I'm noticing this oddity when I create a Datasource and iterate through the features.

Let's say feature FA in my GeoJSON has one attribute: foo: 1. And FB has one attribute: bar: 2. When I create a Datasource and check these features within that structure, FA might have two attributes now: foo: 1, bar: null; and same for FB: bar: 2, foo: null.

So it might seem that Mapnik is giving each feature all the attributes available but with null as the value if the feature did not actually have that feature in the original data.

But this is not happening consistently. In some sample data I've been using, I've found that Mapnik adds the null attributes only sometimes. Whether or not this happens does not seem connected with data type, or the number of times the attribute shows up.

I'm not sure what to make of this, or whether I should file this upstream in Mapnik itself.