Skip to content

Latest commit

 

History

History
175 lines (107 loc) · 4.18 KB

parse_trans_codegen.md

File metadata and controls

175 lines (107 loc) · 4.18 KB
Raw

Module parse_trans_codegen

Parse transform for code generation pseduo functions.

Authors: : Ulf Wiger (ulf@wiger.net).

Description

...

Function Index

format_error/1
parse_transform/2 Searches for calls to pseudo functions in the module codegen, and converts the corresponding erlang code to a data structure representing the abstract form of that code.

Function Details

format_error/1

format_error(E) -> any()

parse_transform/2


parse_transform(Forms, Options) -> NewForms

Searches for calls to pseudo functions in the module codegen, and converts the corresponding erlang code to a data structure representing the abstract form of that code.

The purpose of these functions is to let the programmer write the actual code that is to be generated, rather than manually writing abstract forms, which is more error prone and cannot be checked by the compiler until the generated module is compiled.

Supported functions:

gen_function/2

Usage: codegen:gen_function(Name, Fun)

Substitutes the abstract code for a function with name Name and the same behaviour as Fun.

Fun can either be a anonymous fun, which is then converted to a named function, or it can be an implicit fun, e.g. fun is_member/2. In the latter case, the referenced function is fetched and converted to an abstract form representation. It is also renamed so that the generated function has the name Name.

Another alternative is to wrap a fun inside a list comprehension, e.g.


  f(Name, L) ->
      codegen:gen_function(
          Name,
          [ fun({'$var',X}) ->
               {'$var', Y}
            end || {X, Y} <- L ]).

Calling the above with f(foo, [{1,a},{2,b},{3,c}]) will result in generated code corresponding to:


  foo(1) -> a;
  foo(2) -> b;
  foo(3) -> c.

gen_functions/1

Takes a list of {Name, Fun} tuples and produces a list of abstract data objects, just as if one had written [codegen:gen_function(N1,F1),codegen:gen_function(N2,F2),...].

exprs/1

Usage: codegen:exprs(Fun)

Fun is either an anonymous function, or an implicit fun with only one function clause. This "function" takes the body of the fun and produces a data type representing the abstract form of the list of expressions in the body. The arguments of the function clause are ignored, but can be used to ensure that all necessary variables are known to the compiler.

gen_module/3

Generates abstract forms for a complete module definition.

Usage: codegen:gen_module(ModuleName, Exports, Functions)

ModuleName is either an atom or a {'$var', V} reference.

Exports is a list of {Function, Arity} tuples.

Functions is a list of {Name, Fun} tuples analogous to that for gen_functions/1.

Variable substitution

It is possible to do some limited expansion (importing a value bound at compile-time), using the construct {'$var', V}, where V is a bound variable in the scope of the call to gen_function/2.

Example:


  gen(Name, X) ->
     codegen:gen_function(Name, fun(L) -> lists:member({'$var',X}, L) end).

After transformation, calling gen(contains_17, 17) will yield the abstract form corresponding to:


  contains_17(L) ->
     lists:member(17, L).

Form substitution

It is possible to inject abstract forms, using the construct {'$form', F}, where F is bound to a parsed form in the scope of the call to gen_function/2.

Example:


  gen(Name, F) ->
     codegen:gen_function(Name, fun(X) -> X =:= {'$form',F} end).

After transformation, calling gen(is_foo, {atom,0,foo}) will yield the abstract form corresponding to:


  is_foo(X) ->
     X =:= foo.