Sunday, June 14, 2009

Scons and noweb

I was curious to see how I could integrate noweb and Scons.
You can download my little Sconstruct for this, Sconstruct.example. It defines two builders. NoWeave is used to produce TeX or LaTeX documents, while NoTangle extracts the non-document artefacts, i.e. programs, config files, scripts ... It also includes productions for generating a sample program about Ackermann function:

ackdoc = env.NoWeave('ack.tex', 'ack.nw')
ackcode = env.NoTangle('ack.py', 'ack.nw')
acktest = env.NoTangle('ackTest.py', 'ack.nw')

The noweb source is ack.nw and the companion BibTeX file is ack.bib

It contains the doc chunks describing the function, the source code chunck for the relevant Python code and the code chunk for the unit test

You can try it out executing

$ scons -f Sconstruct.example
You will get the following artefacts, ack.py, ack.tex, ackTest.py and ack.pdf

Remember to run BibTeX first...

Tuesday, February 24, 2009

LaTeX, Python and Literate Programming

In my spare time (a couple of hours per weekend...!) I am implementing calcal, a Python version of calendrica-3.0.cl, the Common Lisp implementation of the calendars from N. Dershowitz, E. M. Reingold Calendrical Calculations, 3rd Edition. (In case you are interested you can find a preview in my google page.)
At some point I decided to go Literate [Programming] using noweb. This is an experiment in the experiment but so far it has been a good choice because I can define all I need in the same place and generate documentation, source code (Python, shell scripts ...) from the same source.
I also found something interesting (on a now disappeared blog ttp://usefreetools.blogspot.com): executing Python from within LaTex! I could use it to avoid to hardcode results in my doc and just calculate them on the fly...
The same blog was showing how to build LaTeX docs using SCons: I will defenitly use it; my Makefile isn't that great nor easy to mantain.

Monday, February 23, 2009

My Giants

I am probably too selective, anyway my models for computer science/software engineering (one of them would disagree on both definitions) are just two: Donald E. Knuth and Alan Kay.
The first one continues to surprise me with the depth, clarity and joy of his works: from TeX (well, I use LaTex but it does not exist without TeX) to The Art of Computer Programming to Literate Programming.
About the latter, I was one of the blessed to be present to his Turing Award Lecture: he shocked me to the point I had two sleepless nights so angry I was about having wasted so much time in useless (computer) matters! After that I have been studying and using a lot of what he wrote and presented from Squeak to Croquet to the Burroughs B5000 and stack computers to his recent line of exploration and the proposal to NSF about reinventing programming (PDF).