FITS Files and Regular Grammars: A DMaSS Design Case Study

摘要

The NOAO Data Management and Science Solutions Platform has already passed through several development iterations and includes a Database (DB) Loader service. Experience using that service has been mixed. Recent discussion of the proposed Metadata service focused on vocabularies. We realized that we could treat FITS headers as a formal language and that this language was described by a regular grammar. It then became clear that part of the database loader's configuration was, in fact, a regular expression (expressed in several files of rather verbose, Spring-based XML!) Furthermore, a separate (and equally confusing) configuration was now clearly related. And we could extend things further: the same data structures could help detect "almost duplicate" data reliably. Reviewing our results in terms of the system architecture, we moved newly identified responsibilities to appropriate services. The end result is a cleaner design, better integrated within our architecture, and in which we have much more confidence. What started as an idle, theoretical curiosity yielded very practical results.