Peri (peri v0.10.0)

Copy Markdown View Source

Peri is a schema validation library for Elixir, inspired by Clojure's Plumatic Schema. It provides a flexible and powerful way to define and validate data structures using schemas. The library supports nested schemas, optional fields, custom validation functions, and various type constraints.

Key Features

  • Simple and Nested Schemas: Define schemas that can handle complex, nested data structures.
  • Optional and Required Fields: Specify fields as optional or required with type constraints.
  • Custom Validation Functions: Use custom functions to validate fields.
  • Comprehensive Error Handling: Provides detailed error messages for validation failures.
  • Type Constraints: Supports various types including enums, lists, maps, tuples, literals, and more.

Usage

To define a schema, use the defschema macro. By default, all fields in the schema are optional unless specified otherwise.

defmodule MySchemas do
  import Peri

  defschema :user, %{
    name: :string,
    age: :integer,
    email: {:required, :string},
    address: %{
      street: :string,
      city: :string
    },
    tags: {:list, :string},
    role: {:enum, [:admin, :user, :guest]},
    geolocation: {:tuple, [:float, :float]},
    preferences: {:map, :string},
    scores: {:map, :string, :integer},
    status: {:literal, :active},
    rating: {:custom, &validate_rating/1}
  }

  defp validate_rating(n) when n < 10, do: :ok
  defp validate_rating(_), do: {:error, "invalid rating", []}
end

You can then use the schema to validate data:

user_data = %{
  name: "John", age: 30, email: "john@example.com",
  address: %{street: "123 Main St", city: "Somewhere"},
  tags: ["science", "funky"], role: :admin,
  geolocation: {12.2, 34.2},
  preferences: %{"theme" => "dark", "notifications" => "enabled"},
  scores: %{"math" => 95, "science" => 92},
  status: :active,
  rating: 9
}

case MySchemas.user(user_data) do
  {:ok, valid_data} -> IO.puts("Data is valid!")
  {:error, errors} -> IO.inspect(errors, label: "Validation errors")
end

Error Handling

Peri provides detailed error messages that include the path to the invalid data, the expected and actual values, and custom error messages for custom validations.

Schema Types

Peri supports the following schema types:

  • :string, :integer, :float, :boolean, :atom, :map, :pid - Basic types
  • {:required, type} - Mark a field as required
  • {:list, type} - List of elements of the given type
  • {:map, type} - Map with values of the given type
  • {:map, key_type, value_type} - Map with keys and values of specified types
  • {:schema, schema} - Explicitly tagged nested schema
  • {:schema, map_schema, {:additional_keys, type}} - Nested schema map, with extra entries validated using another type
  • {:tuple, [type1, type2, ...]} - Tuple with elements of specified types
  • {:enum, [value1, value2, ...]} - One of the specified values
  • {:enum, [value1, value2, ...], opts} - Enum with opts (:type, :error, :gen); :type constrains members to a base type and surfaces it in JSON Schema output
  • {:literal, value} - Exactly matches the specified value
  • {:either, {type1, type2}} - Either type1 or type2
  • {:oneof, [type1, type2, ...]} - One of the specified types
  • {:cond, condition, true_type, false_type} - Conditional validation based on callback
  • {:dependent, callback} - Dynamic type based on callback result
  • {:coerce, source, target} - Coerce a value from source representation to target type during validation
  • {type, {:encode, fun}} - Apply fun to the value when encoding with Peri.encode/3 (inert under validation)
  • {:meta, type, opts} - Attach documentation/example/description to a field; passthrough at validation
  • Nested maps for complex structures

Coercion and Encoding

{:coerce, source, target} validates boundary data (e.g. Phoenix params, where everything arrives as a string) by coercing from the source representation into the target type. Values that already match the target type pass through untouched, so JSON-decoded data is a no-op.

defschema :params, %{
  page: {:coerce, :string, :integer},
  active: {:coerce, :string, :boolean}
}

MySchemas.params(%{"page" => "2", "active" => "true"})
# => {:ok, %{page: 2, active: true}}

Built-in sources support :string into :integer, :float, :boolean, :atom, :date, :time, :naive_datetime, and :datetime. Custom sources are 1-arity functions (or MFA tuples) returning {:ok, value} or :error.

