wardle / dmd Goto Github PK

The current import takes 10 minutes which is too slow. Most of the bottlenecks could be parallelised quite easily when time permits. It would be better to stream the XML fragments in batches with associated context/type information and then parse those data in parallel.

The current approach gets all SNOMED concepts, but we can be much more sophisticated and more efficient by getting all VMPs for a given ATC and then simply getting VTMs and then building expressions for the minimal set - including IS-A HAS-VMP etc. relationships. That would avoid maxing out the size of a single expression, which for more generic purposes, e.g. all cardiovascular drugs, will probably create an expression that can't be realistically parsed. For my current usage pattern, this isn't an issue at all.

VMP result doesn't de-normalise the STRNT_NMRTR_UOMCD or STRNT_DNMTR_UOMCD units of measure, forcing client applications to resolve them manually.

Need a record of dm+d release, and software versions, in case of schema changes and potential incompatibility. Better to fail with a version error, than some obscure error from the internals.