My exercism rust track solutions

I use this repository also to experiment with AI/LLM prompts to solve the exercises.

Prompt within exercise root folder:

# Code Implementation

Implement all _instructions_ defined in `README.md` in the rust file:`src/lib.rs`.

## Enable all tests
Ensure all tests in the `tests` folder are NOT ignored and enabled. 

## Test the code

Run all tests in the `tests` folder, after the implementation of the all _instructions_ in `src/lib.rs` is done.


## Ensure all tests pass

Make sure all tests pass successfully. 
In case test(s) fail, adapt the code implementation and rerun all tests until all tests pass.

## Improve

Do better. Improve the code implementation and increase readability and reduce code complexity.

Claude Code

I used implementing the diamond challenge and the largest-series-product exercises. The .claude folder contains instructions that enable resolved exercises.

Copilot coding agent

I used implementing the Crypto Square exercise. This issue is notably brief compared to others, as the .github folder already includes Copilot prompts and agent instructions.

An earlier attempt was implementing the Wordy Exercise. During this implementation, the Copilot agent detects that no Copilot instructions were present in the repository and flagged this through a PR comment. This led to the creation of a dedicated issue:: ✨ Set up Copilot instructions.

With those instructions in place, I proceeded to implement the custom-set exercise. Interestingly, the existing exercism tests and instructions were already so comprehensive and well-structured that adding Copilot instructions didn’t significantly enhance the development experience. In both the Wordy and Custom Set exercises, the quality of the Exercism framework proved to be the dominant factor.

Windsurf IDE

Windsurf IDE successfully solved the luhn-from and atbash-cipher exercise. Also Windsurf solved the luhn-trait exercise with a much smaller previous version of the prompt.

Cursor IDE

Cursor IDE successfully solved the accumulate and list-ops exercise using its agent.

The result worked and was good code, still in one case clippy caused warnings (ba72c13).

A much smaller prompt:

Implement all _instructions_ defined in `phone-number/README.md` in the rust file:`phone-number/src/lib.rs`.

resolved the phone-number. Similar to above, the first implementation included clippy warnings (5c2175a)

Cursor ran multiple commands (cargo and others) and as a user, I had to confirm every of the requested commands.

Execution Process

Cursor's agent mode performed the following steps autonomously:

• Analyzed the exercise requirements from README.md
• Generated the implementation plan
• Implemented the solution in src/lib.rs
• Enabled all tests by removing #[ignore] attributes
• Executed cargo test to verify the solution
• Refactored the code for improved readability
• Re-tested to ensure refactoring didn't break functionality

User Interaction

• Approval Required: Each command execution (cargo test, file modifications) required user confirmation
• Transparency: All planned changes were shown before execution
• Control: User maintained full control over the process

Gemini CLI

Gemini cli is a terminal coding assitant that can solve exercism exercises: https://www.lotharschulz.info/2025/06/25/getting-started-with-google-gemini-cli-complete-setup-guide-and-rust-testing-experience/

Benchmark Harness Automation

You can auto-generate a Criterion benchmark scaffold for an exercise:

chmod +x .github/scripts/inject_benchmark.sh
.github/scripts/inject_benchmark.sh <exercise-dir> <function-name>
# Example:
.github/scripts/inject_benchmark.sh accumulate accumulate

This will:

• Copy the benchmark template
• Insert a benchmark calling the specified function
• Add criterion to [dev-dependencies] if missing
• Append a [[bench]] section

Run benchmarks with:

cd <exercise-dir>
cargo bench

Copilot Prompt Taxonomy

Role	File	Purpose
Core implementation	solve-exercise.md	Main iterative solve loop (enable tests → implement → iterate)
Legacy alias	implement-exercise.md	Deprecated; forwards to solve-exercise.md
Validation	validate-solution.md	Post-implementation quality gates (tests, fmt, clippy, docs)
Idiomatic refinement	rust-idioms.md	Refactor toward clear, idiomatic Rust
Performance	optimize-solution.md	Benchmark & improve critical paths
Documentation	documentation.md	Add/verify /// docs and README improvements
Troubleshooting	troubleshoot-tests.md	Deep dive when failures persist after solve loop
TDD mode	test-driven.md	Red → Green → Refactor workflow
Shared reference	shared-core-flow.md	Minimal canonical sequence (referenced by others)

Doc Enforcement

If agent.yml sets metadata.default.enforce_doc_on_public: true, CI will fail when public items lack doc comments. To temporarily relax this, set the flag to false in .github/copilot/agent.yml (not recommended for long-term hygiene).

Optimization Benchmarks

The optimize prompt assumes (but does not force) criterion usage. Use the injector script to scaffold when exploring performance changes:

.github/scripts/inject_benchmark.sh <exercise> <function>