Consider

<code>
void main(){
 int [ char[]] aa;
 printf("%d\n", aa["x"]);
 printf("%d\n", aa.length);

 char[]* p=cast(char[]*)("x" in aa);
 p != null && printf("%.*s\n", *p);
}
</code>

<output>
0
1
(null)
</output>

Questions:

1) According to the second line of the output the length of the AA is increased by the access to the value of "x". Where is documented that the access to the _value_ of a key inserts the key into the AA and why is this a useful side effect?

2)
<specs>
The InExpression yields a pointer to the value if the key is in the
associative array, or null if not
</specs>

According to the first line of the output the value of the key `"x"' is "0" and according to the retrieval preceding the `printf' the value of `p' is not `null'. According to the specs the third line should then hold a "0" instead of a "(null)". So what is wrong here: me, the first or the third line of this output?

Hint: multiple choice :-)

-manfred

Manfred Nowak wrote:
> Consider
> 
> <code>
> void main(){
>  int [ char[]] aa;
>  printf("%d\n", aa["x"]);
>  printf("%d\n", aa.length);
> 
>  char[]* p=cast(char[]*)("x" in aa);
>  p != null && printf("%.*s\n", *p);
> }
> </code>
> 
> <output>
> 0
> 1
> (null)
> </output>
> 
> Questions:
> 
> 1) According to the second line of the output the length of the AA is increased by the access to the value of "x". Where is documented that the access to the _value_ of a key inserts the key into the AA and why is this a useful side effect?
> 

I don't think it's documented, but it's known behavior. Some don't like it, some don't mind. I'm in the second group.

> 2)
> <specs>
> The InExpression yields a pointer to the value if the key is in the associative array, or null if not
> </specs>
> 
> According to the first line of the output the value of the key `"x"' is  "0" and according to the retrieval preceding the `printf' the value of `p' is not `null'. According to the specs the third line should then hold a "0" instead of a "(null)". So what is wrong here: me, the first or the third line of this output?
>
> Hint: multiple choice :-)

I modified your code a bit to answer this:

//------------------------------------
void main(){
 int [ char[]] aa;
 printf("%d\n", aa["x"]);
 printf("%d\n", aa.length);

    int * xxx="x" in aa;
    printf("%x, %d\n",xxx,*xxx);
 char[]* p=cast(char[]*)xxx;

//    char[]* p=cast(char[]*)("x" in aa);
    p != null && printf("{%.*s}\n", *p);
}
//------------------------------------

The output I get is (dmd 0.111 for Win):

//------------------------------------
0
1
900fd4, 0
{}
//------------------------------------

So, I'd say you're wrong. "p" is not null: it's pointing to where "xxx" is pointing, which is the address of aa["x"] (900fd4 in my case). Now, printf("%.*s") is expecting a length and a pointer, but since "p" is pointing to a 0, that's the length printf is reading, so it doesn't print anything.

I hope I got all of this right.

> 
> -manfred

_______________________
Carlos Santander Bernal

"Carlos Santander B." wrote:

[...]
> I don't think it's documented, but it's known behavior. Some don't like it, some don't mind. I'm in the second group.

Thanks for the hint.
I found http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D/13789 on
this issue. I have been away for a couple of month :-(

[...]
>      int * xxx="x" in aa;
>      printf("%x, %d\n",xxx,*xxx);

Ahh, now I see my error. Because of the former post, where I mentioned that  one is forced to see the key in the expression `b["hello"]' and not the value, I blindly overread, that a pointer to the value is returned and not to the key. This blindness reached that far, that I casted the unexpected `int*' to a `char[]*', even not noticing this, when writing the post.

[...]
> so it doesn't print anything.

If it hadn't print anything, i.e. "", my error would have gone unnoticed and unpublished, but it printed "(none)" and I still do not understand why.

-manfred

"Manfred Nowak" <svv1999@hotmail.com> wrote in message news:ct6q18$1s4$1@digitaldaemon.com...
> If it hadn't print anything, i.e. "", my error would have gone unnoticed and unpublished, but it printed "(none)" and I still do not understand why.

printf will print (null) on seeing a null pointer, despite the length for .*s being 0. writef will print no characters.

Manfred Nowak wrote:
> "Carlos Santander B." wrote: 
>>so it doesn't print anything.
> 
> 
> If it hadn't print anything, i.e. "", my error would have gone unnoticed and unpublished, but it printed "(none)" and I still do not understand why.
> 
> -manfred

That's just what I got. I don't know either why it gave you something else.

_______________________
Carlos Santander Bernal