-
Notifications
You must be signed in to change notification settings - Fork 1
Home
Minh Pham edited this page May 13, 2020
·
11 revisions
The idea is that we use smaller components (we refer to them as adapters or building blocks) which we ’concatenate’ to form a transformation flow (a "pipeline"). This modular design allows us to reuse existing modules and wrap ready-scripts to create a language-and-format-independent module and pipeline. There are three types of components (adapters):
- Reader Adapter. Used as an entry point in the pipeline. It reads an input file (data) and a description of it (D-REPR language).
- Transformation Adapter. A class which performs a trans- formation done in a form of an API endpoint (remote or local) or an in-code script or library (i.e. using python or R). It does not materialize the data into an output, it just reproduces the data.
- Writer Adapter. Used as an exit point in the pipeline. It writes an output file based on a description file (D-REPR language).
List of currently supported adapters can be found here.