DSL for defining tabular-data validation schemas.

use Scrutinex.Schema (with optional strict: true) imports the column/2, column/3, and check/2/check/3 macros and registers a __schema__/0 callback at compile time.

Supported types

:string, :integer, :float, :boolean, :date, :datetime

Column options

• :required - fail if the column key is missing (default true)
• :coerce - attempt type casting before validation (default false)
• :on_empty - behaviour when a cell is nil or "" (default :error):
  • :error — severity :error, skips remaining checks
  • :warn — severity :warning, skips remaining checks, cell becomes nil in result data
  • :ignore — no error, skips remaining checks
• :nullable - accepted as compile-time sugar for on_empty: :ignore; cannot be combined with :on_empty
• :checks - keyword list of built-in checks:
  • number: [greater_than: 0, less_than: 100]
  • inclusion: ["A", "B", "C"]
  • exclusion: ["X"]
  • format: ~r/^[A-Z]/
  • length: [min: 1, max: 255]
  • custom: &my_function/1

Strict mode

Pass strict: true to use Scrutinex.Schema to reject rows that contain columns not declared in the schema.

Cross-column checks

Use the check macro to validate relationships between columns. The block receives the full row map and must return a truthy value to pass.

Trust Boundary

Schema definitions, including custom check functions and cross-column check blocks, are trusted code. They execute with the same privileges as your application. Do not construct schemas from untrusted user input.

Example

defmodule MyApp.PeopleSchema do
  use Scrutinex.Schema, strict: true

  column "name",  :string,  checks: [length: [min: 1, max: 100]]
  column "age",   :integer, coerce: true, checks: [number: [greater_than: 0]]
  column "email", :string,  checks: [format: ~r/@/]
  column ~r/^score_/, :float, required: false, on_empty: :ignore

  check :age_name_consistency do
    fn row -> !(row["age"] > 150 && row["name"] == "") end
  end
end

result = Scrutinex.validate(data, MyApp.PeopleSchema)