Peri.encode/3 runs the same schema in the opposite direction: it validates data against the target types and produces the wire representation (strings for built-ins, or the encode: opt when given). Peri.decode/3 is an alias of validate/3 for API symmetry. {:transform, fun} is skipped during encoding, and {type, {:encode, fun}} applies only during encoding.

Custom Error Messages

Override the default validation message per field via the error: opt in the type's options list. Accepts either a static string or an MFA tuple {module, function, args}. The MFA receives the %Peri.Error{} (with its content) prepended to args and must return a string.

%{
  age:   {:integer, gte: 18, error: "must be adult"},
  email: {:required, :string, [error: {MyApp.Errors, :email_msg, []}]}
}

For i18n / Gettext, walk the resulting errors with Peri.Error.traverse_errors/2 and translate each leaf message — see that function's docs for an example.

Schema Metadata

Fields can carry metadata via the {:meta, type, opts} wrapper. Metadata is ignored at validation time but available for documentation, JSON Schema export, and tooling. The JSON Schema encoder recognises the standard Draft-7 annotation/format vocabulary (:title, :description, :example, :examples, :deprecated, :default, :format, :pattern, :read_only, :write_only, :content_encoding, :content_media_type); other keys (e.g. :doc) are preserved opaquely for non-encoder tooling.

defschema :user, %{
  email: {:meta, {:required, :string}, doc: "Login email", example: "a@b.io"},
  age: {:meta, {:integer, gte: 0}, description: "Years"}
}, title: "User", description: "Account holder"

Schema-level meta opts are exposed via the generated __schema_meta__/1 function. Validation opts (e.g. :mode) are split out, not surfaced as meta.

Callback Functions for :cond and :dependent

Both :cond and :dependent types support 1-arity and 2-arity callbacks:

  • 1-arity callbacks receive the root data structure (backward compatible)
  • 2-arity callbacks receive (current, root) where:
    • current is the data at the current validation context (e.g., list element)
    • root is the entire root data structure

This is especially useful when validating elements within lists:

defschema :parent, %{
  items: {:list, %{
    type: :string,
    value: {:dependent, fn current, _root ->
      case current.type do
        "number" -> {:ok, :integer}
        "text" -> {:ok, :string}
        _ -> {:ok, :any}
      end
    end}
  }}
}

Custom Generators

When data generation matters (Peri.generate/1), constrained types like {:integer, gt: 1_000_000} or {:string, {:regex, …}} fall back to rejection sampling, which can be slow on tight domains. Provide a gen: opt with an MFA, {mod, fun}, or 0-arity function returning %StreamData{} to skip rejection entirely. Accepted in multi-options, {:required, type, opts}, and {:meta, type, opts} positions.

%{
  age:   {:integer, gte: 18, gen: {MyApp.Gens, :age, []}},
  email: {:meta, :string, doc: "Login", gen: {MyApp.Gens, :email}}
}

Schema Transformation

Peri.walk/2 runs a depth-first rewrite over a schema, useful for derivations like "make every field optional" or "strip private keys from a public DTO". The callback receives {:field, key, value} for entries inside a map/keyword schema and the type expression itself everywhere else; return {:cont, _} to continue or :drop to remove a field. See Peri.Walker for full semantics.

Peri.walk(schema, fn
  {:required, t} -> {:cont, t}
  other -> {:cont, other}
end)

Functions

  • validate/2 - Validates data against a schema.
  • decode/3 - Alias of validate/3; decodes wire data, applying {:coerce, ...} directives.
  • encode/3 - Validates data and produces its wire representation.
  • conforms?/2 - Checks if data conforms to a schema.
  • validate_schema/1 - Validates the schema definition.
  • generate/1 - Generates sample data based on schema (when StreamData is available).
  • walk/2 - Depth-first rewrite of a schema tree.

Example

defmodule MySchemas do
  import Peri

  defschema :user, %{
    name: :string,
    age: :integer,
    email: {:required, :string}
  }
end

user_data = %{name: "John", age: 30, email: "john@example.com"}
case MySchemas.user(user_data) do
  {:ok, valid_data} -> IO.puts("Data is valid!")
  {:error, errors} -> IO.inspect(errors, label: "Validation errors")
end

Summary

Functions

Checks if the given data conforms to the specified schema.

