Re: [Cppcms-users] cppcms::json::value desing question

[Daniel V. <chi...@gm...>]

On Mon, 2010-03-15 at 22:45 -0700, Artyom wrote:
> > Hello Artyom
> >
> >I think that boost::variant type fits exactly as base class for
> > cppcms::json::value.
> >
> > I think that one pro of this approach is that boost::variant interface
> > is more familiar and has good visitor interface.
> > 
> > You think about that?
> > 
> > This break your ABI requirerements?
> 
> Exactly. See http://cppcms.sourceforge.net/wikipp/en/page/cppcms_1x_coding_standards#abi_boost
> 
> I can't use Boost libraries in CppCMS interfaces for very simple
> reason: they ABI can be easily broken even with simple define addition
> in headers. For example boost::shared_ptr changes its strucutres layout
> with special define.
> 
> So I can't use Boost in iterfaces. 2nd issue is that AFAIK boost::variant
> does not have virtual descructor so deriving from  it is bad idea.
As a comment I think no virtual destructor is OK in this case because
the derivation has no useful polymorphic usage. But you are right.

> However I use exactly boost::variant in implementation of cppcms::json.

> 
> > Another question. You define json::object as std::map, which it has
> > O(log n) complexity for access
> >
> > Why you not use boost::unordered_map which it has O(n) complexity?

As you can see I made a mistake, unorderer_map has constant mean
complexity.

> Same reason as above, untill std::unordered_map would be availible
> in major compilers I would use std::map.
> 
> Also, please note, that for relatively small maps, std::map would be 
> even faster (and this is the case in most json objects).

Very good rationale point. 

You convinced me in all points.


> Artyom
> 
> Please, register to mailing list so I would not have to moderate
> incoming messages.

> 
>       
> 
> ------------------------------------------------------------------------------
> Download Intel® Parallel Studio Eval
> Try the new software tools for yourself. Speed compiling, find bugs
> proactively, and fine-tune applications for parallel performance.
> See why Intel Parallel Studio got high marks during beta.
> http://p.sf.net/sfu/intel-sw-dev
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users