[lixx]

Hi Foamers,

I have a question regarding the implementation of mappedVelocityFluxFixedValueFvPatchField::updateCo effs(). The code is listed below:

  162 void Foam::mappedVelocityFluxFixedValueFvPatchField::updateCoeffs()
  163 {
  164     if (updated())
  165     {
  166         return;
  167     }
  168 
  169     // Since we're inside initEvaluate/evaluate there might be processor
  170     // comms underway. Change the tag we use.
  171     int oldTag = UPstream::msgType();
  172     UPstream::msgType() = oldTag+1;
  173 
  174     // Get the mappedPatchBase
  175     const mappedPatchBase& mpp = refCast<const mappedPatchBase>
  176     (
  177         mappedVelocityFluxFixedValueFvPatchField::patch().patch()
  178     );
  179     const fvMesh& nbrMesh = refCast<const fvMesh>(mpp.sampleMesh());
  180     const word& fieldName = dimensionedInternalField().name();
  181     const volVectorField& UField =
  182         nbrMesh.lookupObject<volVectorField>(fieldName);
  183 
  184     surfaceScalarField& phiField = const_cast<surfaceScalarField&>
  185     (
  186         nbrMesh.lookupObject<surfaceScalarField>(phiName_)
  187     );
  188 
  189     vectorField newUValues;
  190     scalarField newPhiValues;
  191 
  192     switch (mpp.mode())
  193     {
  194         case mappedPolyPatch::NEARESTFACE:
  195         {
  196             vectorField allUValues(nbrMesh.nFaces(), vector::zero);
  197             scalarField allPhiValues(nbrMesh.nFaces(), 0.0);
  198 
  199             forAll(UField.boundaryField(), patchI)
  200             {
  201                 const fvPatchVectorField& Upf = UField.boundaryField()[patchI];
  202                 const scalarField& phipf = phiField.boundaryField()[patchI];
  203 
  204                 label faceStart = Upf.patch().start();
  205 
  206                 forAll(Upf, faceI)
  207                 {
  208                     allUValues[faceStart + faceI] = Upf[faceI];
  209                     allPhiValues[faceStart + faceI] = phipf[faceI];
  210                 }
  211             }
  212 
  213             mpp.distribute(allUValues);
  214             newUValues.transfer(allUValues);
  215 
  216             mpp.distribute(allPhiValues);
  217             newPhiValues.transfer(allPhiValues);
  218 
  219             break;
  220         }
  221         case mappedPolyPatch::NEARESTPATCHFACE:
  222         case mappedPolyPatch::NEARESTPATCHFACEAMI:
  223         {
  224             const label nbrPatchID =
  225                 nbrMesh.boundaryMesh().findPatchID(mpp.samplePatch());
  226 
  227             newUValues = UField.boundaryField()[nbrPatchID];
  228             mpp.distribute(newUValues);
  229 
  230             newPhiValues = phiField.boundaryField()[nbrPatchID];
  231             mpp.distribute(newPhiValues);
  232 
  233             break;
  234         }
  235         default:
  236         {
  237             FatalErrorIn
  238             (
  239                 "mappedVelocityFluxFixedValueFvPatchField::"
  240                 "updateCoeffs()"
  241             )   << "patch can only be used in NEARESTPATCHFACE, "
  242                 << "NEARESTPATCHFACEAMI or NEARESTFACE mode" << nl
  243                 << abort(FatalError);
  244         }
  245     }
  246 
  247     operator==(newUValues);
  248     phiField.boundaryField()[patch().index()] == newPhiValues;
  249 
  250     // Restore tag
  251     UPstream::msgType() = oldTag;
  252 
  253     fixedValueFvPatchVectorField::updateCoeffs();
  254 }

When the sampleMode is set to "nearestFace" (line 194-220), why there is a loop

199             forAll(UField.boundaryField(), patchI)
  200             {
  201                 const fvPatchVectorField& Upf = UField.boundaryField()[patchI];

and why not just directly pass the U and phi on the sampled patch to *this patch?

And Upf here is the boundary field (U field at boundary patches) rather than the U field of the sampled, internal patch (given by the keyword "offset" in file constant/polyMesh/boundary). But later

 206                 forAll(Upf, faceI)
  207                 {
  208                     allUValues[faceStart + faceI] = Upf[faceI];
  209                     allPhiValues[faceStart + faceI] = phipf[faceI];
  210                 }

the value on the boundary (Upf) is stored in allUValues and passed to *this patch. My question is where the sampled patch comes in? I suspect it is the following lines

  174     // Get the mappedPatchBase
  175     const mappedPatchBase& mpp = refCast<const mappedPatchBase>
  176     (
  177         mappedVelocityFluxFixedValueFvPatchField::patch().patch()
  178     );

but after checking the constructor of class mappedPatchBase, I did not see how it is achieved.

Can anybody give me a hint? Sorry if this is a silly question. I am now trying to implement a new boundary condition based on this one, but still have some problems.

When using pisoFoam, this updateCoeffs() will be called three times during each time step. Is this right? If I want to keep a running average of the sampled patch, should I do it during each call or do it only once during one time step? If latter, how can I do it? By comparing curTimeIndex with db().time().timeIndex()?