Hi!
I love the language. I would like to see the std. libraries 100% object
oriented, but that's mild considering it fixes a heck of a lot of problems with
C and C++.

I'm trying to do some real coding with D and I've found some annoying bugs.

The first thing I noticed is problems with the toString() method. In std.string there are a great number of methods to convert base types to char[]. Here is an example of the problem I had:

module math.Vector;
import std.string;

class Vector {
private:
double x, y;
public:
char[] toString() {
return "("~toString(x)~","~toString(y)~")";
}
}

Now, immediately you'll notice that this snippet does not compile, even though, intuitively it _should_ compile. The problem appears to be that the compiler does not look outside of Vector for toString() definitions. Java, of course, handles this by not allowing global methods whatsoever. But since D has global methods, D must search _both_ the local scope, and recursively check all parent scopes, so in this case it must check the scope of toString(), then Vector, then math.Vector, then the default module scope, then all imported scopes including std.string. I worked around this bug by using the fully qualified name std.string.toString(double).

The second bug is demonstrated in the following code:

module math.Vector;
import std.math;

Boom... not even these two lines compile. I don't even have to do anything more, it says "module math is multiply defined". I need to use sqrt() in order to calculate the distance between two Vectors, but it looks like I can't call the package I'm using math, otherwise I can't import std.math. Kinda sucky, considering the two names, intuitively, shouldn't conflict at all.

-Ryan

Knaar wrote:

> Hi!
> I love the language. I would like to see the std. libraries 100% object
> oriented, but that's mild considering it fixes a heck of a lot of problems with
> C and C++.
>
> I'm trying to do some real coding with D and I've found some annoying bugs.
>
> The first thing I noticed is problems with the toString() method. In std.string
> there are a great number of methods to convert base types to char[]. Here is an
> example of the problem I had:
>
> module math.Vector;
> import std.string;
>
> class Vector {
> private:
> double x, y;
> public:
> char[] toString() {
> return "("~toString(x)~","~toString(y)~")";
> }
> }
>  
>
Put a dot in front of toString to get access to global scope.

class Vector {
private:
double x, y;
public:
char[] toString() {
return "("~.toString(x)~","~.toString(y)~")";
}
}

> <snip>
>  
>

> The second bug is demonstrated in the following code:
>
> module math.Vector;
> import std.math;
>
> Boom... not even these two lines compile. I don't even have to do anything more,
> it says "module math is multiply defined". I need to use sqrt() in order to
> calculate the distance between two Vectors, but it looks like I can't call the
> package I'm using math, otherwise I can't import std.math. Kinda sucky,
> considering the two names, intuitively, shouldn't conflict at all.
>
> -Ryan
>  
>
Actually its the first line that conflicts the second line has nothing to do with it.   The reason is because there is a

module math;

in phobos, so that namespace is already used.

D doesn't allow you to have a module name the same name as a folder.  You could always create your own namespace ie:

module Knaar.math.Vector;

PS - there is a D bugs newsgroup.

-- 
-Anderson: http://badmama.com.au/~anderson/

Knaar wrote:
> Hi!
> I love the language. I would like to see the std. libraries 100% object
> oriented, but that's mild considering it fixes a heck of a lot of problems with
> C and C++.
> 
> I'm trying to do some real coding with D and I've found some annoying bugs.
> 
> The first thing I noticed is problems with the toString() method. In std.string
> there are a great number of methods to convert base types to char[]. Here is an
> example of the problem I had:
> 
> module math.Vector;
> import std.string;
> 
> class Vector {
> private:
> double x, y;
> public:
> char[] toString() {
> return "("~toString(x)~","~toString(y)~")";
> }
> }
> 
> Now, immediately you'll notice that this snippet does not compile, even though,
> intuitively it _should_ compile. The problem appears to be that the compiler
> does not look outside of Vector for toString() definitions.

It works like a charm if you do this instead:

	return "("~.toString(x)~","~.toString(y)~")";

note the "."

 Java, of course,
> handles this by not allowing global methods whatsoever. But since D has global
> methods, D must search _both_ the local scope, and recursively check all parent
> scopes, so in this case it must check the scope of toString(), then Vector, then
> math.Vector, then the default module scope, then all imported scopes including
> std.string. I worked around this bug by using the fully qualified name
> std.string.toString(double).
> 
> The second bug is demonstrated in the following code:
> 
> module math.Vector;
> import std.math;
> 
> Boom... not even these two lines compile. I don't even have to do anything more,
> it says "module math is multiply defined". I need to use sqrt() in order to
> calculate the distance between two Vectors, but it looks like I can't call the
> package I'm using math, otherwise I can't import std.math. Kinda sucky,
> considering the two names, intuitively, shouldn't conflict at all.
> 
> -Ryan
> 
> 

Thanks for the reply. I guess that makes sense. I'll just stuff all my code in another parent directory. Sorry about the bug report here in this newsgroup. I'll try to find the bugs forum.

-Knaar

Knaar wrote:

>Thanks for the reply. I guess that makes sense. I'll just stuff all my code in
>another parent directory. Sorry about the bug report here in this newsgroup.
>I'll try to find the bugs forum.
>
>-Knaar
>  
>

Are you using a newsgroup program (which I recommend) or the webpage?

Anyway heres the web-based D forums:
http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.*

-- 
-Anderson: http://badmama.com.au/~anderson/