codeasdata

Code as data is the practice of representing source code itself as data that can be inspected, transformed, and executed by software. The central idea is that code and data share representations, enabling metaprogramming, code generation, and program analysis by machines, as well as by humans when needed.

Historically, Lisp and Scheme pioneered code-as-data through their homoiconic syntax, where code and data use the same structure, such as lists. This makes macros possible: code can generate, modify, and expand other code at compile time. Other languages provide analogous facilities via parsing into abstract syntax trees (ASTs), reflection, or runtime code generation, allowing programs to treat code as a manipulable data structure.

Techniques include macro systems, AST-based tooling, templates, and code generators used in compilers, transpilers, and domain-specific languages. Code as data enables program analysis, transformation, instrumentation, and automated refactoring. It supports metaprogramming, where software writes or adapts its own code, and dynamic scripting where code is produced on demand or loaded from external sources.

Benefits include increased expressiveness, reuse, and automation, while challenges involve readability, maintainability, and potential bugs from complex metaprogramming. Performance and safety concerns arise when evaluating or expanding large or untrusted code fragments. Good practice often relies on clear meta-programming constructs and separation between data and code where possible.

See also: macro systems, metaprogramming, abstract syntax tree, reflection, code generation, Lisp.