Jon Aquino's Mental Garden

Engineering beautiful software jon aquino labs | personal blog

Saturday, April 03, 2010, or the world's first executable blog post

I have recently been interested in the old idea of literate programming. Basically, you have a document that describes in detail how a program works, and it has embedded chunks of code. It allows you to see the thoughts of the programmer as he explains how he writes the program using prose. A tool is provided that you can use to extract the working program from chunks of code in the document.

Here's the thing: what you are reading right now is a literate program.

Yes, you can copy this blog post into a file and feed it into the tool, and it will spit out a program. Q: Where do I get the tool? A: That's the program that this document spits out. This document will produce a script that you can use to extract code from noweb-format literate programs.

Why do we need to make a new tool if the noweb tool already exists? Because the noweb tool is hard to install. It's not super-hard, but most people don't want to spend time trying to compile it from source. There are Windows binaries but you have to get them from the Wayback Machine.

Anyway, the noweb tool doesn't seem to do very much, so why not write a little script to emulate it?

And that is what we will do now.


If you are just interested in the script produced by this document, you can download it from GitHub.


The end goal is to produce a Python script that will take a literate program as input (noweb format) and extract code from it as output. For example, -Rhello.php hello.noweb > hello.php

This will read in a file called hello.noweb and extract the code labelled "hello.php". We redirect the output into a hello.php file.


In a literate program, there are named chunks of code interspersed throughout the document. Take the chunk of code below. The name of it is "Reading in the file". The chunk ends with an @ sign.

Let's start by reading in the file given on the command line. We'll build up a map called "chunks", which will contain the chunk names and the lines of each chunk.

<<Reading in the file>>=
file = open(filename)
chunkName = None
chunks = {}
OPEN = "<<"
CLOSE = ">>"
for line in file:
match = re.match(OPEN + "([^>]+)" + CLOSE + "=", line)
if match:
chunkName =
chunks[chunkName] = []
match = re.match("@", line)
if match:
chunkName = None
elif chunkName:


Now that we have a map of chunk names to the lines of each chunk, we need to know which chunk name the user has asked to extract. In other words, we need to parse the command-line arguments given to the script: -Rhello.php hello.noweb

For simplicity, we'll assume that there are always two command-line arguments: in this example, "-Rhello.php" and "hello.noweb". So let's grab those.

<<Parsing the command-line arguments>>=
filename = sys.argv[-1]
outputChunkName = sys.argv[-2][2:]


So far, so good. Now we need a recursive function to expand any chunks found in the output chunk requested by the user. Take a deep breath.

<<Recursively expanding the output chunk>>=
def expand(chunkName, indent):
chunkLines = chunks[chunkName]
expandedChunkLines = []
for line in chunkLines:
match = re.match("(\s*)" + OPEN + "([^>]+)" + CLOSE + "\s*$", line)
if match:
expandedChunkLines.extend(expand(, indent +
expandedChunkLines.append(indent + line)
return expandedChunkLines


The last step is easy. We just call the recursive function and output the result.

<<Outputting the chunks>>=
for line in expand(outputChunkName, ""):
print line,

And we're done. We now have a tool to extract code from a literate programming document. Try it on this blog post!


To generate from this document, you first need a tool to extract the code from it. You can use the original noweb tool, but that's a bit cumbersome to install, so it's easier to use the Python script

Then you can generate from as follows: >


Here's how the pieces we have discussed fit together:

#! /usr/local/bin/python
# By Jonathan Aquino (
# This program extracts code from a literate programming document in "noweb" format.
# It was generated from, itself a literate programming document.
# For more information, including the original source code and documentation,
# see
import sys, re
<<Parsing the command-line arguments>>
<<Reading in the file>>
<<Recursively expanding the output chunk>>
<<Outputting the chunks>>


  • Haskell has some "built in" literate programming features. See

    I like the version with ">".

    By Blogger Thomas David Baker, at 4/03/2010 5:58 PM  

  • That's cool Tom. I guess one thing that it's missing is the ability to name chunks of code so they can be combined elsewhere. But it's definitely a step forward.

    By Blogger Jonathan, at 4/03/2010 6:18 PM  

  • Great post! I ported your tool to PHP:

    I expanded the noweb syntax a bit, you can write multiple code files from same source by just having filenames as chunk names.

    By Blogger Bergie, at 1/15/2011 5:57 AM  

  • Ah - very nice, Bergie!

    By Blogger Jonathan, at 7/21/2011 7:53 PM  

Post a Comment

<< Home