What is the most Pythonic way to count word frequencies in a string, and what does Counter return for missing keys?

How to think about it

This is the classic frequency count, and the interesting part isn’t getting the numbers — it’s knowing the whole Counter API: the zero default for missing keys, top-k with a heap, counter arithmetic, and the dict.get fallback for when you want to show what’s happening underneath.

Counter takes any iterable and hands back a dict-like object of counts. The one behaviour that sets it apart from a plain dict: a missing key returns 0, not a KeyError, which keeps downstream arithmetic safe without guard clauses.

A worked example

from collections import Counter

text = "the cat sat on the mat the cat ate the rat"
words = text.split()

freq = Counter(words)
print("Full counter:", freq)
print()

print("Top 3:", freq.most_common(3))          # heap-based, O(n log k)
print("Count of 'dog':", freq["dog"])          # 0, not KeyError
print()

# Case-insensitive: normalise as you feed the Counter
text2 = "Python python PYTHON is great"
print("Case-insensitive:", Counter(w.lower() for w in text2.split()))
print()

# Counters support arithmetic — merge two with +
print("After merging:", (freq + Counter(["cat", "dog", "cat"])).most_common(4))
print()

# The manual fallback, if collections isn't available
manual = {}
for w in words:
    manual[w] = manual.get(w, 0) + 1           # the zero default, by hand
print("Manual top 3:", dict(sorted(manual.items(), key=lambda x: -x[1])[:3]))

Full counter: Counter({'the': 4, 'cat': 2, 'sat': 1, 'on': 1, 'mat': 1, 'ate': 1, 'rat': 1})

Top 3: [('the', 4), ('cat', 2), ('sat', 1)]
Count of 'dog': 0

Case-insensitive: Counter({'python': 3, 'is': 1, 'great': 1})

After merging: [('the', 4), ('cat', 4), ('sat', 1), ('on', 1)]

Manual top 3: {'the': 4, 'cat': 2, 'sat': 1}

Three things to lift out. most_common(3) returns the top entries already ranked — and uses a heap, so it’s cheaper than sorting the whole thing. freq["dog"] returned 0 instead of raising, which is what makes Counter safe to do arithmetic on. And freq + extra merged the two by adding counts key by key (cat went 2 → 4), turning a fiddly loop into a single operator.

Why Counter beats a manual loop

The hand-rolled dict.get(w, 0) + 1 version is the same O(n) and clarifies what’s happening underneath — worth showing if the interviewer asks “without collections?” But in real code there’s no reason to write it: Counter gives you most_common, the arithmetic operators (+, -, &, |), and the zero-default for free.