pynash wordmark

The idea generator

Welcome to the PyNash group blog! This is a little contest by the Nashville Python user group. We’ll be posting new content daily, Tuesday through Friday, until one of us breaks. I’m @jorendorff and I’ll be your host on Tuesdays.

Topics will vary. You’ll see bits on the standard library, third-party packages and frameworks, core language features, clever hacks, algorithms, style, and so on. The only rule is that every post will have at least 100 words of text and 4 lines of Python code.

Let’s get to it.

Random ideas

I wrote a talk submission form for a local event recently. It unhelpfully suggested a random talk idea, just for laughs:

Here’s the Python code that generates those random ideas. You can take it for a spin pretty easily:

$ wget
2013-01-28 15:46:29 (15.4 MB/s) - “” saved [1260]
$ python
>>> import ideagen
>>> ideagen.random_idea()
'Audio and Autotune'
>>> ideagen.random_idea()
'How to use git to make an amazing mess'
>>> ideagen.random_idea()
'Bootstrap in 140 characters'

Let’s see how it works inside. As usual, if you understand the data, the code is obvious.

The data

First, there’s a constant global dictionary. This dictionary completely describes the form of every talk topic that the idea generator can generate. (This sort of thing is called a grammar.)

You can of course see the full thing, but here I’ll trim it down a bit:

productions = {
    'tech': [
    'talk': [
        '${tech} for ${person}',
        '${tech} + ${tech} = awesome',
        '${tech} with ${other}',

How will our program use this data? It’ll start by picking a random entry from the 'talk' list. Suppose it picks '${tech} with ${other}'. Next, it will replace '${tech}' with an item from the 'tech' list, and ‘${other}' with something from the 'other' list. The result is something like 'HTML5 with bacon'.

The code

def randomly_generated(nt):
    template = random.choice(productions[nt])
    def replace(match):
        return randomly_generated(
    return re.sub(r'\$\{(\w+)\}', replace, template)

def random_idea():
    return randomly_generated('talk')

Short and sweet.

The recursion means that any output string in productions may contain ${variables}, and the function will keep going until they are all fully expanded into random text.

I like this hack because it’s so tiny and because it’s a nice illustration of what recursion is good for.

With the grammar above, we can only generate a few hundred different talks in all. But the code is flexible. We could easily add grammar entries to generate millions of topics.

Next week I think I’ll talk about proxy objects. See you then.