[m-rev.] diff 1/3: Move pointer_equals/2 into the private_builtin module.

[Paul Bone]

On Sun, Jun 30, 2013 at 04:46:49AM +1000, Julien Fischer wrote:
> On Fri, 28 Jun 2013, Paul Bone wrote:
>
>> My other two patches in this series are still okay, and the second one is
>> still waiting for review.
>>
>> Move pointer_equals/2 into the private_builtin module.
>>
>> library/version_hash_table.m:
>> library/private_builtin.m:
>>    As above.
>>
>> library/private_builtin.m:
>>    Adjust a comment to say that some predicates are implemented in the
>>    normal way in this module.
>> ---
>> library/private_builtin.m    | 40 ++++++++++++++++++++++++++++++++++++++--
>> library/version_hash_table.m | 30 +-----------------------------
>> 2 files changed, 39 insertions(+), 31 deletions(-)
>>
>> diff --git a/library/private_builtin.m b/library/private_builtin.m
>> index 1dd866d..cd3bffc 100644
>> --- a/library/private_builtin.m
>> +++ b/library/private_builtin.m
>> @@ -23,8 +23,9 @@
>> % does not get included in the Mercury library reference manual.
>> %
>> % Many of the predicates defined in this module are builtin - they do not have
>> -% definitions because the compiler generates code for them inline. Some others
>> -% are implemented in the runtime.
>> +% definitions because the compiler generates code for them inline. A second
>> +% group are implemented in the runtime.  Whilst a third group are
>> +% implemented normally in this module.
>> %
>> %-----------------------------------------------------------------------------%
>> %-----------------------------------------------------------------------------%
>> @@ -118,6 +119,11 @@
>>     %
>> :- pred typed_compare(comparison_result::uo, T1::in, T2::in) is det.
>>
>> +    % True if the two objects have the same pointer value in the underlying
>> +    % implementation.
>> +    %
>
> I suggest something along the lines of:
>
>      True iff the the two terms occupy the same address in memory.
>      This is useful as a cheap test of equality when implmenting
>      user-defined equality.

Okay.

> (Can I object to the name pointer_equals on the grounds that only a
> minority of our current target languages actually have a notion of
> pointers?  ;-) )

It this a joke based on my clauses/alternatives bug?

Note that even in Java the concept of 'pointer equality' is still present
even though they don't have pointers.  In a language with mutable data this
abstraction will always leak.

>> +:- pred pointer_equals(T::in, T::in) is semidet.
>> +
>
> I suggest s/pointer_equals/pointer_equal/
>

Okay.

> (Likewise, version_hash_table.equals/2 should really have been called
> version_hash_table.equal/2 for the purposes of consistency with the
> rest of the stdlib.)

It's probably not too late to rename those without breaking much code and
using the usual method of deprecating the old names.

Thanks.


-- 
Paul Bone
http://www.bone.id.au