AA key conversion woes

[Timon Gehr]

On 04/17/2012 09:04 PM, Dmitry Olshansky wrote:
> On 17.04.2012 22:59, Timon Gehr wrote:
>> On 04/17/2012 08:53 PM, H. S. Teoh wrote:
>>> On Tue, Apr 17, 2012 at 08:48:04PM +0200, Andrej Mitrovic wrote:
>>>> On 4/17/12, Andrej Mitrovic<andrej.mitrovich at gmail.com> wrote:
>>>>> On 4/17/12, H. S. Teoh<hsteoh at quickfur.ath.cx> wrote:
>>>>>> But even then, I'm considering if .keys should return a mutable
>>>>>> array if
>>>>>> the key is a value type
>>>>>
>>>>> How about having .keys for mutable
>>>>
>>>> Personally I use .keys often and expect them to be mutable. For
>>>> example:
>>>>
>>>> string[] atts = attributes.keys;
>>>> sort(atts);
>>>>
>>>> There is no "sorted" in phobos so I have to use the sort function
>>>> which takes arguments by reference. I can't pass immutable arrays to
>>>> it.
>>>>
>>>> Sometimes I like to store an array of keys from multiple hashes, e.g.:
>>>> string[] arr = hash.keys;
>>>> arr ~= hash2.keys;
>>>>
>>>> Long story short, making .keys immutable would break a lot of my code.
>>>> I don't know if that's true for others as well..
>>>
>>> What if .keys returns a tail-immutable array? You should still be able
>>> to sort it, right?
>>>
>>>
>>> T
>>>
>>
>> No, an immutable array cannot be sorted regardless of whether or not the
>> slice is mutable.
>
> Mm I believe you can sort array with immutable *elements* like say
> strings (at least dchar ones).
>
>

immutable(string)[] <- tail-immutable array, cannot be sorted in place
string[] <- can be sorted, but it is not tail-immutable