olegshulyakov/doocs-leetcode-solutions

Doocs LeetCode Solutions

LeetCode problems with solutions in many programming languages, built from the Doocs LeetCode repository. Each solution comes with the approach name, the reasoning that leads to it, and an explanation with complexity analysis. The dataset is meant for fine-tuning and evaluating code generation models.

The dataset is regenerated monthly from the latest Doocs commit by the leetcode-dataset-generator tool.

Structure

The dataset has two configs:

• default: one row per solution file (a problem in one language for one approach).
• problems: one row per problem; the solution fields below are nested in a solutions list.

Data Fields

Field	Type	Description	Example
id	int64	Problem number	1
title	string	Problem title	"Two Sum"
slug	string	Problem slug	"two-sum"
url	string	Problem URL on leetcode.com	"https://leetcode.com/problems/two-sum"
difficulty	string	Difficulty level	"Easy", "Medium", "Hard"
rating	int64	Contest rating; null when the problem has none	1249
source	string	Contest source; may be empty	"Weekly Contest 379 Q1"
description	string	Problem description as Markdown	"Given an array of integers…"
tags	list<string>	Problem tags	["Array", "Hash Table"]
language	string	Programming language of the solution	"Python", "Java", "C++"
approach	int64	Approach number within the problem	1, 2
approach_name	string	Approach name; may be empty	"Hash Table"
thinking	string	Reasoning that leads to the approach; may be empty	"Checking every pair is…"
explanation	string	Approach explanation with complexity analysis; may be empty	"We can use a hash table…"
solution	string	Complete solution code	"class Solution:\n def…"

Languages: Bash, C, C#, C++, Cangjie, Dart, Go, Java, JavaScript, Kotlin, Nim, PHP, Python, Ruby, Rust, Scala, SQL, Swift, TypeScript.

Data Splits

Split	Share of problems
train	~95%
test	~5%

Problems are assigned to a split by a hash of their id, so all solutions of a problem stay in the same split and existing problems keep their split as new ones are added.

How to Use

from datasets import load_dataset

dataset = load_dataset("olegshulyakov/doocs-leetcode-solutions", split="train")
# One row per problem with nested solutions:
# problems = load_dataset("olegshulyakov/doocs-leetcode-solutions", "problems", split="train")

sample = dataset[0]
print(f"Problem {sample['id']}: {sample['title']} ({sample['difficulty']})")
print(f"Approach {sample['approach']}: {sample['approach_name']}")
print(sample["thinking"])
print(sample["solution"])

Limitations

• Solutions are community-written and may not be optimal.
• Descriptions are converted from HTML to Markdown. Superscripts and subscripts are written as ^x and _x (e.g. 10^4), parenthesized when complex (e.g. 2^(n-1)).
• Alternative approaches (approach > 1) often have only thinking and no explanation.
• Coverage is limited to problems that have solutions in the Doocs repository.

License and Attribution

This dataset is derived from doocs/leetcode by the Doocs community and is distributed under the same CC-BY-SA-4.0 license. If you share or adapt it, you must give appropriate credit and distribute your contributions under the same license.

Problem statements are the property of LeetCode. This dataset is not affiliated with or endorsed by LeetCode or Doocs.

Contact

For questions or issues, please open an issue on GitHub.