Optimize StatusLine with unsafe

[stefannikolei]

Just trying things out how AI can help in this cases. I think Claude Sonnet 4 did a pretty reasonable job.

[Copilot]

Pull Request Overview

This PR refactors StatusLine to use a raw SweepEvent[] buffer with manual resizing and unsafe access for faster inserts/removals.

• Replace List<SweepEvent> with an internal array plus a count field
• Use Unsafe.Add and MemoryMarshal.GetArrayDataReference for element access
• Implement custom binary-search insertion, shifting via Array.Copy, and position updates

[Copilot]

The indexer bypasses bounds checks, which can lead to out-of-range reads/writes. Add an explicit check (e.g., if (index < 0 || index >= this.count) throw new ArgumentOutOfRangeException(nameof(index));) before the unsafe access.

Suggested change

	get => Unsafe.Add(ref MemoryMarshal.GetArrayDataReference(this.events), index);
	get
	{
	if (index < 0 || index >= this.count)
	{
	throw new ArgumentOutOfRangeException(nameof(index));
	}
	return Unsafe.Add(ref MemoryMarshal.GetArrayDataReference(this.events), index);
	}

[JimBobSquarePants]

The bot has a point. We've lost the safety of a bounds check here.

[Copilot]

RemoveAt does not validate that index is within [0, Count). Invalid indices will cause undefined behavior. Consider adding a bounds check and throwing ArgumentOutOfRangeException to match the original List<T> semantics.

Suggested change

	{
	{
	// Validate index
	if (index < 0 || index >= this.count)
	{
	throw new ArgumentOutOfRangeException(nameof(index), "Index must be within the range [0, Count).");
	}

[Copilot]

Growing the array by a fixed amount (+16) leads to O(n²) behavior when many inserts occur. Switch to exponential growth (e.g., newCapacity = this.events.Length * 2) for amortized O(1) resizes.

Suggested change

	// TODO: Is this the best growth strategy?
	int newCapacity = this.events.Length + 16;
	// TODO: Switch to exponential growth for better performance.
	int newCapacity = this.events.Length == 0 ? 16 : this.events.Length * 2;

[JimBobSquarePants]

I was going to do a thorough review here, but Claude has made far too unnecessary changes. I don't think it's safe to rely on AI for these kinds of changes, the result is often very poor.

We should also be providing unit tests and benchmarks to proof performance improvements.

[JimBobSquarePants]

The bot has a point. We've lost the safety of a bounds check here.