Safe Navigation in Python

Written by Ben Wendt

In a chain of method calls, what happens when one of the calls returns a null? The next chained call will throw some kind of reference error. But what if you don’t want to deal with a reference error? What if you want the null value to be the answer? That’s where a safe navigation operator is useful.

Ruby has a safe navigation operator &.. This is really useful in database applications. For example, consider a database model of a tree, so you might have something like:


But what if your tree doesn’t have that leaf, or that branch, or if it has no trunk? But also you don’t mind “null” being the answer to your query. Well that’s where safe navigation is useful. Observe:


This would return null if any of the calls in the chain return null, without having 9 levels of short-circuiting.

Note: I’m ignoring the law of demeter in this post.

But python does not have a safe navigation operator. One has been proposed, but currently it doesn’t exist.

it is however possible to roll one. Consider an input document like this:

<!doctype html>
            <h1>My article</h1>
                <p>Some text <b>yelling <i>curiously</i></b>.</p>
            <h1>My article</h1>
                <p>Some text <b>yelling <i>curiously</i></b>.</p>

Reading the article with beautiful soup, you can access various html nodes using a chained dot notaion:

import bs4
from typing import List, Optional

with open("something.html", "r") as f:
    soup = bs4.BeautifulSoup(, "html.parser")

except AttributeError:
    print("that did not work.")

Here we can read the contents of the b tag, but not the c tag, since the latter doesn’t exist. But if we have the function we talked about:

def safe_navigate(soup: Optional[bs4.BeautifulSoup], tag_list: List[str]):
    if tag_list:
        return safe_navigate(getattr(soup, tag_list[0], None), tag_list[1:])
        return soup

print(safe_navigate(soup, ["body", "article", "div", "p", "b", "i", "text"]))
print(safe_navigate(soup, ["body", "article", "div", "p", "c", "i", "text"]))

Now our document query works in both cases, the query with the b gives the expected result, and the query with c returns None.