Different scientific workflow languages have been developed to help programmers in designing complex data analysis procedures. However, little effort has been invested in comparing and finding a common root for existing approaches. This work is motivated by the search for a scientific workflow language which coherently integrates different aspects of distributed computing. The language proposed is data-driven for easing the expression of parallel flows. It leverages array programming principles to ease data-intensive applications design. It provides a rich set of control structures and iteration strategies while avoiding unnecessary programming constructs. It allows programmers to express a wide set of applications in a compact framework.