www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Issue with casting types

reply Thorsten Sommer <vektoren gmail.com> writes:
Dear all,

I run into an issue with a simple cast:
https://dpaste.dzfl.pl/8e7f7c545eb1

I have a base class and a class A with inherit from that base 
class:
A <- BaseClass

If the base class contains an "alias this", any casting attempt 
fails because the "alias this" type gets considered. Thus, I 
defined the opCast() to solve this. But now, the program crashes 
without any error or exception...

Why does the casting operation consider the "alias this" at all? 
I mean: The "alias this" type is int. If I try to cast with to!A, 
obviously int does not match A!?

Is there any practical solution? Or should I stop using "alias 
this" for my classes? I like the "alias this" concept. But it 
caused some issue for me...


Best regards,
Thorsten
May 17 2016
next sibling parent ag0aep6g <anonymous example.com> writes:
On 05/17/2016 02:24 PM, Thorsten Sommer wrote:

 Dear all,

 I run into an issue with a simple cast:
 https://dpaste.dzfl.pl/8e7f7c545eb1

 I have a base class and a class A with inherit from that base class:
 A <- BaseClass

 If the base class contains an "alias this", any casting attempt fails
 because the "alias this" type gets considered. Thus, I defined the
 opCast() to solve this. But now, the program crashes without any error
 or exception...


You've got an infinite recursion there.

`A a2 = to!A(b1);` calls `b1.opCast!A()` which calls `to!A(this)` which 
is the very same call as `to!A(b1)`, so it calls `opCast!A()` again, and 
so on until the stack is exhausted and the program crashes.


 Why does the casting operation consider the "alias this" at all? I mean:
 The "alias this" type is int. If I try to cast with to!A, obviously int
 does not match A!?


There's an issue on this: https://issues.dlang.org/show_bug.cgi?id=6777

I don't know if there's a good reason for the current behavior. Looks 
silly to me.


 Is there any practical solution? Or should I stop using "alias this" for
 my classes? I like the "alias this" concept. But it caused some issue
 for me...


You can use the reinterpreting style of cast to circumvent features like 
alias this and opCast:
----
A a2 = * cast(A*) &b1;
----
May 17 2016
prev sibling parent reply Steven Schveighoffer <schveiguy yahoo.com> writes:
On 5/17/16 8:24 AM, Thorsten Sommer wrote:

 Dear all,

 I run into an issue with a simple cast:
 https://dpaste.dzfl.pl/8e7f7c545eb1

 I have a base class and a class A with inherit from that base class:
 A <- BaseClass

 If the base class contains an "alias this", any casting attempt fails
 because the "alias this" type gets considered. Thus, I defined the
 opCast() to solve this. But now, the program crashes without any error
 or exception...


It's a stack overflow.

obj1.to!OtherObject is going to basically lower to 
cast(OtherObject)obj1. Which then invokes your opCast. Which then 
invokes to, ...


 Why does the casting operation consider the "alias this" at all? I mean:
 The "alias this" type is int. If I try to cast with to!A, obviously int
 does not match A!?


It shouldn't. This is a bug in the language. alias this should only be 
considered if the expression is invalid for the actual type.


 Is there any practical solution? Or should I stop using "alias this" for
 my classes? I like the "alias this" concept. But it caused some issue
 for me...


I think you need to avoid it for now. Please file an issue.

-Steve
May 17 2016
parent reply Steven Schveighoffer <schveiguy yahoo.com> writes:
On 5/17/16 8:59 AM, Steven Schveighoffer wrote:

 I think you need to avoid it for now. Please file an issue.


I see from ag0aep6g, that there is already an issue. I updated it.

-Steve
May 17 2016
parent reply Thorsten Sommer <vektoren gmail.com> writes:
On Tuesday, 17 May 2016 at 13:13:54 UTC, Steven Schveighoffer 
wrote:

 On 5/17/16 8:59 AM, Steven Schveighoffer wrote:

 I think you need to avoid it for now. Please file an issue.


 I see from ag0aep6g, that there is already an issue. I updated 
 it.

 -Steve


Thanks ag0aep6g and Steve for the fast help :) The D community is 
great :)
May 17 2016
parent Steven Schveighoffer <schveiguy yahoo.com> writes:
On 5/17/16 10:19 AM, Thorsten Sommer wrote:

 On Tuesday, 17 May 2016 at 13:13:54 UTC, Steven Schveighoffer wrote:

 On 5/17/16 8:59 AM, Steven Schveighoffer wrote:

 I think you need to avoid it for now. Please file an issue.


 I see from ag0aep6g, that there is already an issue. I updated it.


 Thanks ag0aep6g and Steve for the fast help :) The D community is great :)


Someone identified a workaround in the bug report:

Object obj = b1;
auto a2 = cast(A)obj;

Very ugly, but it should work.

-Steve
May 17 2016