I tried this:

    immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];

And got a "non-constant expression" error (with or without 'immutable').

What's the correct way?

On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
> I tried this:
>
>     immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
>
> And got a "non-constant expression" error (with or without 'immutable').
>
> What's the correct way?

This example here: https://dlang.org/spec/hash-map.html, doesn't work either.

On Saturday, 24 December 2016 at 00:57:04 UTC, Yuxuan Shui wrote:
> On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
>> I tried this:
>>
>>     immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
>>
>> And got a "non-constant expression" error (with or without 'immutable').
>>
>> What's the correct way?
>
> This example here: https://dlang.org/spec/hash-map.html, doesn't work either.

Is this at global scope?

You need to use a `shared static this() { ... }` to initialise it.

On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
> I tried this:
>
>     immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
>
> And got a "non-constant expression" error (with or without 'immutable').
>
> What's the correct way?

You cannot initialize an AA at compile-time.
Because AA's are provided by druntime and the ABI is not stable.

On Saturday, 24 December 2016 at 11:21:11 UTC, Stefan Koch wrote:

> You cannot initialize an AA at compile-time.
> Because AA's are provided by druntime and the ABI is not stable.

This bit me when I was first starting out as well. I feel like there's really very little documentation on this, and I think most people's expect that you'd be able to initialize AAs from literals at compile time, just like normal arrays. Does anyone know if compile-time initialization of AAs is something that will be fixed?

On Saturday, 24 December 2016 at 11:21:11 UTC, Stefan Koch wrote:
> On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
>> What's the correct way?
>
> You cannot initialize an AA at compile-time.
> Because AA's are provided by druntime and the ABI is not stable.

 For fun I'm throwing together a quick AA struct that you can make at compile time. It's experimental and may be slow for compiling as it needs to find a size of numbers where the hashes don't overlap. Of course you can't add to it, you can't delete elements (key limitations) but you could in theory change the value; So this would be better for immutable objects or having a fixed number of pre-known keys.

 If anyone is interested I can share my sources fairly soon once I'm satisfied the required features are done. So far opIndex and generation works fine, but range/foreach over it is incomplete.

On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
> I tried this:
>
>     immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
>
> And got a "non-constant expression" error (with or without 'immutable').
>
> What's the correct way?

This works:

void main() {
    immutable int[char] xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
}

If you want to initialize global variable, then

immutable int[char] xx;
shared static this() {
	xx = ['Q':0, 'B':1, 'N':2, 'R':3, 'P':4];
}

On Sunday, 25 December 2016 at 05:45:14 UTC, Era Scarecrow wrote:
> On Saturday, 24 December 2016 at 11:21:11 UTC, Stefan Koch wrote:
>> On Saturday, 24 December 2016 at 00:55:01 UTC, Yuxuan Shui wrote:
>>> What's the correct way?
>>
>> You cannot initialize an AA at compile-time.
>> Because AA's are provided by druntime and the ABI is not stable.
>
>  For fun I'm throwing together a quick AA struct that you can make at compile time. It's experimental and may be slow for compiling as it needs to find a size of numbers where the hashes don't overlap. Of course you can't add to it, you can't delete elements (key limitations) but you could in theory change the value; So this would be better for immutable objects or having a fixed number of pre-known keys.
>
>  If anyone is interested I can share my sources fairly soon once I'm satisfied the required features are done. So far opIndex and generation works fine, but range/foreach over it is incomplete.

There are library solutions for AA like containers that can be used at compiletime.
Maybe you could help polishing them ?
So we can eventually get them into druntime ?

On Sunday, 25 December 2016 at 03:12:09 UTC, Carl Vogel wrote:
> Does anyone know if compile-time initialization of AAs is
> something that will be fixed?

I do intend to fix this.
The method is rather simple,
Given the aa implantation in d-runtime is made ctfe-able everything that causes trouble currently will just work.