Feature-oriented systems describe system variants through features as first-class abstractions of optional or incremental units of systems functionality. The choice how to treat modularity and composition in feature-oriented systems strongly influences their design and behavioral modeling. Popular paradigms for the composition of features are superimposition and parallel composition. We approach both in a unified formal way for programs in guarded command language by introducing compositional feature-oriented systems (CFOSs). We show how both compositions relate to each other by providing transformations that preserve the behaviors of system variants. Family models of feature-oriented systems encapsulate all behaviors of system variants in a single model, prominently used in family-based analysis approaches. We introduce family-ready CFOSs that admit a family model and show by an annotative approach that every CFOS can be transformed into a family-ready one that has the same modularity and behaviors.