Decodes wire data against a schema.

Defines a schema with a given name and schema definition.

Encodes data against a schema, producing its wire representation.

Drops the given top-level keys from a map schema.

Decodes a JSON Schema (Draft 7) map into a Peri schema.

Generates sample data based on the given schema definition using StreamData.

Checks if the given data is an enumerable, specifically a map or a list.

Checks if the given data is a numeric value, specifically a integer or a float.

Checks if the given type as an atom is a numeric (integer or float).

Deep merges two map schemas into a new one.

Helper function to put a value into an enum, handling not only maps and keyword lists but also structs.

Keeps only the given top-level keys of a map schema.

Converts a Peri.schema() definition to an Ecto schemaless changesets.

Converts a Peri schema into a JSON Schema (Draft 7) map.

Validates a given data map against a schema with options.

Validates a schema definition to ensure it adheres to the expected structure and types.

Depth-first rewrite of a schema tree.

Types

coerce_def()

@type coerce_def() ::
  {:coerce, coerce_source(), schema_def()}
  | {:coerce, coerce_source(), schema_def(), keyword()}

coerce_source()

@type coerce_source() ::
  :string
  | (term() -> {:ok, term()} | :error)
  | {module(), atom()}
  | {module(), atom(), [term()]}

cond_def()

@type cond_def() ::
  {:cond, condition :: (term() -> boolean()), true_branch :: schema_def(),
   else_branch :: schema_def()}
  | {:cond, condition :: (current :: term(), root :: term() -> boolean()),
     true_branch :: schema_def(), else_branch :: schema_def()}

custom_def()

@type custom_def() ::
  {:custom, validation()}
  | {:custom, {module(), atom()}}
  | {:custom, {module(), atom(), [term()]}}

default_def()

@type default_def() ::
  {schema_def(), {:default, term()}}
  | {schema_def(), {:default, (-> term())}}
  | {schema_def(), {:default, {module(), atom()}}}

dependent_def()

@type dependent_def() ::
  {:dependent, field :: atom(), validation(), type :: schema_def()}
  | {:dependent,
     (term() ->
        {:ok, schema_def() | nil}
        | {:error, template :: String.t(), context :: map() | keyword()})}
  | {:dependent,
     (current :: term(), root :: term() ->
        {:ok, schema_def() | nil}
        | {:error, template :: String.t(), context :: map() | keyword()})}

encode_def()

@type encode_def() ::
  {schema_def(), {:encode, (term() -> term())}}
  | {schema_def(), {:encode, {module(), atom()}}}
  | {schema_def(), {:encode, {module(), atom(), [term()]}}}

explicit_schema_def()

@type explicit_schema_def() ::
  {:schema, schema()}
  | {:schema, map_schema(), {:additional_keys, schema_def()}}

float_def()

@type float_def() ::
  :float | {:float, numeric_option(float()) | [numeric_option(float())]}

int_def()

@type int_def() ::
  :integer | {:integer, numeric_option(integer()) | [numeric_option(integer())]}

literal()

@type literal() :: integer() | float() | atom() | String.t() | boolean()

map_schema()

@type map_schema() :: %{required(String.t() | atom()) => schema_def()}

numeric_option(type)

@type numeric_option(type) ::
  {:eq, type}
  | {:neq, type}
  | {:lt, type}
  | {:lte, type}
  | {:gt, type}
  | {:gte, type}
  | {:range, {min :: type, max :: type}}

schema()

@type schema() :: schema_def() | map_schema() | [{atom(), schema_def()}]

schema_def()

@type schema_def() ::
  :any
  | :atom
  | :boolean
  | :map
  | :pid
  | {:either, {schema_def(), schema_def()}}
  | {:oneof, [schema_def()]}
  | {:required, schema_def()}
  | {:meta, schema_def(), keyword()}
  | {:ref, atom()}
  | {:ref, {module(), atom()}}
  | {:multi, atom(), %{optional(term()) => schema_def()}}
  | {:enum, [term()]}
  | {:enum, [term()], keyword()}
  | {:list, schema_def()}
  | {:map, schema_def()}
  | {:map, key_type :: schema_def(), value_type :: schema_def()}
  | {:tuple, [schema_def()]}
  | {:literal, literal()}
  | time_def()
  | string_def()
  | int_def()
  | float_def()
  | default_def()
  | transform_def()
  | custom_def()
  | coerce_def()
  | encode_def()

string_def()

