 
alias declarations (and other things) in your code.
Sep 11 2014
prev sibling parent reply =?UTF-8?B?QWxpIMOHZWhyZWxp?= <acehreli yahoo.com> writes:
On 09/11/2014 12:08 PM, eles wrote:

 See this:

 http://forum.dlang.org/post/kfdkkwikrfvaukhctrmq forum.dlang.org

 "alias" supports two syntaxes, one of them specifically to address
 writing things like alias A this.

 That's inconsistent. I agree not the most urgent thing in the world, but
 while the fixing things happens (see the  property), why not address
 this too?

 So? Deprecate the old syntax?


The 'alias this = A' syntax did appear for one release as an 
unintentional feature:

 
http://forum.dlang.org/thread/aaflopktcjmljxdnoizj forum.dlang.org#post-aaflopktcjmljxdnoizj:40forum.dlang.org

Ali
Sep 11 2014
parent reply "eles" <eles eles.com> writes:
On Thursday, 11 September 2014 at 23:05:19 UTC, Ali Çehreli wrote:

 On 09/11/2014 12:08 PM, eles wrote:



 The 'alias this = A' syntax did appear for one release as an 
 unintentional feature:


Having two syntaxes for alias makes everything looking worse and 
is an unnecessary noise.

I would rather go with something like:

alias alpha~=beta;

if I would need multiple aliases (see alpha as a list of aliased 
types).
Sep 11 2014
parent Nick Treleaven <ntrel-public yahoo.co.uk> writes:
On 12/09/2014 07:48, eles wrote:

 On Thursday, 11 September 2014 at 23:05:19 UTC, Ali Çehreli wrote:

 On 09/11/2014 12:08 PM, eles wrote:



 The 'alias this = A' syntax did appear for one release as an
 unintentional feature:


 Having two syntaxes for alias makes everything looking worse and is an
 unnecessary noise.


'alias this' is different from normal aliasing. The 'alias ... this' 
syntax should be replaced, but we haven't agreed on by what. I think 
Timon Gehr's opThis solution is best:

http://forum.dlang.org/thread/xjevtasyafnkvusprldp forum.dlang.org?page=2#post-ld5c9r:2418lt:241:40digitalmars.com

Using the opThis identifier (or similar) avoids having to introduce a 
separate 'get' property function when you need to add const qualifiers 
to the forwarded symbol, or do some processing first.

P.S. As I noted in that thread, the suggested syntax for forwarding 
constructors is not so ambiguous, it should be:

class D : C {
     alias this = super.this;	// note .this to name base ctor
}
Sep 12 2014