Composition Patterns in Data Flow Based Mashups

Recently, mashup tools have emerged as popular end-user development platform. Composition languages used in mashup tools provide ways (drag-and-drop based visual metaphor for programming) to integrate data from multiple data sources in order to develop situational applications. However this integration task often requires substantial technical expertise from the developers in order to use basic composition blocks properly in their composition logic. Reusing of existing composition knowledge is one of the possible solutions to ease mashup development process. This reusable composition knowledge can be harvested from composition patterns that have occurred frequently in previously developed mashup. In order to understand composition patterns in mashups, particularly in data flow based mashups, in this paper, we have analyzed the composition language used by one of the most popular data-flow based mashup tools, Yahoo! Pipes. Based upon our analysis we have identified six composition patterns, which represent most commonly used composition steps during mashup application development. To prove the generality of the identified patterns in data-flow based mashup composition languages, we have further shown the applicability of our composition patterns in several other popular data-flow based mashup tools.