@type string_def() :: :string | {:string, string_option() | [string_option()]}

string_option()

@type string_option() ::
  {:regex, Regex.t()}
  | {:eq, String.t()}
  | {:min, integer()}
  | {:max, integer()}

time_def()

@type time_def() :: :time | :date | :datetime | :naive_datetime | :duration

transform_def()

@type transform_def() ::
  {schema_def(), {:transform, (term() -> term()) | (term(), term() -> term())}}
  | {schema_def(), {:transform, {module(), atom()}}}
  | {schema_def(), {:transform, {module(), atom(), [term()]}}}

validation()

@type validation() :: (term() -> validation_result())

validation_result()

@type validation_result() ::
  :ok | {:error, template :: String.t(), context :: map() | keyword()}

Functions

conforms?(schema, data, opts \\ [])

Checks if the given data conforms to the specified schema.

Parameters

  • schema: The schema definition to validate against.
  • data: The data to be validated.

Options

  • :mode - Validation mode. Can be :strict (default) or :permissive.
    • :strict - Only fields defined in the schema are returned.
    • :permissive - All fields from the input data are preserved.

Returns

  • true if the data conforms to the schema.
  • false if the data does not conform to the schema.

Examples

iex> schema = %{name: :string, age: :integer}
iex> data = %{name: "Alice", age: 30}
iex> Peri.conforms?(schema, data)
true

iex> invalid_data = %{name: "Alice", age: "thirty"}
iex> Peri.conforms?(schema, invalid_data)
false

decode(schema, data, opts \\ [])

@spec decode(schema(), data :: term(), keyword()) :: {:ok, term()} | {:error, term()}

Decodes wire data against a schema.

Coercion is a schema directive, so decoding is exactly validate/3: {:coerce, source, target} fields convert their source representation into the target type, and values already matching the target pass through unchanged.

Accepts the same options as validate/3 and returns the same shape.

Examples

iex> schema = %{page: {:coerce, :string, :integer}}
iex> Peri.decode(schema, %{"page" => "2"})
{:ok, %{page: 2}}

iex> schema = %{page: {:coerce, :string, :integer}}
iex> Peri.decode(schema, %{page: 2})
{:ok, %{page: 2}}

defschema(name, schema, opts \\ [])

(macro)

Defines a schema with a given name and schema definition.

Examples

defmodule MySchemas do
  import Peri

  defschema :user, %{
    name: :string,
    age: :integer,
    email: {:required, :string}
  }

  # With permissive mode
  defschema :flexible_user, %{
    name: :string,
    email: {:required, :string}
  }, mode: :permissive

  # With metadata (field-level and schema-level)
  defschema :documented_user, %{
    email: {:meta, {:required, :string}, doc: "Login email", example: "a@b.io"}
  }, title: "User", description: "Account holder"
end

# Schema-level metadata is accessible via __schema_meta__/1:
MySchemas.__schema_meta__(:documented_user)
# => [title: "User", description: "Account holder"]

user_data = %{name: "John", age: 30, email: "john@example.com"}
MySchemas.user(user_data)
# => {:ok, %{name: "John", age: 30, email: "john@example.com"}}

invalid_data = %{name: "John", age: 30}
MySchemas.user(invalid_data)
# => {:error, [email: "is required"]}

# Permissive mode preserves extra fields
flexible_data = %{name: "John", email: "john@example.com", role: "admin"}
MySchemas.flexible_user(flexible_data)
# => {:ok, %{name: "John", email: "john@example.com", role: "admin"}}

encode(schema, data, opts \\ [])

@spec encode(schema(), data :: term(), keyword()) :: {:ok, term()} | {:error, term()}

Encodes data against a schema, producing its wire representation.

Validates data against the schema (target types), then applies the reverse of each codec directive:

  • {:coerce, source, target} - built-in :string sources render with to_string/1 (to_iso8601/1 for date/time structs); when the encode: opt is given, that function/MFA is applied instead. Custom function/MFA sources without an encode: opt pass the validated value through unchanged.
  • {type, {:encode, fun}} - applies fun (or MFA) to the validated value. This directive is inert under validate/3 and decode/3.
  • {:transform, fun} - skipped; transforms are decode-only.

