sorting

Unicode string equivalence is complex, especially with collations, leading to surprising results like deletion of control characters and non-deterministic grouping. The author discusses challenges in implementing proper Unicode support in database systems.

A Go engineer recounts an incident where an in-memory datastore became overloaded due to slow sorting, and they implemented context cancellation inside sort functions by using panics and recover for non-local flow control, similar to how encoding/json handles errors.

Explains why the classic Win32 TreeView control offers separate sort methods for name and lParam, and recommends switching to a data model pattern when both are needed for sorting.