Overview

An off-curriculum Python/TDD session. James is experienced in TypeScript/JavaScript but new to Python. We set up a uv project, wrote pytest tests for two data structures, and James implemented both. Code reviews after each implementation surfaced Python idioms and real bugs.

Questions

Project Setup with uv and pytest

James set up a minimal Python project using uv init --no-package and uv add --dev pytest, then ran tests with uv run pytest -v.

Delivery: N/A (hands-on setup) Content: Clean — correct commands, correct mental model of what --no-package does.

HashSet Implementation

James implemented a HashSet backed by an array of buckets with chaining. All tests passed.

Delivery: N/A Content grade: A−

Correct use of __contains__, __len__, and exception translation (ValueError → KeyError in remove). Two small feedback points:

• sum([len(x) for x in ...]) — wrapping a list comprehension in sum() builds an unnecessary intermediate list; generator expression sum(len(x) for x in ...) is idiomatic
• for i in range(...) with unused i — Python convention is _ for unused loop variables

Python Type-Agnostic Hashing

James asked how to detect whether an item is a number or string, and how to convert a string to a number (referencing the TypeScript isNaN pattern). Key insight:

Python’s hash() is already polymorphic — hash(42), hash("hello"), hash((1,2)) all return integers. The likely issue was using item % self.capacity directly instead of hash(item) % self.capacity. Swapping that one call fixed the failing test.

Content grade: A — once the pattern was clear, James immediately identified the fix.

MinHeap Implementation

James implemented a MinHeap with push/pop/peek, _sift_up, _sift_down, _heapify, and a _assert_heap debugging method. Showed good instinct for private helpers and defensive assertions.

Delivery: N/A Content grade: B− — structure solid, but three real bugs found in review (see below).

MinHeap Code Review

Three bugs found:

1. _assert_heap off-by-one — if bi < length - 1 skipped the last valid parent-child pair. For a 2-element heap [5, 1], the assertion never fired. Fix: bi < length.

2. _sift_down equal-children case — when both children are equal and less than the parent (e.g. [5, 3, 3]), b < a and b < c fails because b == c. Neither branch fires; the heap property is violated silently. Fix: change b < c to b <= c in the first branch.

3. pop() is O(n) — list.pop(0) and list.insert(0, ...) both shift every element. Standard fix: swap root with last element, list.pop() (O(1)), then sift down.

Style issues also addressed:

• min shadows the built-in; use result or smallest
• initial if initial else [] should be list(initial) if initial is not None else [] to handle empty-list input and avoid mutating the caller’s list
• _sift_up was more complex than necessary; standard implementation only compares with parent, not siblings
• _heapify iteration used a confusing length - ai - 1 mapping; cleaner as range((length-1)//2, -1, -1)