Source-to-source compiler

A source-to-source compiler, transcompiler, or transpiler is a type of compiler that takes the source code of a programming language as its input and outputs the source code into another programming language. A source-to-source compiler translates between programming languages that operate at approximately the same level of abstraction, while a traditional compiler translates from a higher level programming language to a lower level programming language. For example, a source-to-source compiler may perform a translation of a program from Pascal to C. An automatic parallelizing compiler will frequently take in a high level language program as an input and then transform the code and annotate it with parallel code annotations (e.g., OpenMP) or language constructs (e.g. Fortran's forall statements).^[1]

Another purpose of source-to-source-compiling is translating legacy code to use the next version of the underlying programming language or an API that breaks backward compatibility. It will perform automatic code refactoring which is useful when the programs to refactor are outside the control of the original implementer (for example, converting programs from Python 2 to Python 3, or converting programs from an old API to the new API) or when the size of the program makes it impractical or time consuming to refactor it by hand.

Transcompilers may either keep translated code as close to the source code as possible to ease development and debugging of the original source code, or else they may change the structure of the original code so much, that the translated code does not look like the source code.^[2] There are also debugging utilities that map the transpiled source code back to the original code; for example, JavaScript source maps allow mapping of the JavaScript code executed by a web browser back to the original source in a transpiled-to-JavaScript language.^[3]

History

One of the earliest programs of this kind was Digital Research's XLT86 in 1981, a program written by Gary Kildall, which translated .ASM source code for the Intel 8080 processor into .A86 source code for the Intel 8086. Using global data flow analysis on 8080 register usage, the translator would also optimize the output for code size and take care of calling conventions, so that CP/M-80 and MP/M-80 programs could be ported to the CP/M-86 and MP/M-86 platforms automatically. XLT86 itself was written in PL/I-80 and was available for CP/M-80 platforms as well as for DEC VMS (for VAX 11/750 or 11/780).^[4]

A similar, but much less sophisticated program was TRANS.COM, written by Tim Paterson in 1980 as part of 86-DOS. It could translate some Z80 assembly source code into .ASM source code for the 8086, but supported only a subset of opcodes, registers and modes, often still requiring significant manual correction and rework afterwards. Also it did not carry out any register and jump optimizations.^[5][6]

Programming language implementation

The first implementations of some programming languages started as transcompilers, and the default implementation for some of those languages are still transcompilers. In addition to the table below, a CoffeeScript maintainer provides a list of languages that compile to JavaScript.^[7]

Porting a codebase

When developers want to switch to a different language while retaining most of an existing codebase, it might be better to use a transcompiler compared to rewriting the whole software by hand. In this case, the code often needs manual correction because the automated translation might not work in all cases.

Examples

ROSE

The ROSE compiler framework, developed at Lawrence Livermore National Laboratory (LLNL), is an open-source compiler infrastructure that generates source-to-source analyzers and translators for multiple source languages, including C, C++ and Fortran. It also supports OpenMP, UPC and certain binary files, while also supporting auto-parallelizing compilers by generating source code annotated with OpenMP directives. Unlike most other research compilers, ROSE is aimed at enabling non-experts to leverage compiler technologies for building their own custom software analyzers and optimizers.

DMS Software Reengineering Toolkit

DMS Software Reengineering Toolkit is a source-to-source program transformation tool, parameterized by explicit source and target (may be the same) computer language definitions. It can be used for translating from one computer language to another, for compiling domain-specific languages to a general purpose language, or for carrying out optimizations or massive modifications within a specific language. DMS has a library of language definitions for most widely used computer languages (including full C++, and a means for defining other languages which it does not presently know).

LLVM

LLVM can translate from any language supported by gcc 4.2.1 (Ada, C, C++, Fortran, Java, Objective-C, or Objective-C++) or by clang to any of: C, C++, or MSIL by way of the "arch" command in llvm-gcc.

% llvm-g++ -emit-llvm x.cpp -o program.bc -c
% llc -march=c program.bc -o x.c
% cc x.c -lstdc++
 
% llvm-g++ x.cpp -o program.bc -c
% llc -march=msil program.bc -o program.msil

Translation to C has been removed from LLVM since version 3.1. It had numerous problems, to the point of not being able to compile any nontrivial program.^[26]

Emscripten

Emscripten is a source-to-source compiler that transforms LLVM bytecode, typically created by compiling from C or C++, to JavaScript. As a result, native applications can be converted into JavaScript applications that can run in web browsers.

An example of using the Emscripten C compiler:

% emcc helloworld.c -o helloworld.html

An example of using a make file with Emscripten:

% emmake make

An example of using a configure script with emscripten:

% emconfigure ./configure

Emscripten is very powerful and is remarkably able to compile most large applications that are system-independent with almost no modifications to the source code. Some examples include Unreal Engine 3,^[27] a fork of Cube Engine 2 known as BannanaBread,^[28] and ammo.js, which is an exact JavaScript port of the Bullet physics engine compiled using Emscripten.^[29]

Other examples

• Haxe is an open-source high-level multiplatform programming language and compiler designed to support multiple targets, using a source language similar to the ActionScript language used by Adobe Flash. Depending on the selected target, Haxe may act as a transpiler or compiler.
• Dart is an open-source web programming language that transpiles to JavaScript, developed by Google.
• TypeScript is a programming language designed as a strict superset of JavaScript that transpiles to JavaScript. It is developed by Microsoft.

Refactoring tools

The refactoring tools automate transforming source code into another:

• Python's 2to3 tool transforms non-forward-compatible Python 2 code into Python 3 code.
• Qt's qt3to4 tool convert non–forward-compatible usage of the Qt3 API into Qt4 API usage.
• Coccinelle uses semantic patches to describe refactoring to apply to C code. It's been applied successfully to refactor the drivers of the Linux kernel due to kernel API changes.^[30]
• pfff can do the same things as Coccinelle, but targets more languages. It can also features a bug finder, offers structural searches and source code visualization. There is good support for C, Java, PHP, JavaScript, HTML, CSS, and preliminary support for a lot more languages^[31]
• RefactoringNG is a NetBeans module for refactoring Java code where you can write transformations rules of a program's abstract syntax tree.

See also

References