Re: [boost] [iterator] Using zip_iterator with std::pair

Subject: Re: [boost] [iterator] Using zip_iterator with std::pair
From: Claas H. Köhler (claas.koehler_at_[hidden])
Date: 2012-10-18 09:56:10

On 18/10/12 15:00, Jeffrey Lee Hellrung, Jr. wrote:
> On Thu, Oct 18, 2012 at 3:11 AM, "Claas H. Köhler" <claas.koehler_at_[hidden]>wrote:
> 
>>
>>
>> On 17/10/12 13:21, Jeffrey Lee Hellrung, Jr. wrote:
>>> On Wed, Oct 17, 2012 at 12:08 AM, "Claas H. Köhler" <
>> claas.koehler_at_[hidden]>wrote:
>>>
>>>>
>>>>
>>>> On 17/10/12 04:28, Jeffrey Lee Hellrung, Jr. wrote:
>>>>> On Tue, Oct 16, 2012 at 10:06 AM, "Claas H. Köhler" <
>>>> claas.koehler_at_[hidden]>wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>> On 16/10/12 14:27, Jeffrey Lee Hellrung, Jr. wrote:
>>>>>>> On Tue, Oct 16, 2012 at 4:21 AM, "Claas H. Köhler" <
>>>> claas.koehler_at_[hidden]
>>>>>>> wrote:
>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 16/10/12 12:01, Giovanni Piero Deretta wrote:
>>>>>>>>> On Tue, Oct 16, 2012 at 9:06 AM, "Claas H. Köhler" <
>>>>>> claas.koehler_at_[hidden]>
>>>>>>>> wrote:
>>>>>>>>>> Hi!
>>>>>>>>>>
>>>>>>>>>> I would like to use the boost::zip_iterator with std::pair.
>>>>>>>> Unfortunately this does not work out of
>>>>>>>>>> the box, but there exists a namespace
>>>>>>>>  boost::detail::tuple_impl_specific, which apparently needs to
>>>>>>>>>> be specialised for each tuple type.
>>>>>>>>
>>>>>>>
>>>>>>> Hmmm...I guess it makes sense you might have to specialize something
>> if
>>>>>>> zip_iterator doesn't specifically know about your desired tuple type.
>>>>>>>
>>>>>>>
>>>>>>>>>> I managed to re-implement tuple_meta_transform and
>>>>>>>> tuple_meta_accumulate for std::tuple, which seems
>>>>>>>>>> to be straightforward. However, there are functions of the form
>>>>>>>>>>
>>>>>>>>>> template<typename Tuple, typename Fun>
>>>>>>>>>> complex_result_type tuple_transform(const Tuple&, Fun)
>>>>>>>>>>
>>>>>>>>>> which have to be overloaded as well, if I am not mistaken.
>> However,
>>>>>>>> overloading this with something like
>>>>>>>>>> template<typename... Args, typename Fun>
>>>>>>>>>> complex_result_type tuple_transform(const std::tuple<Args...>&,
>> Fun)
>>>>>>>>>>
>>>>>>>>>> is probably not going to work, since this would be a partial
>>>> template
>>>>>>>> specialisation of a function.
>>>>>>>>>
>>>>>>>>> That's an overload actually and should work fine. Did you try it?
>>>>>>>>>
>>>>>>>>> HTH,
>>>>>>>>
>>>>>>>> Thanks for the hint. I have to admit I did not try it yet. I have
>> just
>>>>>>>> created a similar simple test
>>>>>>>> case, which indeed compiles/works fine. Shame on me for not trying
>> it
>>>>>> out.
>>>>>>>> For me this was always a
>>>>>>>> classical example of template function specialisation. I guess I
>>>> should
>>>>>>>> take a look at template
>>>>>>>> basics again :-)
>>>>>>>>
>>>>>>>> Sorry for that and thanks for the quick response.
>>>>>>>>
>>>>>>>
>>>>>>> I'm glad you got it working, but ideally, I think, zip_iterator
>> should
>>>>>> work
>>>>>>> out of the box for std::pair and std::tuple. Could I trouble you to
>>>>>> provide
>>>>>>> patches based on what you needed to get this to work?
>>>>>>>
>>>>>>> - Jeff
>>>>>>
>>>>>> Unfortunately things are not working yet. As far as I can judge, the
>>>>>> problem is that some of the
>>>>>> functions are called from within the zip_iterator.hpp header file,
>> which
>>>>>> is before I can actually
>>>>>> define the overloaded functions. As a result the overloaded functions
>>>> are
>>>>>> not called.
>>>>>>
>>>>>> Should I get this working, however, I am glad to share the wisdom.
>>>>>>
>>>>>
>>>>> Well, I'd consider it fair game to edit the zip_iterator.hpp header
>> file
>>>>> directly :)
>>>>>
>>>>> - Jeff
>>>>>
>>>> Looks like I found a working solution. I transformed the tuple_xxx
>>>> template functions into template
>>>> structs, for which overloading seemingly works fine. Since I am not
>>>> familiar with boost::mpl I used
>>>> my own versions of the meta transformations based on C++11. For tuple
>> this
>>>> may not be an issue, since
>>>> its a C++11 feature, but for std::pair it could be.
>>>>
>>>> Furthermore I would suggest to implement these functions exclusively for
>>>> boost::tuple and leave the
>>>> general template unimplemented. This may ease specialisations of other
>>>> classes. Using C++11 for
>>>> example the specialisations for std::pair and std::tuple are almost
>>>> identical, but have to be
>>>> specialised individually. Just my 0.02$ .
>>>>
>>>> Shall I post the modified code directly to the mailing list or somewhere
>>>> else?
>>>>
>>>
>>> Create a trac ticket, attach any patches, and I'll take a look at it
>> when I
>>> get a chance.
>>>
>>> - Jeff
>>>
>> I just created ticket #7526 with a proposed patch. I am not sure, whether
>> I got the different
>> compiler macros right, though, since I do not have access to the compilers
>> they are intented for.
>> Furthermore you may want to enclose the std::tuple stuff (including
>> namespace helper) into a macro,
>> which enables it if C++11 is used (I presume such a macro does already
>> exist, but I did not know its
>> exact syntax).
>>
> 
> Great, thanks! I'll try to get to this before 1.53.0 is released. Did you
> add a unit test by chance?
I wrote one but did not include it in the patch. Do you want a simple source file or is there a boost
template for unit tests?

> As for conditioning on the existence of std::tuple, looks like
> BOOST_NO_CXX11_HDR_TUPLE and BOOST_HAS_TR1_TUPLE would be macros to use.
Thanks.

By the way: Is there a way to remove attachments from track tickets? I accidentally attached a
non-working patch before actually attaching the working one. Unfortunately I did not manage to remove
the wrong attachment. If I start to modify the patch further, I might hit the restrictions of max.
space for attachments, if I cannot delete the deprecated ones.

Regards
Claas