On 13/11/2010 20:02, spir wrote:
> On Sat, 13 Nov 2010 17:27:08 +0000
> div0<div0@sourceforge.net>  wrote:
>
>>> How about "associative array literal" then? Regardless of what you call
>>> it I shouldn't get an assertion failure.
>>>
>>
>> True. It's been fixed in dmd2 though, you get:
>>
>> Error: Integer constant expression expected instead of ""
>>
>> When it's auto the compiler thinks you are doing static initialisation
>> of a flat array, not an AA. You have to do what Ellery said.
>
> But the compiler (D2) accepts nested aa literals remaining anonymous:
>      writeln(["a" : ["b" : "c"]]);	// -->  "a:b:c"
> (where "auto aa = ..." fails)
>
> Denis
> -- -- -- -- -- -- --
> vit esse estrany ☣
>
> spir.wikidot.com
>

yeah, but dmd's parser was written by hand so it's not surprising there are inconsistency's with what works where.

I finally found the bit where it describes associative array literals and they look identical to initialising a flat array, so god only knows which one gets picked when. It would be better if they where made different.

-- 
My enormous talent is exceeded only by my outrageous laziness.
http://www.ssTk.co.uk

div0:

> I finally found the bit where it describes associative array literals and they look identical to initialising a flat array, so god only knows which one gets picked when. It would be better if they where made different.

One of my many enhancement requests, no one disturbs their sleep in Bugzilla :-)

http://d.puremagic.com/issues/show_bug.cgi?id=4703

Bye,
bearophile

On 11/13/2010 02:02 PM, spir wrote:
> On Sat, 13 Nov 2010 17:27:08 +0000
>
> But the compiler (D2) accepts nested aa literals remaining anonymous:
>      writeln(["a" : ["b" : "c"]]);	// -->  "a:b:c"
> (where "auto aa = ..." fails)
>

The difference is initializer vs expression.
initializers occur on the rhs of a decl/assignment, and they are not parsed the same as expressions (e.g. you can have 'void' and struct literals as initializers -

int i = void;
X x = {a: {b: 1}, c: "hi"};

- you can't have these in expressions).

Initializers are also stored and treated differently in the compiler, which results in this and other inconsistencies

On 14/11/2010 00:28, Ellery Newcomer wrote:
>
>
> On 11/13/2010 02:02 PM, spir wrote:
>> On Sat, 13 Nov 2010 17:27:08 +0000
>>
>> But the compiler (D2) accepts nested aa literals remaining anonymous:
>> writeln(["a" : ["b" : "c"]]); // --> "a:b:c"
>> (where "auto aa = ..." fails)
>>
>
> The difference is initializer vs expression.
> initializers occur on the rhs of a decl/assignment, and they are not
> parsed the same as expressions (e.g. you can have 'void' and struct
> literals as initializers -
>
> int i = void;
> X x = {a: {b: 1}, c: "hi"};
>
> - you can't have these in expressions).

OIC, ty.

> Initializers are also stored and treated differently in the compiler,
> which results in this and other inconsistencies

Yeah I know. :( I've been working on implementing some missing CTFE functionality so that one day I might actually get to upgrade from 2.028, and having dealing with initializers vs expressions is a real pita.

-- 
My enormous talent is exceeded only by my outrageous laziness.
http://www.ssTk.co.uk

On Sun, 14 Nov 2010 00:03:48 +0000
div0 <div0@sourceforge.net> wrote:


> yeah, but dmd's parser was written by hand so it's not surprising there are inconsistency's with what works where.

Well, most (all?) truely used languages have hand-written parsers, AFAIK. The issue is not with writing the parser, I guess, but with having a clear, unambiguous, complete, definition of the language ;-)

> I finally found the bit where it describes associative array literals and they look identical to initialising a flat array, so god only knows which one gets picked when. It would be better if they where made different.

... with the tiny difference AAs are supposed to have (explicite) keys ;-)


Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

On Sun, 14 Nov 2010 10:35:42 +0100
spir <denis.spir@gmail.com> wrote:

> > I finally found the bit where it describes associative array literals and they look identical to initialising a flat array, so god only knows which one gets picked when. It would be better if they where made different.
> 
> ... with the tiny difference AAs are supposed to have (explicite) keys ;-)

Oops, sorry for the noise...

Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

On 2010-11-13 18:27, div0 wrote:
> On 13/11/2010 15:49, Jacob Carlborg wrote:
>> On 2010-11-13 14:56, div0 wrote:
>>> On 13/11/2010 11:02, Jacob Carlborg wrote:
>>>> On 2010-11-12 17:44, Ellery Newcomer wrote:
>>>>> Should be. Are you having problems?
>>>>>
>>>>> (I don't use them much, but fwiw, it seems like tango had some
>>>>> [trivial?] problems with them)
>>>>>
>>>>> On 11/12/2010 10:08 AM, Jacob Carlborg wrote:
>>>>>> Is D supposed to be able to handle nested associative arrays ?
>>>>>>
>>>>>>
>>>>
>>>> Well, yes. The following code:
>>>>
>>>> module main;
>>>>
>>>> void main ()
>>>> {
>>>> auto tree = ["" : ["" : ""]];
>>>> }
>>>>
>>>> Using DMD 1.065 results in:
>>>>
>>>> Assertion failed: (0), function toExpression, file init.c, line 437.
>>>>
>>> That's static initialisation not an AA.
>>>
>>> alias string[char] innerAA;
>>> alias innerAA[int] outerAA;
>>>
>>> outerAA x;
>>> innerAA t;
>>>
>>> x[2] = t;
>>>
>>> auto y = x[2];
>>> y['c'] = "cat";
>>>
>>> I've no idea what'll happen if you start passing it around though.
>>
>> How about "associative array literal" then? Regardless of what you call
>> it I shouldn't get an assertion failure.
>>
>
> True. It's been fixed in dmd2 though, you get:
>
> Error: Integer constant expression expected instead of ""
>
> When it's auto the compiler thinks you are doing static initialisation
> of a flat array, not an AA. You have to do what Ellery said.

Well, that's no fun. How many are using the Static Initialization of Static Arrays syntax: http://www.digitalmars.com/d/2.0/arrays.html#static-init-static ?

-- 
/Jacob Carlborg