Normalize

better-supabase/normalize is a set of pure functions for import mappers. Locale rules are data: a country is a calling code, a postal shape or a national-id length, not a branch inside the function. The same checks exist in SQL as better_supabase.validate_iban and better_supabase.normalize_phone.

import {
  TEXT_PACKS,
  normalizePhone,
  sanitizeText,
  splitPersonName,
  validateIban,
} from "better-supabase/normalize";

splitPersonName("A.M. Gerritsen").initials; // "A.M."
normalizePhone("06 12345678", { defaultRegion: "NL" }); // "+31612345678"
validateIban("NL91 ABNA 0417 1643 00"); // true

sanitizeText("café", {
  repairs: TEXT_PACKS["utf8-as-latin1"],
  stripControl: true,
  csvQuotes: true,
});

sanitizeText applies each repair once, as a literal string. The named packs are utf8-as-latin1, windows-1252, cp437 and cp850. No pattern for one source is built in. detectMojibake returns a score from 0 to 1, and looksShouted is true when most longer words are capitals.

splitPersonName keeps initials and never expands them, and puts particles such as van der in infix. classifyParty uses that split. auto treats a legal suffix (bv, gmbh, ltd) as a company and stores legalName. A person keeps initials, so A. Gerritsen and A.M. Gerritsen stay different.