Handling Errors

Iron Laws

1. Never swallow errors - Empty catch blocks hide bugs that surface later in unrelated places, making them much harder to trace
2. Never convert errors to booleans - Loses all context and forces callers to guess what went wrong
3. Preserve error context when wrapping or propagating - upstream handlers need the original cause to make good decisions
4. Log once where handled, not at every layer - duplicate logs across layers create noise that obscures the real signal

Error Messages

Every error message answers: What happened? Why? How to recover?

For logs (developers):