Combining deep and shallow embedding of domain-specific languages

Folding domain-specific languages

“…Good practice in the design of embedded DSLs is to distinguish between a minimal 'core' language and a more useful 'everyday' language [26]. The former is more convenient for the language designer, but the latter more convenient for the language user.…”

The Geometry of Computation-Graph Abstraction

“…Reaching a language comparable in sophistication and e ciency with TF is a long path, but we are making the rst steps in that direction [3]. e advantages of using a stand-alone language, especially when there is evidence that it has a reasonably well behaved semantics, are signi cant, as EDSLs su er from well known pitfalls [18].…”

Formalising Executable Specifications of Low-Level Systems

“…Differences and complementarity between shallow and deep embedding have been widely discussed in functional programming, in relationship with the development of embedded domain specific languages (EDSLs) [24,25,26,27]. Combinations of shallow and deep embedding have been proposed e.g.…”

“…Combinations of shallow and deep embedding have been proposed e.g. in [25] to deal at once with the expression problem (related to extending a deeply embedded language) and the interpretation problem (related to extending the semantic interpretation of a shallow embedding). Their approach consists in extending a deeply embedded core language with a shallowly embedded frontend, thus the opposite of what we do with a DLE.…”