Accepts the same options as validate/3 and returns the same shape.

Examples

iex> schema = %{page: {:coerce, :string, :integer}}
iex> Peri.encode(schema, %{page: 2})
{:ok, %{page: "2"}}

iex> schema = %{upcased: {:string, {:encode, &String.downcase/1}}}
iex> Peri.encode(schema, %{upcased: "HELLO"})
{:ok, %{upcased: "hello"}}

except(schema, keys)

@spec except(map_schema(), [atom() | String.t()]) ::
  {:ok, map_schema()} | {:error, [Peri.Error.t()]}

Drops the given top-level keys from a map schema.

Keys not present in the schema are ignored. The resulting schema is checked with validate_schema/1, returning its error tuple when invalid.

Examples

iex> Peri.except(%{name: :string, age: :integer}, [:age])
{:ok, %{name: :string}}

iex> Peri.except(%{name: :string}, [:missing])
{:ok, %{name: :string}}

from_json_schema(json_schema, opts \\ [])

@spec from_json_schema(map(), [opt]) :: {:ok, schema()} | {:error, term()}
when opt: {:keys, :strings | :atoms | :atoms!}

Decodes a JSON Schema (Draft 7) map into a Peri schema.

Returns {:ok, schema} if the resulting Peri schema is valid, otherwise {:error, errors}.

generate(schema)

Generates sample data based on the given schema definition using StreamData.

This function validates the schema first, and if the schema is valid, it uses the Peri.Generatable.gen/1 function to generate data according to the schema.

Note that this function returns a Stream, so you traverse easily the data generations.

Parameters

  • schema: The schema definition to generate data for.

Returns

  • {:ok, stream} if the data is successfully generated.
  • {:error, errors} if there are validation errors in the schema.

Examples

iex> schema = %{name: :string, age: {:integer, {:range, {18, 65}}}}
iex> {:ok, stream} = Peri.generate(schema)
iex> [data] = Enum.take(stream, 1)
iex> is_map(data)
true
iex> data[:age] in 18..65
true

get_enumerable_value(enum, key)

is_enumerable(data)

(macro)

Checks if the given data is an enumerable, specifically a map or a list.

Parameters

  • data: The data to check.

Examples

iex> is_enumerable(%{})
true

iex> is_enumerable([])
true

iex> is_enumerable(123)
false

iex> is_enumerable("string")
false

is_numeric(n)

(macro)

Checks if the given data is a numeric value, specifically a integer or a float.

Parameters

  • data: The data to check.

Examples

iex> is_numeric(123)
true

iex> is_numeric(0xFF)
true

iex> is_numeric(12.12)
true

iex> is_numeric("string")
false

iex> is_numeric(%{})
false

is_numeric_type(t)

(macro)

Checks if the given type as an atom is a numeric (integer or float).

Parameters

  • data: The data to check.

Examples

iex> is_numeric(:integer)
true

iex> is_numeric(:float)
true

iex> is_numeric(:list)
false

iex> is_numeric({:enum, _})
false

is_of_numeric_type(val, t)

(macro)

is_type_with_multiple_options(t)

(macro)

merge(schema1, schema2)

@spec merge(map_schema(), map_schema()) ::
  {:ok, map_schema()} | {:error, [Peri.Error.t()]}

Deep merges two map schemas into a new one.

When both schemas define the same key as a plain map (a nested schema), the nested schemas are merged recursively. Otherwise the value from the right side wins: type tuples and directives are never merged structurally.

The resulting schema is checked with validate_schema/1, returning its error tuple when the composed schema is invalid.

Examples

iex> Peri.merge(%{name: :string}, %{age: :integer})
{:ok, %{name: :string, age: :integer}}

iex> Peri.merge(%{name: :string}, %{name: :atom})
{:ok, %{name: :atom}}

iex> Peri.merge(%{user: %{name: :string}}, %{user: %{age: :integer}})
{:ok, %{user: %{name: :string, age: :integer}}}

put_in_enum(enum, key, val)

Helper function to put a value into an enum, handling not only maps and keyword lists but also structs.

