HACKER Q&A
📣 techn00

Guidance on writing a source to source compiler (transpiler)


I would like to learn to write a transpiler from scratch. Where would I get started?

Preferably something that doesn't involve going though big projects to read the source code.

  👤 samsaga2 Accepted Answer ✓
If the languages are similar:

  1. Tokenize
  2. Build AST tree
  3. Emit target code from the AST tree
If they are very different:

  1. Tokenize
  2. Build AST tree
  3. Create a simple intermediate language from the AST tree
  4. Emit target code from the intermediate language
It is the same as building a whole compiler, but without the optimization/register allocator passes.
👤 stevekemp
You could start here:

https://github.com/jamiebuilds/the-super-tiny-compiler

That converts from lisp-like to javascript. Really though this is a big field, and there are lots of resources out there.

To get started look at your input language; you'll need to lex and parse that. Then massage the parsed structure into the appropriate output.

You can see me convert brainfuck to C, or x86 assembly language here:

https://github.com/skx/bfcc

👤 quectophoton
If you don't mind a video, Destroy All Software has a nice screencast on the topic:

https://www.destroyallsoftware.com/screencasts/catalog/a-com...

Really easy to understand, even if Ruby is not your main language. The video is about compiling from a tiny Ruby-like language[1] into real JavaScript.

[1]: It's only a single, simple function, but the concepts are well explained and it should be easy to build on top of that.

👤 cdhdc
http://www.txl.ca/txl-index.html
👤 ricklamers
A good option would be to take a MooC on compilers https://www.edx.org/course/compilers
👤 kunley
Doesn't matter much if the output is also a source, you can't avoid lexing and parsing, so the best thing to get started seems to be... the Dragon Book, right?