Metaprogramming is the writing of computer programs that write or manipulate other programs (or themselves) as their data, or that do part of the work at compile time that would otherwise be done at runtime. In some cases, this allows programmers to minimize the number of lines of code to express a solution (hence reducing development time), or it gives programs greater flexibility to efficiently handle new situations without recompilation.
The language in which the metaprogram is written is called the metalanguage. The language of the programs that are manipulated is called the object language. The ability of a programming language to be its own metalanguage is called reflection or reflexivity.
Reflection is a valuable language feature to facilitate metaprogramming. Having the programming language itself as a first-class data type (as in Lisp, Forth or Rebol) is also very useful; this is known as "homoiconicity". Generic programming invokes a metaprogramming facility within a language, in those languages supporting it.
Metaprogramming usually works in one of three ways. The first way is to expose the internals of the run-time engine to the programming code through application programming interfaces (APIs). The second approach is dynamic execution of expressions that contain programming commands, often composed from strings, but can also be from other methods using arguments and/or context. Thus, "programs can write programs." Although both approaches can be used in the same language, most languages tend to lean toward one or the other.
The third way is to step outside the language entirely. General purpose program transformation systems, which accept language descriptions and can carry out arbitrary transformations on those languages, are direct implementations of general metaprogramming. This allows metaprogramming to be applied to virtually any target language without regard to whether that target language has any metaprogramming abilities of its own.