The extension requires a top file with the name CGNS_VAL_USER_<key>.py with the CGNS_VAL_USER_Checks declaration. This file has to be importable, which means the file should be in one of the PYTHONPATH directories.
Writing an extension requires good Python skills. The user has to define a new class, with CGNS.VAL.grammars.SIDS.SIDSbase as one of its base class, a set of per-node check methods and a set of associated diagnostics.
A diagnostic is composed of a key, a level and a message:
('U101',CGM.CHECK_WARN,'No Zone in this Base')
The CGM.CHECK_WARN enumerate comes from the CGNS.VAL.parse.messages module, in that example impported as CGM.
All diagnostics are set into the class as a dictionnary, using the method addMessages of the log object of the class. This is performed in the __init__ method of the user class:
def __init__(self,log): CGS.SIDSbase.__init__(self,log) self.log.addMessages(USER_MESSAGES)
Do not forget the base class initialisation. The log argument is managed by the CGNS.VAL internals, you should not take care about it.
The CGNS.VAL internals are in charge of parsing the tree. Each time a node is entered, the corresponding check function is called. The functions args are always the same: the path of the node in which you are entering, the node itself as a CGNS/Python list, the parent node as a CGNS/Python list, the complete CGNS/Python tree and the log object.
The function should have the name of the entered node type. For example, the Zone_t is called each time the parser enters into a Zone_t node:
def Zone_t(self,pth,node,parent,tree,log): rs=self.sids.Zone_t(self,pth,node,parent,tree,log) if (CGK.GridCoordinates_s not in CGU.childNames(node)): rs=log.push(pth,NOGRIDZONE) if (not CGU.hasChildNodeOfType(node,CGK.ReferenceState_ts)): rs=log.push(pth,NOZREFSTATE) return rs
The first step is to call the Zone_t for the SIDS checker. This is strongly recommanded, but not mandatory... The rs variable contains the current status for this node: that can be CHECK_GOOD,`CHECK_FAIL` or CHECK_WARN. This status should be the return of your check function.
We check values of the node and in case of problem, we push the diagnostic into the log. The path of the node is pushed as well.
Now each check is a test on several values of the current node, the use of CGNS.PAT.cgnsutils, CGNS.PAT.cgnstypes and CGNS.PAT.cgnskeywords would help.
You should not parse the tree by yourself, unless you have to check consistency with other node values. The example below shows a control on the mandatory ElementRange_s node of an Elements_t node (The context object is detailled in the next section).:
def Elements_t(self,pth,node,parent,tree,log): rs=CGM.CHECK_OK if (CGU.getShape(node)!=(2,)): rs=log.push(pth,BADVALUESHAPE) else: et=node eb=node self.context[CGK.ElementType_s]=et self.context[CGK.ElementSizeBoundary_s]=eb if (et not in range(0,len(CGK.ElementType)+1)): rs=log.push(pth,UKELEMTYPE) if (eb==0): bad_eb=False elif (eb<0): bad_eb=True else: bad_eb=True ecnode=CGU.getNodeByPath(tree,pth+'/'+CGK.ElementRange_s) if ( (ecnode is not None) and (CGU.getShape(node)==(2,)) and (CGU.getValueDataType(ecnode)==CGK.I4) and (ecnode>eb)): bad_eb=False if (bad_eb): rs=log.push(pth,BADELEMSZBND) return rs
The parser has a context for global and local data. It is a dictionnary with SIDS names as keys, the values are overwritten during the parse but a value is always correct for a given sub-tree. For example to set the base dimension attributes:
cd=node pd=node self.context[CGK.CellDimension_s]=cd self.context[CGK.PhysicalDimension_s]=pd
And to get them later on in another node function:
if (not CGS.transformIsDirect(tr,self.context[CGK.CellDimension_s])): rs=log.push(pth,NOTRHTRANSFORM)
In this test, CGS stands for CGNS.APP.sids.utils.