Re: persistence : current state, and plans for future ?

From: krempp_at_[hidden]
Date: 2001-04-17 17:15:53

> We've moved the files into CVS under the branch 
> "persistence-initial".  Jens and I have had some email over the 
last couple 
> of days, and I've written another test program, which got put in 
the CVS 
> branch yesterday.

Hum.. 
I checked out branch "persistence-initial" but I am not sure this is 
actually what you are speaking about,
since the persistence_demo.cpp still throws (probleme of EOF and 
file.exceptions set in definiton of save_file )

PLUS there is an obvious bug here :

template<class T, class Reader, class OutputIterator>
inline void load_sequence(Reader & reader, OutputIterator out)
{
  std::size_t n;
  reader.start_sequence(n);
  T tmp;  // BOOOUUH !
  while(reader.good() && n-- > 0 && !reader.end_sequence()) {
    load(reader, tmp);
    *out = tmp;
    ++out;
  }
}

when T is something like a vector<vector>, load(reader,tmp) will do 
back_insertions into tmp, and thus tmp needs to be initialised at 
*each iteration*, if you dont want to accumulate all elts, step 
after step, making something like 
[ [0] [1] [2] ]
-load->
[ [0] [0 1] [0 1 2] ]

movig 'T tmp' inside the loop is definitely needed.

> Overall, the Persistence Library seems pretty strong to me.

It's truly a nice library :  very useful though pretty lightweight,
so it's fast to get used-to.

> There are a number of compiler issues, particularly with wchar_t 
and FP 
> types, the documentation needs a bit of updating, and more work 
needs to be 
> done on the XDR reader/writer, but most of these seem pretty minor 
to me.

about the XDR : with my g++ (3.0-pre) I had to define 
void write(std::size_t n)
void write(int n)
[ and same with laod)
because it was called with those type, but they did not match any of 
the boost *least32 types.
Is it normal ?

( I dont really care about XDR, but this prevented me from compiling 
test1, so I felt the need to make it run..)

-- 
Sam