Re: ANN: Seed7 Release 2008-02-17

thomas.mertes@[EMAIL PROTECTED]
 wrote:
> On 18 Feb., 22:02, Bart <b...@[EMAIL PROTECTED]
> wrote:
>> On Feb 17, 10:13 pm, thomas.mer...@[EMAIL PROTECTED]
 wrote:
>>
>> [From Seed 7 FAQ]
>>
>>> Variables with object types contain references to object values....
>>> a := b
>>> For primitive types a different logic is used...
>>> a := b
>>> both variables are still distinct and changing one variable has no
>>> effect on the other.
>>
>> This is more of a general question about a:=b for a class object
>> being different from a:=b for an ordinary object. It seems to be the
>> case in several oo languages.
>>
>> Why?
> IMHO it is a cornerstone of classic OO programming languages
> like Smalltalk and its descendants: Object variables and
> parameters contain just pointers to the actual object value
> which is at the heap. It is almost not mentioned by OO
> proponents but in classic OO programming not everything is
> done with methods. Assignments are always done as pointer
> assignments, independend of the object class. Also comparisons
> are done as pointer comparisons. Therefore a 'clone' method
> to do deep copies and an 'compare' method to do logical
> comparisons is necessary.
>
> The advantage of this OO pointer philosophy is that container
> cl***** are easier to implement (they store just pointers in
> the container).
>
>> OK, objects are implemented by reference and all that, and maybe this
>> saves some copying, but that should all be transparent.
>>
>> Surely the meaning of something as basic as a:=b should not depend on
>> something as arbitrary as whether a and b have datatypes T or U. If
>> someone changes their mind then a lot of recoding might be needed!
>
> Most OO programmers have the different assignment logic for
> object variables in their mind. They would be confused when
> those assignments would work as deep copies (As Seed7 does
> for all other types).
>
> IIRC Simula had two different assignment operators:
>  := For value copy assignments
>  :- For reference assignments
>
> Seed7 is an extensible programming language.
> It would be possible to define the assignment to do a (deep)
> copy of the implementation (object) value and to have
> a different operator to do reference assignment.
>
> I don't belive that OO fan's would be happy without
> reference assignment?
>
> What do you think?

I think it's a bit late to change things now.

I suppose I just have to think about objects as pointers instead, with 
whatever they're pointing to being automatically managed:

ptrA := ptrB    ( programming equivalent of training wheels :-)

--
Bart