Abstract: Aspects of the present invention provide a solution for compiling data. In an embodiment, an input query is received in a first language. The input query is translated to a nested relational algebra (NRA) in the form of a first intermediate representation (IR), wherein the first IR comprises a high-level functional language including algebraic operators. A set of algorithms of the algebraic operators is implemented to compile at least some of the first IR into a second IR, wherein the second IR comprises a data-flow language. At least one of the first IR and the second IR is compiled into a low-level code.

Abstract: Aspects of the present invention provide a solution for compiling data. In an embodiment, an input query is received in a first language. The input query is translated to a nested relational algebra (NRA) in the form of a first intermediate representation (IR), wherein the first IR comprises a high-level functional language including algebraic operators. A set of algorithms of the algebraic operators is implemented to compile at least some of the first IR into a second IR, wherein the second IR comprises a data-flow language. At least one of the first IR and the second IR is compiled into a low-level code.

Abstract: Aspects of the present invention provide a solution for compiling data in a plurality of modes. In an embodiment, at least one optimal mode is specified for each of a set of program language constructs and each of a set of language primitives in a first language. A set of optimal mode code is generated in the at least one mode in a second language. A set of bridge code is generated. A set of additional mode code is generated in a plurality of other modes in the second language, wherein the generating utilizes the bridge code. The generated optimal mode code and additional mode code is compiled.