An Ordered Map Can Answer a Range Query Directly

A transcript viewer may need every segment whose start time falls within the visible window, such as 100 through 199 milliseconds in a small example. If segments are stored under ordered start times, the query can name that interval directly. Rust’s BTreeMap::range returns an iterator over entries within the requested key bounds.

use std::collections::BTreeMap;

let mut segments = BTreeMap::new();
segments.insert(100_u64, "open settings");
segments.insert(160, "choose audio");
segments.insert(220, "save changes");

for (&start_ms, text) in segments.range(100..200) {
    println!("{start_ms}: {text}");
}

The 100..200 range includes 100 and excludes 200, so this example yields the first two entries. The map maintains entries in key order, and range gives a double-ended iterator over the selected subrange. Other bound shapes are possible: a pair of Bound values can make the lower end exclusive or the upper end inclusive. That is useful when a cursor page should start strictly after the last key already shown.

The key has to express the product’s ordering. A time-window query works when keys sort by time; a map keyed only by segment ID cannot answer a time interval through range unless those IDs encode the same order. Also, a map has one value per key. If two transcript segments start at the same millisecond, storing both under that millisecond alone would replace one. A composite key whose order compares time first and a unique tie-breaker second preserves both while keeping chronological order. This key design is part of the feature, not a detail to patch after data disappears.

Validate bounds at the API edge. The standard library specifies a panic when the range start exceeds its end, and another when both bounds exclude the same key. An ordinary half-open start..end with equal endpoints simply selects an empty interval. A user dragging a timeline backward can produce reversed inputs, so normalize or reject them before calling range.

An ordered map is a good fit when the interface repeatedly asks for windows or neighboring segments in key order. Its ordering has a maintenance cost, so use the product’s actual query pattern to choose the index. For transcript timelines, define a stable sortable key first, then test inclusive, exclusive, empty, and reversed windows at the boundary.