Key Error in Python: Missing Keys vs Present None Values

Start with a small, explicit task

Inspect a fictional location record whose city field is present but unknown and whose postal field is absent. Check membership, direct lookup and get with an explicit sentinel. Keep required-field handling separate from an optional-value display rule.

Run the complete example

This original example was checked with Python 3.12.14. Save it as python_key_error.py and run it with Python 3.12. It uses only the standard library. The assertions check the stated results.

record = {"location": {"city": None}}
location = record["location"]
missing = object()
assert "city" in location and location["city"] is None
assert "postal" not in location
assert location.get("city", missing) is None
assert location.get("postal", missing) is missing
try:
    location["postal"]
except KeyError as exc:
    absent_key = exc.args[0]
else:
    raise AssertionError("expected missing key")
assert absent_key == "postal"
print("city present:", "city" in location, "value:", location["city"])
print("missing key:", absent_key)

Explain the result

Direct city lookup succeeds even though its value is None. The postal lookup fails because no such key exists. A plain get without an explicit default would return None for the absent key too, making the two situations look alike. The unique object sentinel preserves the distinction using identity checks.

The targeted except records the missing key from the error’s arguments. That is useful when explaining the failing field without turning every exception into a default record. In a deeper structure, an absent location key would fail at an earlier lookup; a location value of None would raise a different type-related error when used like a dictionary.

Avoid these mistakes

Use defaults only when they represent an actual allowed absence, not to conceal a broken required-data contract. Avoid checking only truthiness: an empty string or numeric zero can be present, legitimate values. Keep error messages free of unnecessary private record contents when adapting this pattern.