One identifier in the catalog: a regex that finds values of its shape, and the check a value must pass before it counts.

Fields

• name – the stable identifier, an atom such as :iban or :personnummer_se. Once a pattern is released its name never changes, so it can be stored.
• label, short_label, plural – how to name it in a full sentence ("US Social Security number"), in a compact list ("SSN") and in running text ("US social security numbers").
• regex – finds candidate values. Compiled without the u flag, so it matches bytewise; see PiiCatalog.Patterns for the boundaries this implies.
• regions – where the identifier is used, as region atoms (PiiCatalog.regions/0). The first is the region it is listed under.
• confidence – how a match is confirmed:
  • :checksum – a validator confirms the value (a check digit or a structure the regex cannot express);
  • :format – the regex alone decides;
  • :format_with_context – the identifier has no check digit, or one too weak on its own, so the regex also requires a word naming it directly before the value, and starts the match after that word with \K. The validator still runs on the value.
• check – what the validator checks: :check_digit (a check digit, letter or character computed from the rest of the value), :structure (a structure only, such as a date of birth or ranges that are never issued, with no check digit), :mixed (a check digit for some values and a structure only for others, as for EU VAT numbers, where some member states' numbers have no check), or nil (no validator).
• keyword_required – true when the regex matches only directly after a word naming the identifier. That is every :format_with_context pattern, where the word is not part of the match, and :date_of_birth, where it is.
• validator – the name of the check in PiiCatalog.Validators, or nil when the format alone decides.
• requires – what a text must contain for the pattern to be scanned at all: a literal every match contains ("@"), or a list of regexes, one of which every text holding a match also matches. nil means always scan.
• example – a test or invented value, never a real person's. For a :format_with_context pattern it includes the keyword.
• notes – what the regex accepts, in a sentence or two.