Problem reading ROOT branch containing Strings with UnROOT.jl

jling · June 17, 2024, 7:01pm

JuliaHEP/UnROOT.jl/blob/01969242f7006350ba0f194e615be63fd448df48/src/root.jl#L480


      
                  # since no classname were found, we now try to determine
                  # type based on leaf information
                  _type, _jaggtype = leaf_jaggtype(leaf, _jaggtype)
              end
              return _type, _jaggtype
          end
          
          function leaf_jaggtype(leaf, _jaggtype)
                  _type = primitivetype(leaf)
                  leafLen = leaf.fLen
                  if leafLen > 1 # treat NTuple as Nojagg since size is static
                      _fTitle = replace(leaf.fTitle, "[$(leafLen)]" => "")
                      # looking for more `[var]`
                      m = match(r"\[\D+\]", _fTitle)
                      _type = FixLenVector{Int(leafLen), _type}
                      if isnothing(m)
                          return _type, Nojagg
                      else
                          #FIXME this only handles [var][fix] case
                          return Vector{_type}, Nooffsetjagg
                      end

this is the root cause of the bug – in this case even though leafLen is > 1, somehow it’s not really fixed length?

github.com

root-project/root/blob/86a5a713474e37d96735231da74c000056d86d0c/tree/tree/inc/TLeaf.h#L72


      
          
          protected:
          
             using Counts_t = std::vector<Int_t>;
             struct LeafCountValues {
                Counts_t fValues;
                Long64_t fStartEntry{-1}; ///<! entry number of corresponding to element 0 of the vector.
             };
          
             Int_t            fNdata;           ///<! Number of elements in fAddress data buffer.
             Int_t            fLen;             ///<  Number of fixed length elements in the leaf's data.
             Int_t            fLenType;         ///<  Number of bytes for this data type
             Int_t            fOffset;          ///<  Offset in ClonesArray object (if one)
             bool             fIsRange;         ///<  (=true if leaf has a range, false otherwise).  This is equivalent to being a 'leafcount'.  For a TLeafElement the range information is actually store in the TBranchElement.
             bool             fIsUnsigned;      ///<  (=true if unsigned, false otherwise)
             TLeaf           *fLeafCount;       ///<  Pointer to Leaf count if variable length (we do not own the counter)
             TBranch         *fBranch;          ///<! Pointer to supporting branch (we do not own the branch)
             LeafCountValues *fLeafCountValues; ///<! Cache of collection/array sizes
          
             TLeaf(const TLeaf&);
             TLeaf& operator=(const TLeaf&);

according to ROOT, this should indeed be fixed length