(: MetaJS)

Problem

Today compilers of computer programs analyze only the grammar of the language and ignore the semantics of each specific program. As a result, we get the source code with high level of redundancy. With the increasing size of the program this naturally leads to difficulties in its support. Major changes in the application source code become virtually impossible without causing of new bugs and regressions.

Solution

Imagine that you can formalize some aspects of the creation of computer programs. Not code itself, but the code generation process. Then the compiler will not just parse and compile the code, but partly will understand the meaning of your code and generate new code, depending on the surrounding context.

For example, you have a function save-thing that expects two required parameters: thing and db.

(defn save-thing (thing db))

Somewhere later in the program you call this function without parameters (save-thing), but the compiler knows that this function requires thing and db, finds these parameters in the current lexical scope and generates a complete call, for example:

(save-thing user-thing environment.db)

Elsewhere the program and, respectively, in another context compiler for the same original string (save-thing) will generate another call, for example:

(save-thing backup-thing (get-secondary-db))

If you provide all the parameters of the called function, the compiler will not think instead of you. It will just checks the compliance of the expected values to the passed parameters.

Which of the available symbols in the context of the call to be used as a function missing parameter, can be determined in several ways. The easiest way, by the name. We can look for symbol with the name thing in the context of the call and use it in place of missed parameter thing. Very often, parameters required by a function are defined with the same name in the outer scope, for example:

(defn process-thing (thing)
  (log thing)
  (save-thing))

Example of name matching shown above is the simplest mode that is used by MetaJS. You can find more information about semantic code transformations, examples of symbolic and entitative MetaJS to JavaScipt transformations in the metajs_semantic_code_transformations.pdf. Please look also at the high-level MetaJS language overview metajs_lisp.pdf.

Similarly, you can search for matches in the meta-information associated with function parameters and symbols available in the context of call. Such matching may be based on static typing or relations between symbols defined in logoses.

MetaJS allows compiler to generate source code. Will a computer create programs instead of a human?

The compiler will execute the instructions exactly as before, but in addition to grammar instructions, it will also execute semantic instructions defined specifically for your program.

Imagine that you're explaining how does your program work to a grandmother, who knows nothing about programming — it's an old grammar compiler. Now imagine that you're explaining the same thing to a girl with a degree in computer science — it's a new semantic compiler. But you will have to explain in both cases.

Source code

MetaJS is free open-source language available on GitHub. Please look also at the interactive documentation where you can try MetaJS without leaving your browser and see MetaJS magic in action.

Dmytro Dogadailo,
github:dogada, twitter:@d0gada.