Re: [boost] [fusion] is there an adapter for C-arrays?

Subject: Re: [boost] [fusion] is there an adapter for C-arrays?
From: vicente.botet (vicente.botet_at_[hidden])
Date: 2010-02-07 15:37:09

----- Original Message ----- 
From: "Christopher Schmidt" <mr.chr.schmidt_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Sunday, February 07, 2010 8:45 PM
Subject: Re: [boost] [fusion] is there an adapter for C-arrays?

> 
> vicente.botet schrieb:
>> ----- Original Message ----- 
>> From: "Christopher Schmidt" <mr.chr.schmidt_at_[hidden]>
>> To: <boost_at_[hidden]>
>> Sent: Sunday, February 07, 2010 2:19 AM
>> Subject: Re: [boost] [fusion] is there an adapter for C-arrays?
>>> vicente.botet schrieb:
>>>>> On 2/7/2010 3:34 AM, vicente.botet wrote:
>>>>>> Hi,
>>>>>>
>>>>>> I need to visit the members of a C++ type using Fusion sequences. For structures I plan
>>>>>> to use BOOST_FUSION_ADAPT_STRUCT. I need also to visit also c-arrays (with static
>>>>>> size). I see that there is already an adapter for boost::array. Is there an adapter for
>>>>>> C-arrays?
>>>>>
>>>>> Sorry, no. Would you want to contribute one? It's easy if you
>>>>> take a look at the boost::array adapter.
>>>>
>>>> Hi, 
>>>>
>>>> Well if doesn't exists I will need to see how to do it :) I will inform you when I reach to get something.
>>>
>>> I just added an implementation for adapting plain arrays to the c++0x
>>> branch of fusion. You can find the code here:
>>>
>>> https://svn.boost.org/trac/boost/browser/sandbox/SOC/2009/fusion/boost/fusion/adapted/array.hpp
>>> https://svn.boost.org/trac/boost/browser/sandbox/SOC/2009/fusion/boost/fusion/adapted/detail/po_array
>>>
>>>>
>>> [snip]
>>>>
>>>> Is there a simple way to implement this flat view? Or even better, is it already available?
>>>
>>> It is. Check the (undocumented) code in the ext_-directories.
>>>
>>> https://svn.boost.org/trac/boost/browser/trunk/boost/fusion/container/ext_/tree.hpp
>>> https://svn.boost.org/trac/boost/browser/trunk/boost/fusion/algorithm/iteration/ext_/for_each_s.hpp
>>>
>>> Unfortunately the code is pretty much broken at the moment.
>>>
>>> https://svn.boost.org/trac/boost/ticket/3858
>>>
>>> -Christopher
>> 
>> Hi,
>> 
>> thanks for all these pointers. I was starting to implement it,when DING, your post appeared. I'm sure I would made  a lot of errors before reaching to run the test.
> 
> I ported my code to the trunk. Update to the HEAD and include
> <boost/fusion/adapted/array.hpp> in your code.
> 
>> Do you know if it is planned to document the tree extension soon? Is there any post about this subject?
> 
> The segmented extension is pretty much proto specific in its current
> form. From my limited point of view, it still needs some work in order
> to be a useful component aside of Spirit/Proto-alike use-cases.
> Unfortunately Eric Niebler, the original author, has not continued
> development on this extension. For more information, see the old posts
> in the archives.
> 
> http://lists.boost.org/Archives/boost/2006/09/110708.php
> 
> As the code is broken, undocumented and may change in the future, I
> suggest you to connect your fusion sequences recursively via
> fusion::fold and fusion::join.
> 
>> Just a remark most of the .hpp files on po_array are not self contained as the structure po_array_tag is not declared in th file.
> 
> the *.hpp-files in the detail folder are not meant to be included by the
> user. po_array_tag is defined in /detail/tag_of.hpp, which is always
> included by array.hpp before any other /detail/*_impl.hpp-header.
> 
> -Christopher
> 

Hi, 

I've the impression you have introduced a interface break as before the array directory was for boost_array, and now is for POD arrays?

I would prefer a hierarchycal directory

fusion/adapted/boost/array.hpp
fusion/adapted/boost/mpl.hpp
fusion/adapted/boost/tuple.hpp
fusion/adapted/boost/detail/array
fusion/adapted/boost/detail/mpl
fusion/adapted/boost/detail/tuple
fusion/adapted/builtin/array.hpp 
fusion/adapted/builtin/struct.hpp
fusion/adapted/builtin/detail/array
fusion/adapted/builtin/detail/struct
fusion/adapted/std/pair.hpp
fusion/adapted/std/detail/pair

The old files can be made deprecated includig the new ones.

Now all the files are at the array level (detail directory has desapeared), so can be directly include by the user, thus my previous remark is still valid.

Update
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\detail
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\array_iterator.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\size_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\tag_of.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\value_at_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\value_of_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\category_of_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\is_sequence_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\at_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\begin_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\deref_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\is_view_impl.hpp
C:\cygwin\home\Vicente\boost\trunk\boost\fusion\adapted\array\end_impl.hpp
At revision: 59579

Best,
Vicente