Data composition is used in programming languages and data transfer representation languages (such as JSON, XML and network protocols) to build complex objects, but the particular language constructs may differ. The paper presents common primitives underlying structural data composition techniques in both groups of languages: sets, multisets, sequences, maps, multimaps and variant types. The primitives are described using the patterns methodology. Application of the patterns is considered with respect to the declarative parts of the languages (compile-time types in programming languages, XML Schema for XML) and usage parts (run-time types, JSON objects, XML documents). The patterns are intended for programming language designers, developers of high-level data transfer protocols, as well as students studying common approaches in different languages. The patterns may be used to facilitate conversion between different forms of data (e.g., arrays in memory and sequences of values in a data transfer channel).