Python has great support for mapping over lists or tuples, creating new structures containing the results of calling a function on the members of the original. I was looking for something similar for the values in a dictionary, maintaining the original keys, and I didn’t find it described anywhere.

I’d written a little webapp using Flask and one of the routes took a number of float values in the URL. To make sure that these were indeed the floats I was looking for it seemed prudent to pass them into the float type and catch any ValueError that was raised (Flask does have a handler for floats in the routing it inherits from Werkzeug, werkzeug.routing.FloatConverter, but “This converter does not support negative values”). For the first pass at this I listed all the parameters to the function and then checked each one explicitly like this:

@app.route("/route-with-floats/<first>/<second>/<third>/<fourth>")
def route_with_floats(first, second, third, fourth):
    try:
        first = float(first)
        second = float(second)
        third = float(third)
        fourth = float(fourth)
    except ValueError:
        abort(404)

That worked fine but I wanted to find a way to condense it down a bit and it occurred that maybe I could use the **kwargs dictionary; I’d just need to map over that and pass each value into float. For a list (or set or tuple etc.) the map function or list comprehensions would be a perfect fit and second nature to Python developers, but I couldn’t recall ever doing anything similar for a dict.

The solution I came up with passes a generator expression which iterates over the items in a dictionary (each item is a tuple of the key and the value) into the dict callable. For an arbitrary function, func, and a dictionary, d, the solution looks like this:

dict((v[0], func(v[1])) for v in d.items())

Plugging that into the Flask example above it now looks like this:

@app.route("/route-with-floats/<first>/<second>/<third>/<fourth>")
def route_with_floats(**kwargs):
    try:
        params = dict(
            (v[0], float(v[1])) for v in kwargs.items()
        )
    except ValueError:
        abort(404)

Arguably this is less readable and less explicit than the original form but I like the fact that it’ll be the same regardless of the number of parameters i.e. values in the dictionary.

This should actually be described as a dictionary comprehension, a feature which is built into Python 3, and in fact the exact technique shown above is described in PEP 274. It’s good to know that someone cleverer than me has already had the same idea.