Sprint Report

I worked on a source-to-source translation system at the Google Python sprint. The system is intended to allow us to write transformations that upgrade programs from Python 2.x to Python 3.x. I finished a toy example that demonstrated that the approach will work. I translated a small program that use has_key to an equivalent program that uses in.

d = dict(x=1)
if d.has_key("x"):
print "yes"
else:
print "no"

Is transformed to

d = dict(x=1)
if "x" in d             :
print "yes"
else:
print "no"

(Yes. There are 13 spaces between the d and the :.) The code is checked into the Python sandbox under refactor. I had a bunch of helpful conversations and did a little pair programming with Martin von Loewis.

Guido wrote a summary post about the Python 3000 sprint work. Update: I fleshed out many of the details in several subsequent posts.

• Python 3000 Translation Strategy describes how to decorate an AST with concrete tokens.
• Python tokenize module explains some details of the tokenize and untokenize functions I'm using.
• An example code transformation describes how the simple has_key() to in transformer works.