A new class -->String

[Jari-Matti Mäkelä]

Daniel Keep wrote:

> 
> 
> Jari-Matti Mäkelä wrote:
>> David B. Held wrote:
>>> Jari-Matti Mäkelä wrote:
>>>> David B. Held wrote:
>>>>> If there were a way to switch between method call and free
>>>>> function syntax, then this distinction would become moot (and maybe we
>>>>> will get that as well...).
>> 
>>>> D has that already :)
>> 
>>> Well, it isn't general:
>>>
>>> void foo(int x) { int y = x; }
>>>
>>> unittest
>>> {
>>>     int z = 5;
>>>     z.foo();
>>> }
>>>
>>> test.d(43): Error: no property 'foo' for type 'int'
>>> test.d(43): Error: function expected before (), not 1 of type int
>> 
>> Hm, I wonder if that's a compiler bug. I have not used this feature
>> recently, but I recall it used to work. I even have bindings where this
>> feature is used to provide an alternative syntax for a wrapper class.
>> Now that I tested it, it only seems to work with arrays and AAs anymore.
>> But it's still enough to implement those string functions though.
> 
> From what I understand, it's not a feature, it's a bug.  No, I'm
> serious: this happens because of a bug in DMD.  The only reason it's
> never been fixed is because people actually really, really like that it
> works :)

Ah, you're right. I didn't find it in the language docs either.

> The whole "switch function to member of first arg" thing is only usable,
> as you said, on arrays and AAs.  Having it usable in general would be
> wicked-nifty-awesome-keen, to mix my colloquialisms. :)

In Scala language this can be "safely" done with views and implicit type
conversions. If this "poor man's version" isn't a good solution in the long
run, at least there is another way to do this that has strong academic
backing behind it. That alternative is also much more flexible.