Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ddlm-group] Split _model_site.adp_eigen_system into eigen_vectorand eigen_value

Dear DDLm gurus,

As part of tidying up the core CIF dictionary in preparation for Vol G publication, we have come across the following fundamental issue: only one unit can be specified in a definition, which means that those units apply to all elements of a matrix or list. However, _model_site.adp_eigen_system bundles both eigenvalues (dimensionless) and eigenvectors (dimensioned) in one matrix. This is purely for computational convenience, because eigen-vectors and -values are efficiently computed together. Note that this data name cannot and did not exist in the old DDL1 dictionary, so we have considerable freedom to fiddle.

I would like to suggest that we split this into two data names corresponding to the eigenvectors and eigenvalues. The dREL definitions will precisely define how they are supposed to be used and how they relate to one another.

My main reason for approaching it this way is that compound data names whose components are handled differently by code should be split into separate items to reflect the implied distinct meanings. We have done this for all other cases in the core dictionary. Differing units is an obvious criterion for separating components, as different units implies that these components have different meanings.

Other alternatives that I find less attractive are:
(1) Enhance the DDLm _units.code attribute to somehow handle compound data items - complex and disproportionate for a single data name (there are no others in cif core)
(2) Add a _units.code enumerated value of something like 'derived' which signals that the units should be derived from the dREL calculation, which in this case would simply join the eigenvectors and eigenvalues together. Possible but excludes software that doesn't want to interpret dREL just to get units.

Please let me know your thoughts. If none are forthcoming, I'll go with my preference.

T +61 (02) 9717 9907
F +61 (02) 9717 3145
M +61 (04) 0249 4148
ddlm-group mailing list

Reply to: [list | sender only]