www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Conflicting UDA

reply =?UTF-8?B?TcOhcmNpbw==?= Martins <marcioapm gmail.com> writes:
I came across an issue with UDAs and was wondering if there 
really is no way or if I just missed something...

Basically, my library has an  ignore UDA, which conflicts with 
vibe.d's vibe.data.serialization.

If both mine and vibe's module are imported, DMD will fail with a 
very non-descriptive error message, seen below... The obvious 
solution would be to prefix my UDAs, but then I suppose every 
library writer would have to abide by this convention, which in 
practice won't happen, and would bring us back to the C-style 
redundant symbol names all over. It's unpleasant to have to 
disambiguate even when not necessary... I suppose I could also do 
 mylibattr("ignore") instead, but this is also hideous and overly 
verbose...

I tried  mylib.ignore, which would not be too bad, if necessary 
only to disambiguate, but it seems like the parser doesn't 
understand it.

struct Score {
	uint id;
	ulong score;
	 ignore ulong score2 = 10; // Error: expression ignore is void 
and has no value
}
static assert(hasUDA!(Score.score2, IgnoreAttribute));


Something like would be intuitive:

 mylib.ignore  vibe.ignore  std.somemodule.ignore ulong score2 = 
10;


What are other people doing to go around this? To me, this seems 
like a major UDA usability issue for library writers in the wild.
Feb 06 2016
parent reply Marc =?UTF-8?B?U2Now7x0eg==?= <schuetzm gmx.net> writes:
On Saturday, 6 February 2016 at 13:36:32 UTC, Márcio Martins 
wrote:

 I came across an issue with UDAs and was wondering if there 
 really is no way or if I just missed something...

 Basically, my library has an  ignore UDA, which conflicts with 
 vibe.d's vibe.data.serialization.

 If both mine and vibe's module are imported, DMD will fail with 
 a very non-descriptive error message, seen below... The obvious 
 solution would be to prefix my UDAs, but then I suppose every 
 library writer would have to abide by this convention, which in 
 practice won't happen, and would bring us back to the C-style 
 redundant symbol names all over. It's unpleasant to have to 
 disambiguate even when not necessary... I suppose I could also 
 do  mylibattr("ignore") instead, but this is also hideous and 
 overly verbose...

 I tried  mylib.ignore, which would not be too bad, if necessary 
 only to disambiguate, but it seems like the parser doesn't 
 understand it.


` (mylib.ignore)` should work. You could open an enhancement 
request to enable the paren-less syntax.
Feb 06 2016
parent =?UTF-8?B?TcOhcmNpbw==?= Martins <marcioapm gmail.com> writes:
On Saturday, 6 February 2016 at 15:01:44 UTC, Marc Schütz wrote:

 On Saturday, 6 February 2016 at 13:36:32 UTC, Márcio Martins 
 wrote:

 [...]


 ` (mylib.ignore)` should work. You could open an enhancement 
 request to enable the paren-less syntax.


Thanks, that does work indeed and is not that verbose...

Cheers!
Feb 06 2016