Performance

Start with the idea

Iterator chains compile to roughly the same code as a hand-written loop, so readability is rarely the cost. What does cost is allocating in a hot loop, cloning to satisfy the borrow checker, and choosing the wrong container. Reserve capacity when the size is known, borrow instead of cloning, and measure in release — a debug build is not evidence.

Performance work begins with a workload and a measurement. A release build is essential because debug checks and missing optimizations can dominate a microbenchmark. Use realistic input sizes and record latency distributions or throughput, not one attractive timing. Profile before changing code: time may be in parsing, allocation, I/O, locking, or an algorithmic choice far from the line that looks complicated. An iterator chain is often optimized well, but verify hot paths rather than treating any syntax as inherently free or expensive.