Select data structures and algorithms based on their performance characteristics and expected usage patterns. Consider the size of collections, access patterns, and computational complexity when making implementation choices. **Key considerations:** - Use `HashSet` instead of `Vec` for membership testing when collections can be large - Choose data structures that preserve order when sequence matters (e.g., `IndexMap` vs `HashMap`) - Prefer more efficient algorithms when available (e.g., checking if a commit is a head vs. evaluating all children) - Combine operations to reduce computational overhead (e.g., union revset expressions rather than evaluating separately) **Example:** ```rust // Instead of Vec for large collections with frequent lookups let already_tracked_bookmarks: Vec = ...; if already_tracked_bookmarks.contains(&bookmark) { ... } // O(n) lookup // Use HashSet for O(1) average lookup time let already_tracked_bookmarks: HashSet = ...; if already_tracked_bookmarks.contains(&bookmark) { ... } // O(1) lookup ``` This applies especially when dealing with user-facing operations that may scale with repository size or when processing large collections where the choice of data structure significantly impacts performance.