Visual analytics has been widely studied in the past decade. One key to make visual analytics practical for both research and industrial applications is the appropriate definition and implementation of the visual analytics pipeline which provides effective abstractions for designing and implementing visual analytics systems. In this paper we review the previous work on visual analytics pipelines and individual modules from multiple perspectives: data, visualization, model and knowledge. In each module we discuss various representations and descriptions of pipelines inside the module, and compare the commonalities and the differences among them.