git @ Cat's Eye Technologies Wierd / master

Tree @master (Download .tar.gz)

The Wierd Programming Language

This is Cat's Eye Technologies' distribution of Wierd, a two-dimensional esoteric programming language (a "fungeoid".) Unlike similar languages where the symbols in a program determine which instructions are executed, in Wierd, it is the bends in the chain of arbitrary symbols that determine which instructions are executed.


Wierd has a long and colourful history of language fuzziness.

Wierd was originally designed by Ben Olmstead, John Colagioia, and Chris Pressey, in a three-way email discussion about how Befunge and brainfuck could be combined in an interesting way.

Based on this discussion, John wrote an interpreter in C for his interpretation of Wierd (which, as comments in its source code explain, differs from "official" Wierd), and wrote a version of the classic "Hello, world!" program which runs on his interpreter.

Later on, based on his implementation, John wrote a specification for the language accepted by his interpreter, calling it Wierd.

Even later on, Milo van Handel (who was not privy to the email conversation) wrote an interpreter, also in C, for his interpretation of Wierd, which was apparenly based largely on John's interpreter and spec, but interpreting some conditions slightly differently, and filling in some gaps (such as treatment of EOF.) Milo also wrote several example programs that run on his interpreter, including several versions of the classic cat program and a Wierd/INTERCAL polyglot.

Unfortunately, the language implemented by Milo's interpreter has different semantics from the language implemented by John's interpreter — and from what I've been able to tell, the two languages are largely incompatible. It may of course be possible to write polyglot programs which are accepted by both interpreters, perhaps even having the same behaviour in both, but I'm pretty sure that none of the included example programs fall into this category.

Later still, Chris attempted to implement John's interpretation of Wierd in the yoob framework, based on John's spec, but when trying to run hello.w on it, found that it would only get so far before entering an infinite loop (back and forth along the chain.) This suggested a possible bug in either the yoob implementation or in wierd.c or in the spec.

Shortly after this, Chris also patched Milo's implementation to take standard long options, for portability (NetBSD doesn't have getopt_long_only.) (Sorry Milo, hope you don't mind.)

Shortly after that, Chris began writing an interpreter in Javascript, using the yoob.js framework, of John's interpretation of Wierd, and discovered the source of the problems with his previous attempt: hello.w relies on incorrectly-documented behaviour. Specifically, while John's spec and the comments in John's interpreter say that during the "putget" operation, a zero value means "get" and a non-zero value means "put", in the implementation, it is actually the other way around.

Given all this...

Given all this, well, here's how I see it.

The name Wierd refers to the language defined (however fuzzily) by that original email conversation.

Given that that email thread is, as far as I know, lost and gone forever, Wierd has no specification, and no reference implementation. (Therefore, there are no src, doc, or eg directories in the root directory of this repository.)

Both John Colagioia and Milo van Handel designed and implemented dialects of Wierd. (Therefore there is a directory called dialect in this repository.) I have tended to call them "John's Wierd" and "Milo's Wierd" in the past, but anything else that distinguishes them by the name of their author would suffice. (Therefore there are subdirectories dialect/wierd-jnc and dialect/wierd-mvh in this repository, and each of those contains the standard src, doc, and eg subdirectories. And in addition, because Chris's interpreter implements John's Wierd, it is in the impl directory of dialect/wierd-jnc. It is also installed online at

And in light of all this, it might also be acceptable to consider Wierd to be a language family rather than a language. I'm not yet decided on this point.

Pull Requests

You are perfectly welcome to open pull requests on this repository, but please observe the layout described above:

  • implementations of John's Wierd go into dialect/wierd-jnc/impl
  • example programs in John's Wierd go into dialect/wierd-jnc/eg
  • implementations of Milo's Wierd go into dialect/wierd-mvh/impl
  • example programs in Milo's Wierd go into dialect/wierd-mvh/eg
  • any other dialects of Wierd go into dialect/your-dialect-name

In light of the following section, I would also ask that you provide some license information regarding any sources you submit. Open-source licensing would definitely be preferable.


The Wierd distribution's licensing matches the language's fuzziness.

With the exception of Milo's quine.w, which is licensed under the GPL (no version specified), no license was ever explicitly placed on any of John's or Milo's sources or documentation, so they are all implicitly copyrighted by their respective authors.

However, Cat's Eye Technologies has been redistributing these sources in the form of this Wierd distribution for years now, with no objections from the authors, so I think it's safe to consider them to be freely redistributable, unmodified and for non-commercial purposes; however, I am not a lawyer, your mileage may vary, caveat emptor, etc. etc.

In stark (I hope) contrast to this, Chris's implementation, wierd-jnc.js, is placed into the public domain (see the file UNLICENSE in its directory.)