@spec build_state([{[String.t()], [String.t()]}]) :: Lightning.Scrubber.State.t()

Builds the state scrub_string/2 takes, without starting a process.

scrub/2 needs an agent because a run accumulates samples as it goes. A caller that already knows every sample would otherwise start - and stay linked to - an agent per string it scrubs.

Takes {samples, basic_auth} pairs, the same shape add_samples/3 takes.

Returns a specification to start this module under a supervisor.

See Supervisor.

@spec encode_samples(samples :: [String.t()], basic_auth :: [String.t()]) :: [
  String.t()
]

Prepare a list of sensitive samples (strings) into a potentially bigger list composed of variations a sample may appear.

This is used for scrubbing logs (is_list) and JSON (is_binary)

@spec scrub_string(Lightning.Scrubber.State.t(), String.t() | nil) :: String.t() | nil

Process-free counterpart to scrub/2 for a single string.

@spec scrub_values(any(), non_neg_integer()) :: any()

Recursively scrubs all values from a data structure, replacing them with type placeholders. Keeps the structure and the field names, and hides the values.

Field names are not values and survive as they are, so a body keyed by record identifier would carry those identifiers out. Long lists keep two samples and keys are capped across the whole structure, and a map that was truncated says how many it dropped under a "..." key. Arrays are sampled up to array_limit elements with a "...N more" indicator if truncated.

Examples

iex> scrub_values(%{"name" => "John", "age" => 30})
%{"name" => "string", "age" => "number"}

iex> scrub_values([1, 2, 3])
["number", "number", "...1 more"]