cs50/problems/2022/python/um: check50 accepts substring-based solution that counts words like "umbrella" as "um"

[kali-mx]

Summary

Raising this to help improve the autograder tests for future students.

For cs50/problems/2022/python/um, check50 currently accepts a count implementation that treats "um" as a prefix/substring rather than a whole word. As a result, count("umbrella") returns 1, but check50 cs50/problems/2022/python/um still passes.

This appears to contradict the spec, which states that "um" should be counted only "as a word unto itself, not as a substring of some other word."

---

Minimal example

Here is a (deliberately incorrect) implementation of count:

def count(s):
    tally = 0
    pieces = s.split(" ")
    for item in pieces:
        if item.lower().startswith("umm"):
            continue
        if item.lower().startswith("um"):
            tally += 1
    return tally

With this implementation:

>>> count("um")
1
>>> count("Um, thanks, um...")
2
>>> count("umbrella")
1        # <-- incorrect per spec, but not caught by check50

Despite this clearly incorrect behavior for "umbrella" (and other words like "umlaut" or "umami"), check50 cs50/problems/2022/python/um reports all tests passing.

Expected vs actual behavior

Expected: check50 should fail implementations that count "um" inside larger words, per the problem specification.

Actual: A substring-/prefix-based implementation like the one above passes all check50 tests.