We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Lobsters·22d ago·5 min read
Should Libraries Log or Propagate Errors?
TL;DR
Libraries often log errors instead of simply propagating them, which can complicate error handling. This approach allows for enriched error context and adjustable log levels.
✦ Why It Matters
Developers should evaluate their logging strategies to balance error visibility and performance in production environments.
Key Takeaways
- Consider logging levels to manage verbosity in production.
- Use enriched error propagation for better debugging.
- Implement debug-only assertions to catch issues early.
- Evaluate the cost of additional error checks in production.
Full Summary
Traditionally, applications log errors while libraries are expected to simply return them. However, many programming ecosystems, including Go, emphasize logging within libraries, leading to a need for developers to manage log verbosity.
This shift allows for enhanced error context and the ability to adjust log levels, which can be crucial for debugging. For instance, in Go's slog, developers can propagate errors with additional details, rather than just returning them.
This method can be particularly useful in development environments where extra error checks are beneficial but too costly for production. Examples include Vulkan validation layers and assertions that are only active in debug mode.
Recognizing these practices can help engineers design better error handling strategies.