Why is the OpenDylan IDE only on Windows?

We're often asked when the IDE will show up on Linux or Mac OS X.

Unfortunately, the answer to this is rather complex because the IDE consists of multiple pieces, all of which have separate portability issues.


The Existing IDE

DUIM - The GUI Framework

The biggest reason that the IDE is only on Windows is that it uses the DUIM framework which hasn't yet been ported away from Windows in a functioning and well-maintained manner. (An experimental GTK+ port has existed in the past, but it doesn't currently compile.)

Code Browsers and Inspectors

The code browsers and inspectors pull their information from frameworks that are independent of the GUI. It is possible to use this information separately from the IDE itself. The DIME environment for emacs supplies some of this functionality. The Hula project (now defunct) provided some of this functionality as well via a web-based interface.

This is actually an exciting set of features that our compiler provides. While many other languages have to glue IDE functionality on as a separate set of libraries, Open Dylan's compiler supports it internally from the ground up.

Debugging

The debugger depends on a couple of key components:

  • A debugger nub that provides the actual debugging functionality.
  • The debug information generated during compilation.

A new debugger nub would need to be written for each supported platform or an interface would need to be built that allowed the use of lldb or GDB's machine interface (gdb-mi).

The bigger issue however is the Dylan-level debug information. On Mac OS X and on 64 bit platforms, we use a C compiler backend, so there is no Dylan-level debug information at all. On 32 bit Linux and FreeBSD where we use HARP, we are also missing debug info due to how we generate the assembly (among other reasons).

Combined, this makes it difficult for the debugger to work away from Windows.

REPL / Listener

In the OpenDylan IDE, the REPL or listener was built on top of the debugger interfaces. Without the debugger, that code won't work.


Future Directions

We aren't satisfied with the status quo and would love to see it change.

WINE

Some components of the Windows IDE actually work under WINE (1.5 and later). However, this tends to crash a lot and the debugger doesn't work as WINE doesn't implement that aspect of the Windows APIs.

LLVM

The upcoming LLVM compiler backend features much improved support for debug information generation. This, combined with lldb, should result in debugging being possible. It is likely that some work will need to be done for this to work well.

DIME / Emacs

There are many improvements to DIME that could be made.

Web-based Tools

Much like Hula demonstrated, we can build a new web-based tool for doing code browsing, inspecting libraries and other traditional IDE functions. We can even use it to integrate documentation.

Porting the existing IDE

Porting the existing IDE would mean getting DUIM up and running on other platforms. This is doable but would require a substantial amount of work. DUIM has also aged and hasn't kept up with all the things that modern toolkits can do.

On top of that, it lacks some features like syntax highlighting that might be easier to provide in other ways.

A new REPL

We're investigating some options for a new REPL. These include using the LLVM backend and using a JIT, using the existing backend and loading shared libraries, and the more appealing option of using DynASM from LuaJIT.

Sublime Text

Sublime Text is a popular editor these days and other languages have built IDE-like extensions for it. We could do the same for Dylan as well, perhaps sharing some of the backend code with DIME.

Other Ideas

It is likely that other people will find new and interesting ways to apply our compiler frameworks and we'd enjoy seeing what other people are interested in trying out.

Projects like Light Table, the iPython Notebook and others show that there is a lot of fertile ground for exploration and experimentation.


Moving Forward

We'd love to hear what people think should happen next.

In the short term, we're working on the LLVM backend so that we can gain debugging support on more platforms (among other benefits from the LLVM backend).

It would be wonderful to see some volunteers to help out with some of these projects. There's a lot of exciting and potentially groundbreaking work that can be done and we're more than happy to help provide guidance and direct assistance.

We'll be covering some of these items in more depth soon! Be sure to follow @DylanFoundry or our RSS feed.


Comments