Collections & Iterators Best Practices

How to Use This List

• Profile before micro-optimizing iterators.
• Enable clippy needless_collect and iter_overeager_cloned.

A - Collection Choice

• Default to Vec for sequences. Specialized containers when measured need.
• HashMap for speed, BTreeMap for order/range. Document choice.
• Avoid LinkedList unless proven. VecDeque usually wins.
• Reserve capacity when size known. with_capacity reduces realloc.
• Use entry API for map updates. Avoid double lookup.

B - Iterators

• Prefer iterators over manual index loops. enumerate when index needed.
• Avoid needless collect between adapters. Chain lazily.
• Use try_collect / Result collect for fallible maps. Short-circuit errors.
• Borrow with iter() unless ownership needed. into_iter consumes.
• Bound infinite iterators with take. Prevent hangs.

C - Allocation

• Reuse buffers in hot loops. clear + reuse Vec.
• Prefer &str keys only when lifetimes allow. Owned String keys common.
• Consider SmallVec for usually-small lists. Fewer heap hits.
• Profile allocator on hot paths. dhat, heaptrack.
• Do not premature shrink_to_fit everywhere. Only after spikes.

D - Correctness

• Do not mutate vec during non-adapter iteration. Use retain, drain.
• Handle empty reduce/fold. Provide seed or Option.
• Document sort stability requirements. sort vs sort_unstable.
• Test equality independent of map order. Sort keys or use BTreeMap in tests.
• Use typed collect explicitly in complex chains. Aid inference.

FAQs

Related

• Collections Basics
• Iterator Adapters
• Entry API & In-Place Updates
• Closures Best Practices

Stack versions: This page was written for Rust 1.97.0 (edition 2024), Tokio 1.x, Axum 0.8, serde 1.0, sqlx 0.8, clap 4, and Polars 0.46+.