Thursday, May 15, 2008

cmd2 0.3.0

I'm speaking on SQL*Plus alternatives again on Wednesday at GCOUG, and trying to polish up sqlpython a bit in preparation for that. That required polishing cmd2; I've been increasingly embarrassed that cmd2 used homemade code (flagReader) to interpret --options --like --these, instead of using a perfectly good module from the standard library (optparse).

Optparse is not designed for iterative cmd-type apps where you remain at an application prompt after issuing each command - it keeps issuing sys.exit()s to kick you back to the shell. That was my excuse for not using it earlier, but once I decided to fix it, a little subclassing took care of it.

Next I got a little deeper into decorators than I'd ever been, to provide a concise way to specify those options:

@options([make_option('-k', '--ketchup', action="store_true", help="Ketchup on that?"),
[make_option('-r', '--relish', help="relish (sweet or dill)")])
def do_hotdog(self, arg, opts):
if opts.relish:
self.stdout.write('One dog with %s relish\n' % opts.relish)
if opts.ketchup:
self.stdout.write('Hope you're not wearing white!\n')

That was pretty fun. Took some experimentation and this article, though, to figure out how to write a decorator that would accept an argument.

I didn't throw away the homemade ("flagReader") code, for backward compatibility; I issue a DeprecationWarning instead. I did throw away the guts and reimplement it as a wrapper around optparse. I managed to do that part on a single 20-minute bus ride, thanks to good doctests on flagReader. I feel so grown-up!

Finally, inspired by Brandon Rhodes' article in the March 2008 Python magazine, I tried supporting buildout deployments this time. I'm not sure I'm getting it quite right yet; expect it to actually work in cmd2 0.3.1.

1 comment:

Carl said...

I've always been interested in the cmd module but its lack of advanced features kept me from bothering. cmd2 looks great - can't wait to play with it.