MetadataDrivenAction¶
Definition¶
MetadataDrivenActions are Acts in which their logic is defined by validation and transformation rules expressed in some standard rule language.
Synonyms¶
Explanatory notes¶
Whereas ComputationActions are used in statistical packages like SPSS, Stata, SAS and R to perform data management and data transformations, MetadataDrivenActions are used by data pipelines, e.g. ETL (Extract Transform Load) platforms, along with ComputationActions. In a data pipeline the user is presented with a menu of MetadataDrivenActions that are captured here in an external controlled vocabulary. Users enter into a dialog with the platform through which they customize the MetadataDrivenAction. The user writes no code. The dialog is saved. In the course of this dialog the user might specify a data source, specify which variables to keep or drop, rename variables, specify join specifics, create a value map between two variables and so forth. MetadataDrivenActions represent this dialog in some standard rule language, e.g. Structured Data Transformation Language (SDTL) , Validation and Transformation Language (VTL), etc.. In fact, MetaDataDrivenAction can use VTL with certain “extensions”. Together they form quasi-VTL. quasi-VTL in turn conforms to a known schema. Using this schema users can render the quasi-VTL as XHTML and record it in the MetadataDrivenAction quasiVTL property. quasiVTL supports structured text. Note that we would like NOT to create specific subtypes of MetadataDrivenAction – one for each transformation. In fact there are a dozen or more such actions that will be captured in the MetadataDrivenAction types external controlled vocabulary. As an alternative, users can optionally write structured text based on a known schema. In this approach we are able to “evolve” SDTL as needed without changing our model. Instead, as our language “evolves”, we one have to update its schema.