Strings Best Practices

How to Use This List

• Review APIs taking String everywhere.
• Audit UTF-8 assumptions before Windows/path work.

A - API Types

• Accept &str (or impl AsRef<str>) for read-only inputs. Caller flexibility.
• Return String when caller owns result. Borrow only when tied to input lifetime.
• Use Cow<str> when mutation is rare. Avoid forced allocation.
• Use Path/PathBuf for filesystem paths. Not String paths.
• Use OsString when preserving non-UTF-8 OS bytes. Windows env vars.

B - UTF-8 Correctness

• Never index str by byte without is_char_boundary. Prevent panics.
• Use chars() for Unicode scalars; graphemes for UI. unicode-segmentation.
• Validate external bytes with from_utf8. Do not unwrap in production.
• Do not assume len() is character count. Document byte length.
• Normalize when comparing user text (NFC). unicode-normalization if needed.

C - Performance

• Preallocate with with_capacity when size known. Join/build loops.
• Prefer push_str over + in loops. + moves left operand.
• Reuse String buffers with clear. Hot formatting paths.
• Compile regex once (LazyLock). Not per line.
• Avoid to_string/clone in tight loops. Profile with dhat.

D - FFI and Security

• CString rejects interior NUL. Validate before FFI.
• Document lifetime of pointers passed to C. Who frees?
• Redact secrets in Display/Debug logs. Custom impls.
• Validate parse input; no unwrap on user data. Return Result.
• Sanitize before shell invocation. Not regex alone.

FAQs

Related

• Strings Basics
• UTF-8 & Indexing
• Cow
• Modules Basics

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+.