• Andrei Alexandrescu (11/11) Apr 14 2016 https://dlang.org/library/std/exception/may_point_to.html
• Vladimir Panteleev (8/17) Apr 14 2016 Generally, if you don't understand why something is there, you
• Andrei Alexandrescu (14/30) Apr 14 2016 I know, but my point is that shouldn't be the way. Rationale for why
• Vladimir Panteleev (10/16) Apr 14 2016 I don't disagree with that. But another argument is that finding

Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:

https://dlang.org/library/std/exception/may_point_to.html
https://dlang.org/library/std/exception/does_point_to.html

We have an unfortunate pattern in Phobos: things that start out as 
simple ideas with nice implementations (in this case pointsTo) slowly 
change into complicated artifacts with undocumented rationale. We really 
should rein in documentation, not only external but also internal. In 
this case I cannot for the life of me figure what Tdummy does, and the 
only appearance of the name is in the respective definitions.

Not to mention it uses incorrect spacing ("Tdummy=void", should be 
"Tdummy = void") :o).


Andrei

Vladimir Panteleev <thecybershadow.lists gmail.com> writes:

On Thursday, 14 April 2016 at 18:47:41 UTC, Andrei Alexandrescu 
wrote:
 https://dlang.org/library/std/exception/may_point_to.html
 https://dlang.org/library/std/exception/does_point_to.html

 We have an unfortunate pattern in Phobos: things that start out 
 as simple ideas with nice implementations (in this case 
 pointsTo) slowly change into complicated artifacts with 
 undocumented rationale. We really should rein in documentation, 
 not only external but also internal. In this case I cannot for 
 the life of me figure what Tdummy does, and the only appearance 
 of the name is in the respective definitions.

Generally, if you don't understand why something is there, you 
could...

1. git blame

2. Delete it and see which tests fail :)

In this case it appears to be used to disambiguate between the 
overloaded template.

Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:

On 04/14/2016 02:50 PM, Vladimir Panteleev wrote:
 On Thursday, 14 April 2016 at 18:47:41 UTC, Andrei Alexandrescu wrote:
 https://dlang.org/library/std/exception/may_point_to.html
 https://dlang.org/library/std/exception/does_point_to.html

 We have an unfortunate pattern in Phobos: things that start out as
 simple ideas with nice implementations (in this case pointsTo) slowly
 change into complicated artifacts with undocumented rationale. We
 really should rein in documentation, not only external but also
 internal. In this case I cannot for the life of me figure what Tdummy
 does, and the only appearance of the name is in the respective
 definitions.

 Generally, if you don't understand why something is there, you could...

 1. git blame

I know, but my point is that shouldn't be the way. Rationale for why 
something is there should be in there, not in the accumulated history of 
commit explanations.

 2. Delete it and see which tests fail :)

 In this case it appears to be used to disambiguate between the
 overloaded template.

Thanks. That's the shared overload. Isn't there a much better solution? 
Why isn't shared handled inside the body along with the other cases?

Also, why is this:

foreach (i, Subobj; typeof(source.tupleof))
   if (mayPointTo(source.tupleof[i], target)) return true;

not written easier like this:

foreach (ref member; source.tupleof)
   if (mayPointTo(member, target)) return true;

?


Andrei

Vladimir Panteleev <thecybershadow.lists gmail.com> writes:

On Thursday, 14 April 2016 at 18:56:19 UTC, Andrei Alexandrescu 
wrote:
 I know, but my point is that shouldn't be the way. Rationale 
 for why something is there should be in there, not in the 
 accumulated history of commit explanations.

I don't disagree with that. But another argument is that finding 
out the reason (by deleting it and looking at the error messages) 
took me less time than it took you to write this thread.

 Thanks. That's the shared overload. Isn't there a much better 
 solution? Why isn't shared handled inside the body along with 
 the other cases?

I have not even been aware of this function's existence before 
this thread, so I'll let someone else answer that. Perhaps 
finding the author is the use case for git blame. (This is one 
reason why I think people should be reachable by PR comments even 
after PRs are merged, BTW.)