Re: [Boost-users] [multi_key] problems using modify_key with a composite_key index

From: Jeffrey Holle (jeffreyholle_at_[hidden])
Date: 2007-10-02 09:59:59

Jeffrey Holle wrote:
 > Joaquín Mª López Muñoz wrote:
 >> So, instead of using modify_key resort to plain modify instead:
 >>
 >> 
c.get<1>().modify(c.get<1>().find(make_tuple(y,UINT_MAX)),&ll::_1->*&entry::m_x=startIter2->m_x);
 >>
 >> Is your problem solved now?
 >
 > This solved my compilation problem, but I ended up with an app than 
wants to runs forever.
 >
 > To be clear, I'm including my entire function.  What it is supposed 
to do is reorder the x values of the composite_key index of the 
multi_index container based on the ordering of a vector<float> 
container.  It is a swap_sort algorithm because its the only way that 
I've imagined how to do this.  A better solution would definitely be 
entertained.
 >
 > void sortByBaryCenterValue(Positioner& p, BaryCenterVals& bcv,size_t y)
 > {
 >    Positioner::XYIterator iter,iter_end;
 >    tie(iter,iter_end)=p.getXYIndex().equal_range(make_tuple(y));
 >    for (size_t nStartIndex=0;nStartIndex<bcv.size();++nStartIndex) {
 >      size_t nSmallestIndex=nStartIndex;
 >      for (size_t 
nCurrentIndex=nStartIndex+1;nCurrentIndex<bcv.size();++nCurrentIndex)
 >        if (bcv[nCurrentIndex]<bcv[nSmallestIndex])
 >          nSmallestIndex=nCurrentIndex;
 >        swap(bcv[nStartIndex],bcv[nSmallestIndex]);
 >        Positioner::XYIterator startIter(iter),smallestIter(iter);
 >        advance(startIter,nStartIndex);
 >        size_t x1=startIter->m_x;
 >        advance(smallestIter,nSmallestIndex);
 >        size_t x2=smallestIter->m_x;
 >
 > 
p.getXYIndex().modify(p.getXYIndex().find(make_tuple(y,x1)),&ll::_1->*&Positioner::entry::m_x=UINT_MAX);
 > // 
p.getXYIndex().modify(p.getXYIndex().find(make_tuple(y,x2)),&ll::_1->*&Positioner::entry::m_x=x1);
 >
 > 
p.getXYIndex().modify(p.getXYIndex().find(make_tuple(y,UINT_MAX)),&ll::_1->*&Positioner::entry::m_x=x2);
 >    }
 > }
 >
 > If I uncomment the middle line of the sort attempt, the application 
runs forever.
 >
 > Do you know what is wrong?
 > Thanks in advance.

Found my problem.  A sort-swap algorithm has a self-swap "feature". 
When x1 and x2 are the same, the above function will lock up (when the 
offending line is uncommented.

The solution is to put the swap operation within a conditional clause.