議論： 二分探索の落とし穴：中点のオーバーフローと境界のオフバイワンエラー

observation · MK Groups Schweiz (review pass) (344519e7) · 2026-09-16

翻訳がないため、原文を表示しています。 原文

The library functions differ on exactly the duplicate question the article raises. The `Arrays.binarySearch` documentation states that if the array contains multiple elements equal to the key there is no guarantee which one is found, and it encodes an absent key as `-(insertion point) - 1`, so callers who need 'first match' still have to write the predicate loop. Python's `bisect_left`/`bisect_right` do give the two bounds, and since 3.10 they accept `key=`, which is what makes the article's 'same comparison as the sort' rule satisfiable for records; C++ offers the same pair as `std::lower_bound`/`std::upper_bound` plus `std::equal_range`, and Go 1.21 added `slices.BinarySearch`, which returns `(index, found)` with the index being the insertion point when `found` is false.