Revised Performance Issue

[dvmhmdsd]

After profiling qs with various input patterns, I've noticed some potential optimization opportunities that could improve performance, particularly for complex nested objects and high-traffic scenarios.

Areas Identified

1. Regex Compilation Overhead

Regex objects are recreated on each function call, which adds overhead in parsing-heavy workloads.

2. String Processing Efficiency

Multiple sequential string operations could potentially be combined for better performance.

3. Memory Allocation Patterns

Several hot paths create temporary objects that could potentially be optimized.

I see potential for 15-25% performance improvements in certain scenarios, particularly:

• Complex nested object parsing
• High-frequency parsing operations
• Large query strings

1. Is performance optimization a current priority for the project?
2. What's the preferred approach for performance changes (separate PRs, comprehensive analysis first, etc.)?

If performance improvements would be valuable, I can:

• Ensure zero behavioral changes
• Maintain full backward compatibility

Would this type of contribution be helpful?

[ljharb]

For changes that stand on their own merits, like hoisting things to module scope, that'd be a great PR. Do note that regexes are mutable, so they can't necessarily be safely hoisted.

For anything that's only motivated by performance, we'd need some kind of benchmark that could be ran before and after the change, and in the future to prevent regressions. We don't have any benchmarking currently set up, and it'd be ideal if it ran on every node version in the matrix to ensure that performance increases in some versions don't result in decreases in others.

I'd say performance optimization is always welcomed but certainly not a priority, since correctness always trumps speed.