Sort Lines Online — Alphabetize Lines (A-Z, Numeric, Random)

Sorting accented characters depends on the active locale's collation rules, not raw Unicode code-point order. The Unicode Collation Algorithm (UTS #10, Whistler & Scherer eds., revision 51 tied to Unicode 16.0) defines the Default Unicode Collation Element Table — base letters first, with accents and case as secondary and tertiary weights — but locale-specific tailorings from CLDR (Common Locale Data Repository) override the defaults for languages whose conventions differ. In Spanish, for example, ñ is a distinct letter sorting after n (so 'naranja' precedes 'ñame'), while default UCA places ñ together with n. Both ECMA-262 String.prototype.localeCompare and ECMA-402 Intl.Collator delegate to UCA + CLDR, so the result depends on which locale is active when the sort runs.

The shuffle implements Fisher-Yates — originally Fisher & Yates 1938 (Statistical Tables for Biological, Agricultural and Medical Research, pen-and-paper using a random-number table), then Durstenfeld 1964 (CACM 7(7):420, Algorithm 235) for the in-place computer form, popularized as Algorithm P (Shuffling) in Knuth's The Art of Computer Programming Vol 2 §3.4.2. The algorithm walks the array right-to-left, at index i picking a random index j in [0, i] and swapping. Every permutation appears with equal probability provided the random source is unbiased — this tool draws entropy from crypto.getRandomValues() (W3C Web Cryptography API, Recommendation 26 January 2017) and applies rejection sampling to map 32-bit values into [0, i+1] without the modulo bias that affects naive Math.floor(Math.random() * n) implementations.

parseFloat on a non-numeric input returns NaN. Per ECMA-262, when a comparator returns NaN the spec treats it as 0 (equal), and ES2019 made Array.prototype.sort stable — so non-parseable lines keep their original relative order rather than getting a deterministic numeric position, while the parseable lines around them are sorted numerically. In practice this means a numeric column with stray text, dates, or empty lines does not silently push non-conforming entries somewhere predictable — those entries cluster wherever they originally were among other non-numbers. Pre-filtering the input with text-replace to strip non-numeric lines first gives a clean numeric-only sort.

The case option controls only the comparator. With case-insensitive mode on, both sides are lowercased before comparison, so 'Apple' and 'apple' are treated as equal — but the lines that appear in the output retain their original casing. This matches the convention of database COLLATE NOCASE and Excel's case-insensitive sort: capitalization, leading whitespace, and other display-only differences are preserved even though they did not influence ordering. The behaviour is deterministic across re-runs because both localeCompare and the input lines are stable.

The sorted result region is marked aria-live="polite", the W3C WCAG Success Criterion 4.1.3 (Status Messages, introduced in WCAG 2.1, Recommendation 5 June 2018; carried unchanged into WCAG 2.2, Recommendation 5 October 2023) pattern. Polite is appropriate because the sort isn't urgent — it queues announcements behind any speech the user is hearing rather than interrupting mid-sentence. Screen readers (NVDA, JAWS, VoiceOver) consume the live region automatically; nothing else is required from the user.