v3.0.0

[beningreenjam]

Full Changelog: v2.2.1...v3.0.0

---

This discussion was created from the release v3.0.0.

[beningreenjam]

Important Update on Performance Benchmarks - DeepRedact v3.0.0

With the release of v3.0.0, I want to address an important matter regarding our previous performance claims and share some exciting improvements.

Acknowledgment and Apology

I need to be transparent with you: the benchmark tests in previous versions of DeepRedact contained inaccuracies that led to misleading performance comparisons. Specifically, our claims that DeepRedact was faster than Fast Redact were incorrect due to flawed testing methodology. I sincerely apologise for this oversight and any confusion it may have caused.

The Real Performance Picture

After correcting our benchmark tests, I discovered that previous versions of DeepRedact were approximately 4.5-5x slower than Fast Redact. This was a significant gap that needed addressing, and it's why I've removed the benchmark results from our README - I believe in honest, accurate performance reporting.

Significant Improvements in v3.0.0

However, I'm excited to share that v3.0.0 represents a major leap forward:

Performance Improvements:

• The performance gap with Fast Redact has been dramatically reduced from ~4.5-5x to approximately 1.6x
• Single object redaction: ~0.011ms vs Fast Redact's ~0.007ms
• Bulk operations (1000 objects): ~10.9ms vs Fast Redact's ~6.6ms

Additional Enhancements:

• Significantly more configurable - extensive customization options for your specific use cases
• More reliable - improved stability and consistency across different scenarios
• Better developer experience - cleaner API and more intuitive configuration

Moving Forward

While DeepRedact may not be the fastest redaction library available, it offers a compelling combination of performance, flexibility, and reliability that makes it ideal for complex redaction scenarios where configurability matters as much as speed.

I'm committed to continued transparency in our performance reporting and will only share benchmark results that have been thoroughly validated. Going forward, benchmark tests will be used primarily for internal development comparisons to ensure we're always improving.

Thank you for your continued support and trust in DeepRedact.