# `Plug.Parsers`
[🔗](https://github.com/elixir-plug/plug/blob/v1.20.1/lib/plug/parsers.ex#L1)

A plug for parsing the request body.

It invokes a list of `:parsers`, which are activated based on the
request content-type. Custom parsers are also supported by defining
a module that implements the behaviour defined by this module.

Once a connection goes through this plug, it will have `:body_params`
set to the map of params parsed by one of the parsers listed in
`:parsers` and `:params` set to the result of merging the `:body_params`
and `:query_params`. In case `:query_params` have not yet been parsed,
`Plug.Conn.fetch_query_params/2` is automatically invoked.

This plug will raise `Plug.Parsers.UnsupportedMediaTypeError` by default if
the request cannot be parsed by any of the given types and the MIME type has
not been explicitly accepted with the `:pass` option.

`Plug.Parsers.RequestTooLargeError` will be raised if the request goes over
the given limit. The default length is 8MB and it can be customized by passing
the `:length` option to the Plug. `:read_timeout` and `:read_length`, as
described by `Plug.Conn.read_body/2`, are also supported.

Parsers may raise a `Plug.Parsers.ParseError` if the request has a malformed
body.

This plug only parses the body if the request method is one of the following:

  * `POST`
  * `PUT`
  * `PATCH`
  * `DELETE`

For requests with a different request method, this plug will only fetch the
query params.

## Options

  * `:parsers` - a list of modules or atoms of built-in parsers to be
    invoked for parsing. These modules need to implement the behaviour
    outlined in this module.

  * `:pass` - an optional list of MIME type strings that are allowed
    to pass through. Any mime not handled by a parser and not explicitly
    listed in `:pass` will `raise UnsupportedMediaTypeError`. For example:

      * `["*/*"]` - never raises
      * `["text/html", "application/*"]` - doesn't raise for those values
      * `[]` - always raises (default)

  * `:query_string_length` - the maximum allowed size for query strings

  * `:validate_utf8` - boolean that tells whether or not we want to
      validate that parsed binaries are utf8 strings.

  * `:body_reader` - an optional replacement (or wrapper) for
    `Plug.Conn.read_body/2` to provide a function that gives access to the
    raw body before it is parsed and discarded. It is in the standard format
    of `{Module, :function, [args]}` (MFA) and defaults to
    `{Plug.Conn, :read_body, []}`. Note that this option is not used by
    `Plug.Parsers.MULTIPART` which relies instead on other functions defined
    in `Plug.Conn`.

All other options given to this Plug are forwarded to the parsers.

## Examples

    plug Plug.Parsers,
         parsers: [:urlencoded, :multipart],
         pass: ["text/*"]

Any other option given to Plug.Parsers is forwarded to the underlying
parsers. Therefore, you can use a JSON parser and pass the `:json_decoder`
option at the root:

    plug Plug.Parsers,
         parsers: [:urlencoded, :json],
         json_decoder: Jason

Or directly to the parser itself:

    plug Plug.Parsers,
         parsers: [:urlencoded, {:json, json_decoder: Jason}]

It is also possible to pass the `:json_decoder` as a `{module, function, args}` tuple,
useful for passing options to the JSON decoder:

    plug Plug.Parsers,
         parsers: [:json],
         json_decoder: {Jason, :decode!, [[floats: :decimals]]}

A common set of shared options given to Plug.Parsers is `:length`,
`:read_length` and `:read_timeout`, which customizes the maximum
request length you want to accept. For example, to support file
uploads, you can do:

    plug Plug.Parsers,
         parsers: [:urlencoded, :multipart],
         length: 20_000_000

However, the above will increase the maximum length of all request
types. If you want to increase the limit only for multipart requests
(which is typically the ones used for file uploads), you can do:

    plug Plug.Parsers,
         parsers: [
           :urlencoded,
           {:multipart, length: 20_000_000} # Increase to 20MB max upload
         ]

## Built-in parsers

Plug ships with the following parsers:

  * `Plug.Parsers.URLENCODED` - parses `application/x-www-form-urlencoded`
    requests (can be used as `:urlencoded` as well in the `:parsers` option)
  * `Plug.Parsers.MULTIPART` - parses `multipart/form-data` and
    `multipart/mixed` requests (can be used as `:multipart` as well in the
    `:parsers` option)
  * `Plug.Parsers.JSON` - parses `application/json` requests with the given
    `:json_decoder` (can be used as `:json` as well in the `:parsers` option)

## File handling

If a file is uploaded via any of the parsers, Plug will
stream the uploaded contents to a file in a temporary directory in order to
avoid loading the whole file into memory. For such, the `:plug` application
needs to be started in order for file uploads to work. More details on how the
uploaded file is handled can be found in the documentation for `Plug.Upload`.

When a file is uploaded, the request parameter that identifies that file will
be a `Plug.Upload` struct with information about the uploaded file (e.g.
filename and content type) and about where the file is stored.

The temporary directory where files are streamed to can be customized by
setting the `PLUG_TMPDIR` environment variable on the host system. If
`PLUG_TMPDIR` isn't set, Plug will look at some environment
variables which usually hold the value of the system's temporary directory
(like `TMPDIR` or `TMP`). If no value is found in any of those variables,
`/tmp` is used as a default.

## Custom body reader

Sometimes you may want to customize how a parser reads the body from the
connection. For example, you may want to cache the body to perform verification
later, such as HTTP Signature Verification. This can be achieved with a custom
body reader that would read the body and store it in the connection, such as:

    defmodule CacheBodyReader do
      def read_body(conn, opts) do
        with {:ok, body, conn} <- Plug.Conn.read_body(conn, opts) do
          conn = update_in(conn.assigns[:raw_body], &[body | &1 || []])
          {:ok, body, conn}
        end
      end
    end

which could then be set as:

    plug Plug.Parsers,
      parsers: [:urlencoded, :json],
      pass: ["text/*"],
      body_reader: {CacheBodyReader, :read_body, []},
      json_decoder: Jason

# `init`

```elixir
@callback init(opts :: Keyword.t()) :: Plug.opts()
```

# `parse`

```elixir
@callback parse(
  conn :: Plug.Conn.t(),
  type :: binary(),
  subtype :: binary(),
  params :: Plug.Conn.Utils.params(),
  opts :: Plug.opts()
) ::
  {:ok, Plug.Conn.params(), Plug.Conn.t()}
  | {:error, :too_large, Plug.Conn.t()}
  | {:next, Plug.Conn.t()}
```

Attempts to parse the connection's request body given the content-type type,
subtype, and its parameters.

The arguments are:

  * the `Plug.Conn` connection
  * `type`, the content-type type (e.g., `"x-sample"` for the
    `"x-sample/json"` content-type)
  * `subtype`, the content-type subtype (e.g., `"json"` for the
    `"x-sample/json"` content-type)
  * `params`, the content-type parameters (e.g., `%{"foo" => "bar"}`
    for the `"text/plain; foo=bar"` content-type)

This function should return:

  * `{:ok, body_params, conn}` if the parser is able to handle the given
    content-type; `body_params` should be a map
  * `{:next, conn}` if the next parser should be invoked
  * `{:error, :too_large, conn}` if the request goes over the given limit

---

*Consult [api-reference.md](api-reference.md) for complete listing*