This program

```
void aa_stat(T, U) (T[U] aa)
{
   import std.stdio;
   writefln ("aa        : %s", aa);
   writefln ("aa.length : %d", aa.length);
   writefln ("aa.keys   : %s", aa.keys);
   writefln ("aa.values : %s", aa.values);
   foreach (k, v; aa)
      writeln (k, ": ", v);
}

void main ()
{
   string[int] aa = [
      0: "null",
      0: "eins"
   ];
   aa.aa_stat;
}
```

produces this output:

   aa        : [0:"eins", ]
   aa.length : 2
   aa.keys   : [0, 32767]
   aa.values : ["eins", ""]
   0: eins

I would have expected either a compile time or a runtime error or this result:

   aa        : [0:"eins"]
   aa.length : 1
   aa.keys   : [0]
   aa.values : ["eins"]
   0: eins

What's happening here?

On Thu, Nov 30, 2017 at 11:50:13PM +0000, kdevel via Digitalmars-d-learn wrote:
> This program
> 
> ``` void aa_stat(T, U) (T[U] aa) { import std.stdio; writefln ("aa
> : %s", aa); writefln ("aa.length : %d", aa.length); writefln ("aa.keys
> : %s", aa.keys); writefln ("aa.values : %s", aa.values); foreach (k,
> v; aa) writeln (k, ": ", v); }
> 
> void main () { string[int] aa = [ 0: "null", 0: "eins" ]; aa.aa_stat;
> } ```
> 
> produces this output:
> 
>    aa        : [0:"eins", ] aa.length : 2 aa.keys   : [0, 32767]
>    aa.values : ["eins", ""] 0: eins
[...]

This looks like a bug in the implementation of AA literals.  Please file a bug at http://issues.dlang.org/.


--T

Hmm, which compiler are you using, and which version?  I cannot reproduce this bug with DMD git head.  It may be have fixed since the release you're using?


--T

On Thu, Nov 30, 2017 at 03:57:37PM -0800, H. S. Teoh via Digitalmars-d-learn wrote:
> Hmm, which compiler are you using, and which version?  I cannot reproduce this bug with DMD git head.  It may be have fixed since the release you're using?
[...]

Sorry, I was testing the wrong code.  This problem does still happen on DMD git head.  Looking into the code right now...  Please still file the bug if you haven't already.


T

-- 
"How are you doing?" "Doing what?"

Seems like this problem is already known:

	https://issues.dlang.org/show_bug.cgi?id=15290

Here's the fix:

	https://github.com/dlang/druntime/pull/1980


--T

On Friday, 1 December 2017 at 00:42:00 UTC, H. S. Teoh wrote:
> Here's the fix:
>
> 	https://github.com/dlang/druntime/pull/1980

Great. By the way: It it true, that there cannot be more than 2^32 keys in an associative array (due to the use of uint)?

On Friday, 1 December 2017 at 00:42:00 UTC, H. S. Teoh wrote:
> Here's the fix:
>
> 	https://github.com/dlang/druntime/pull/1980

And wouldn't it be reasonable to add

   assert(aa.values == [ "b" ]);

to the unittest?

On 12/1/17 11:01 AM, kdevel wrote:
> On Friday, 1 December 2017 at 00:42:00 UTC, H. S. Teoh wrote:
>> Here's the fix:
>>
>>     https://github.com/dlang/druntime/pull/1980
> 
> Great. By the way: It it true, that there cannot be more than 2^32 keys in an associative array (due to the use of uint)?

Yes, and technically even less because there is a requirement to grow the array when it's 4/5 full. But such a thing is easily fixed if we need to.

Note that due to the way it's currently implemented (and you aren't going to get much smaller than this), each key takes 16 bytes (bucket ptr + hash) + at least 16 bytes per key/value pair (depending on what's stored there).

So you are talking minimum 4GB * 32 = 128GB RAM required if you wanted to have such an AA!

-Steve

On Fri, Dec 01, 2017 at 04:06:50PM +0000, kdevel via Digitalmars-d-learn wrote:
> On Friday, 1 December 2017 at 00:42:00 UTC, H. S. Teoh wrote:
> > Here's the fix:
> > 
> > 	https://github.com/dlang/druntime/pull/1980
> 
> And wouldn't it be reasonable to add
> 
>    assert(aa.values == [ "b" ]);
> 
> to the unittest?

I did think about adding that, but then I didn't want the unittest to dictate which value should take precedence when there are duplicate keys, since that is currently implementation-dependent.  At the very least, it merits further discussion.


T

-- 
Never trust an operating system you don't have source for! -- Martin Schulze

On Friday, 1 December 2017 at 16:23:33 UTC, H. S. Teoh wrote:
> On Fri, Dec 01, 2017 at 04:06:50PM +0000, kdevel via Digitalmars-d-learn wrote:
>> On Friday, 1 December 2017 at 00:42:00 UTC, H. S. Teoh wrote:
>> > Here's the fix:
>> > 
>> > 	https://github.com/dlang/druntime/pull/1980
>> 
>> And wouldn't it be reasonable to add
>> 
>>    assert(aa.values == [ "b" ]);
>> 
>> to the unittest?
>
> I did think about adding that, but then I didn't want the unittest to dictate which value should take precedence when there are duplicate keys, since that is currently implementation-dependent.  At the very least, it merits further discussion.

I personally would have preferred a run time or even a compile-time error in the case of my orginal example. That would support the "fail fast" philosohphy. IMHO it makes no sense at all to initialize an AA in such a way. The way the initialization works hides copy-and-paste errors like the one of my example code.