Anatomy of a problem

What lcpy gen creates, what each file is for, and which one you own.

Every generated problem is the same six files. Learn the shape once and every problem in the catalog feels like home.

lcpy gen -n 1

leetcode/two_sum/
├── README.md           # problem statement, difficulty, topics, tags
├── solution.py         # your code; the only file you edit
├── test_solution.py    # 10+ parametrized cases, ready to run
├── helpers.py          # run_/assert_ pair for readable test output
├── playground.py       # percent-format scratchpad notebook
└── __init__.py

​

README.md

The problem statement: difficulty, topics, collection tags, a link to the original LeetCode problem, the description, and worked examples. Everything you need to solve without leaving the terminal.

​

solution.py: the only file you edit

Generated with a class skeleton, the method signature with type hints, and a TODO where your implementation goes. Solved examples in this repo also carry Time/Space complexity comments at the top of the method, the house convention.

class Solution:
    # Time: O(n)
    # Space: O(n)
    def two_sum(self, nums: list[int], target: int) -> list[int]:
        ...

​

test_solution.py

A parametrized pytest suite with 10+ cases per problem, edge cases included. It runs red the moment you generate the problem and stays the contract: make red go green, never edit the tests. See Testing.

​

helpers.py

Two functions per problem: run_<name> instantiates your solution class and calls the method; assert_<name> compares result to expected, normalized so order-insensitive answers compare cleanly. Failures print the inputs and expected output instead of a raw object dump.

def run_two_sum(solution_class: type, nums: list[int], target: int):
    implementation = solution_class()
    return implementation.two_sum(nums, target)


def assert_two_sum(result: list[int], expected: list[int]) -> bool:
    assert sorted(result) == sorted(expected)
    return True

​

playground.py

A scratchpad in jupytext percent format: plain Python in git, a notebook when you want one. See Notebooks.

​

Generated, not hand-rolled

Every problem directory comes from a JSON template, so files arrive byte-identical on every machine and CI verifies they regenerate exactly. If something looks off, regenerate the problem with bake p-gen rather than editing generated files by hand.