Simple typed information access stress test, useful for profiling, as well as for quickly checking high-level performance effects of changes (albeit not very accurately, obviously -- need longer running composite tests for such verifications).

Type of data is auto-detected, and is assumed to be homogenous. Basically, data is either within attributes, or as element content, but not both. In either case structure should be shallow, with the root and only immediate leaf-level elements containing attribute or element data. Type of this data is auto-detected from the first instance; data must be in canonical format to be properly recognized (non-first values can be non-canonical).