Examples

iex> Peri.put_in_enum(%{}, :hello, "world")
iex> Peri.put_in_enum(%{}, "hello", "world")
iex> Peri.put_in_enum(%User{}, :hello, "world")
iex> Peri.put_in_enum([], :hello, "world")

select(schema, keys)

@spec select(map_schema(), [atom() | String.t()]) ::
  {:ok, map_schema()} | {:error, [Peri.Error.t()]}

Keeps only the given top-level keys of a map schema.

Keys not present in the schema are ignored. The resulting schema is checked with validate_schema/1, returning its error tuple when invalid.

Examples

iex> Peri.select(%{name: :string, age: :integer}, [:name])
{:ok, %{name: :string}}

iex> Peri.select(%{name: :string}, [:name, :missing])
{:ok, %{name: :string}}

to_changeset!(s, attrs)

@spec to_changeset!(schema(), attrs :: map()) :: Ecto.Changeset.t()

Converts a Peri.schema() definition to an Ecto schemaless changesets.

to_json_schema(schema, opts \\ [])

@spec to_json_schema(schema(), Peri.JSONSchema.Encoder.opts()) :: map()

Converts a Peri schema into a JSON Schema (Draft 7) map.

Reads {:meta, type, opts} annotations and emits title, description, examples, deprecated. Dynamic types degrade per :on_unsupported (:omit | :true_schema | :raise, default :omit).

Pass :exclude_meta_keys to drop annotation keywords from the output — commonly [:default] when the consumer-facing schema should not surface validation defaults.

Examples

iex> Peri.to_json_schema(%{name: {:required, :string}})
%{"type" => "object", "properties" => %{"name" => %{"type" => "string"}}, "required" => ["name"]}

iex> Peri.to_json_schema({:integer, {:default, 0}}, exclude_meta_keys: [:default])
%{"type" => "integer"}

validate(schema, data, opts \\ [])

Validates a given data map against a schema with options.

Returns {:ok, data} if the data is valid according to the schema, or {:error, errors} if there are validation errors.

Parameters

  • schema: The schema definition map.
  • data: The data map to be validated.
  • opts: Options for validation.

Options

  • :mode - Validation mode. Can be :strict (default) or :permissive.
    • :strict - Only fields defined in the schema are returned.
    • :permissive - All fields from the input data are preserved.

Examples

schema = %{name: :string, age: :integer}
data = %{name: "John", age: 30, extra: "field"}

# Strict mode (default)
Peri.validate(schema, data)
# => {:ok, %{name: "John", age: 30}}

# Permissive mode
Peri.validate(schema, data, mode: :permissive)
# => {:ok, %{name: "John", age: 30, extra: "field"}}

validate_schema(schema)

Validates a schema definition to ensure it adheres to the expected structure and types.

This function can handle both simple and complex schema definitions, including nested schemas, custom validation functions, and various type constraints.

Parameters

  • schema - The schema definition to be validated. It can be a map or a keyword list representing the schema.

Returns

  • {:ok, schema} - If the schema is valid, returns the original schema.
  • {:error, errors} - If the schema is invalid, returns an error tuple with detailed error information.

Examples

Validating a simple schema:

  schema = %{
    name: :string,
    age: :integer,
    email: {:required, :string}
  }
  assert {:ok, ^schema} = validate_schema(schema)

Validating a nested schema:

  schema = %{
    user: %{
      name: :string,
      profile: %{
        age: {:required, :integer},
        email: {:required, :string}
      }
    }
  }
  assert {:ok, ^schema} = validate_schema(schema)

Handling invalid schema definition:

  schema = %{
    name: :str,
    age: :integer,
    email: {:required, :string}
  }
  assert {:error, _errors} = validate_schema(schema)

walk(schema, fun)

@spec walk(schema(), Peri.Walker.walker_fun()) :: schema()

Depth-first rewrite of a schema tree.

The callback is invoked on every subtree (pre-order). It must return either {:cont, new_node} to replace the node and continue, or :drop to remove it (only valid for values inside a map or keyword schema).

Building block for transforms like "make every field optional" or "strip internal-only fields from a public DTO". See Peri.Walker for details.

Examples

iex> schema = %{name: {:required, :string}, age: {:required, :integer}}
iex> Peri.walk(schema, fn
...>   {:required, t} -> {:cont, t}
...>   other -> {:cont, other}
...> end)
%{name: :string, age: :integer}