Refactoring MMG5_chkBdryTria

[corentin-prigent]

This update refactorizes the function MMG_chkBdryTria. It is now written as a sequence of several sub-functions. It does not modify the behaviour or features of mmg. However, this update is necessary for parmmg, in order to take into account the fact that we pack the mesh->tria array when extra triangles are deleted from it.
This update primarily ensures the maintainability of the code and avoids redundancy in mmg and parmmg.

[Algiane]

I think that we can remove / extract this from the loop and just delete extra triangles inside the loop.
Then, after the loop (so outside the MMG5_chkbdryTria_deleteExtraBoundaries() func), it should be possible to :

• travel triangles
• detect the deleted tria
• pack the array accordingly
• update the perm array for ParMmg only

[Algiane]

something like that :

nt=0; nbl=1;
for (k=1; k<=mesh->nt; k++) {
    ptt = &mesh->tria[k];
    
    if ( !MG_EOK(ptt) ) continue;
    
    ++nt;
    if ( k!=nbl ) {
      pttnew = &mesh->tria[nbl];
      memcpy(pttnew,ptt,sizeof(MMG5_Tria));
      if ( permtria ) {
         permtria[k] = nbl;
       }
    }
}
    

[Algiane]

Or more elegant but also more error-prone (I am not sure that it works ;-) )

    k = 1;
    do {
      ptt = &mesh->tria[k];
      if ( !MG_EOK(ptt) ) {
        pttnew = &mesh->tria[mesh->nt];
        assert( ptt && pttnew && MG_EOK(pttnew) );
        memcpy(ptt,pttnew,sizeof(MMG5_Tria));

       if ( permtria ) permtria[mesh->nt] = k;

        ptt->v[0] = 0;
        mesh->nt--;
      }
    }
    while ( ++k < mesh->nt );

[Algiane]

something like that :

nt=0; nbl=1;
for (k=1; k<=mesh->nt; k++) {
    ptt = &mesh->tria[k];
    
    if ( !MG_EOK(ptt) ) continue;
    
    ++nt;
    if ( k!=nbl ) {
      pttnew = &mesh->tria[nbl];
      memcpy(pttnew,ptt,sizeof(MMG5_Tria));
      if ( permtria ) {
         permtria[k] = nbl;
       }
    }
}
    

[Algiane]

I think that here we can insert the tria deletion in order to perform the compression of triangle array outside the loop (see comment where the array is currently packed).

With something like : ptt->v[0] = 0; MG_EOK(ptt) should be false.

[Algiane]

Or more elegant but also more error-prone (I am not sure that it works ;-) )

    k = 1;
    do {
      ptt = &mesh->tria[k];
      if ( !MG_EOK(ptt) ) {
        pttnew = &mesh->tria[mesh->nt];
        assert( ptt && pttnew && MG_EOK(pttnew) );
        memcpy(ptt,pttnew,sizeof(MMG5_Tria));

       if ( permtria ) permtria[mesh->nt] = k;

        ptt->v[0] = 0;
        mesh->nt--;
      }
    }
    while ( ++k < mesh->nt );

[codecov]

Codecov Report

Attention: Patch coverage is 65.48673% with 39 lines in your changes are missing coverage. Please review.

Project coverage is 43.98%. Comparing base (058f9a3) to head (cf436e4).

Files	Patch %	Lines
src/mmg3d/hash_3d.c	65.48%	17 Missing and 22 partials ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #259      +/-   ##
===========================================
+ Coverage    43.96%   43.98%   +0.01%     
===========================================
  Files          179      179              
  Lines        54102    54124      +22     
  Branches     10249    10252       +3     
===========================================
+ Hits         23786    23805      +19     
  Misses       22599    22599              
- Partials      7717     7720       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Algiane]

Thanks!