path: root/doc/src/manual/cowboy_rest.asciidoc

= cowboy_rest(3)

== Name

cowboy_rest - REST handlers

== Description

The module `cowboy_rest` implements the HTTP state machine.

Implementing REST handlers is not enough to provide a REST
interface; this interface must also follow the REST
constraints including HATEOAS (hypermedia as the engine
of application state).

== Callbacks

REST handlers implement the following interface:

[source,erlang]
----
init(Req, State)
    -> {cowboy_rest, Req, State}

Callback(Req, State)
    -> {Result, Req, State}
     | {stop, Req, State}
     | {{switch_handler, Module}, Req, State}
     | {{switch_handler, Module, Opts}, Req, State}

terminate(Reason, Req, State) -> ok  %% optional

Req    :: cowboy_req:req()
State  :: any()
Module :: module()
Opts   :: any()
Reason :: normal
        | {crash, error | exit | throw, any()}

Callback - see below
Result   - see below
Default  - see below
----

The `init/2` callback is common to all handlers. To switch
to the REST handler behavior, it must return `cowboy_rest`
as the first element of the tuple.

The `Callback/2` above represents all the REST-specific
callbacks. They are described in the following section
of this manual. REST-specific callbacks differ by their
name, semantics, result and default values. The default
value is the one used when the callback has not been
implemented. They otherwise all follow the same interface.

The `stop` tuple can be returned to stop REST processing.
If no response was sent before then, Cowboy will send a
'204 No Content'. The `stop` tuple can be returned from
any callback, excluding `expires`, `generate_etag`,
`last_modified` and `variances`.

A `switch_handler` tuple can be returned from these same
callbacks to stop REST processing and switch to a different
handler type. This is very useful to, for example, to stream
the response body.

The optional `terminate/3` callback will ultimately be called
with the reason for the termination of the handler.
Cowboy will terminate the process right after this. There
is no need to perform any cleanup in this callback.

The following terminate reasons are defined for loop handlers:

normal::
    The handler terminated normally.

{crash, Class, Reason}::
    A crash occurred in the handler. `Class` and `Reason` can be
    used to obtain more information about the crash. The function
    `erlang:get_stacktrace/0` can also be called to obtain the
    stacktrace of the process when the crash occurred.

== REST callbacks

=== AcceptCallback

[source,erlang]
----
AcceptCallback(Req, State) -> {Result, Req, State}

Result  :: true | {true, URI :: iodata()} | false}
Default  - crash
----

Process the request body.

This function should create or update the resource using the
request body.

For PUT requests, the body is a representation of the resource
that is being created or replaced.

For POST requests, the body is typically application-specific
instructions on how to process the request, but it may also
be a representation of the resource. When creating a new
resource with POST at a different location, return `{true, URI}`
with `URI` the new location.

For PATCH requests, the body is a series of instructions on
how to update the resource. Patch files or JSON Patch are
examples of such media types.

A response body may be sent. The appropriate media type, charset
and language for the response can be retrieved from the Req
object using the `media_type`, `charset` and `language` keys,
respectively. The body can be set using
link:man:cowboy_req:set_resp_body(3)[cowboy_req:set_resp_body(3)].

=== allowed_methods

[source,erlang]
----
allowed_methods(Req, State) -> {Result, Req, State}

Result  :: [binary()]  %% case sensitive
Default :: [<<"GET">>, <<"HEAD">>, <<"OPTIONS">>]
----

Return the list of allowed methods.

=== allow_missing_post

[source,erlang]
----
allow_missing_post(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether POST is allowed when the resource doesn't exist.

Returning `true` here means that a new resource will be
created. The URI for the newly created resource should be
returned from the `AcceptCallback` function.

=== charsets_provided

[source,erlang]
----
charsets_provided(Req, State) -> {Result, Req, State}

Result  :: [binary()]  %% lowercase; case insensitive
Default  - skip this step
----

Return the list of charsets the resource provides in order
of preference.

During content negotiation Cowboy will pick the most
appropriate charset for the client. The client advertises
charsets it prefers with the accept-charset header. When
that header is missing, Cowboy picks the first charset
from the resource.

// @todo We should explain precisely how charsets are picked.

Cowboy will add the negotiated `charset` to the Req object
after this step completes:

[source,erlang]
----
req() :: #{
    charset => binary()  %% lowercase; case insensitive
}
----

Note that Cowboy will only append the charset to the
content-type header of the response if the media type is text.

=== content_types_accepted

[source,erlang]
----
content_types_accepted(Req, State) -> {Result, Req, State}

Result     :: [{binary() | ParsedMime, AcceptCallback :: atom()}]
ParsedMime :: {Type :: binary(), SubType :: binary(), '*' | Params}
Params     :: [{Key :: binary(), Value :: binary()}]

Default     - crash
----

// @todo Case sensitivity of parsed mime content?

Return the list of media types the resource accepts in
order of preference.

A media type is made of different parts. The media type
`text/html;charset=utf-8` is of type `text`, subtype `html`
and has a single parameter `charset` with value `utf-8`.

// @todo Cowboy needs to ignore the boundary parameter for
// multipart, as we never want to match against it. Or allow
// ignoring specific parameters at the very least.

Cowboy will match the content-type request header against
the media types the server accepts and select the appropriate
callback. When that header is missing, or when the server does not
accept this media type, the request fails and an error response
is returned. Cowboy will execute the callback immediately otherwise.

// @todo We should explain precisely how media types are picked.

An empty parameters list `[]` means that no parameters will be
accepted. When any parameter is acceptable, the tuple form
should be used with parameters as the atom `'*'`.

Cowboy treats all parameters as case sensitive, except for the
`charset` parameter, which is known to be case insensitive. You
should therefore always provide the charset as a lowercase
binary string.

// @todo Maybe this should be in the user guide instead.
//This function will be called for POST, PUT and PATCH requests.
//It is entirely possible to define different callbacks for different
//methods if the handling of the request differs. Simply verify
//what the method is with `cowboy_req:method/1` and return a
//different list for each methods.

=== content_types_provided

[source,erlang]
----
content_types_provided(Req, State) -> {Result, Req, State}

Result     :: [{binary() | ParsedMime, ProvideCallback :: atom()}]
ParsedMime :: {Type :: binary(), SubType :: binary(), '*' | Params}
Params     :: [{Key :: binary(), Value :: binary()}]

Default     - [{{ <<"text">>, <<"html">>, '*'}, to_html}]
----

// @todo Case sensitivity of parsed mime content?
// @todo Space required for the time being: https://github.com/spf13/hugo/issues/2398

Return the list of media types the resource provides in
order of preference.

A media type is made of different parts. The media type
`text/html;charset=utf-8` is of type `text`, subtype `html`
and has a single parameter `charset` with value `utf-8`.

// @todo Cowboy needs to ignore the boundary parameter for
// multipart, as we never want to match against it. Or allow
// ignoring specific parameters at the very least.

During content negotiation Cowboy will pick the most appropriate
media type for the client. The client advertises media types it
prefers with the accept header. When that header is missing,
the content negotiation fails and an error response is returned.

The callback given for the selected media type will be called
at the end of the execution of GET and HEAD requests when a
representation must be sent to the client.

// @todo We should explain precisely how media types are picked.

An empty parameters list `[]` means that no parameters will be
accepted. When any parameter is acceptable, the tuple form
should be used with parameters as the atom `'*'`.

Cowboy treats all parameters as case sensitive, except for the
`charset` parameter, which is known to be case insensitive. You
should therefore always provide the charset as a lowercase
binary string.

When a charset is given in the media type parameters in the
accept header, Cowboy will do some additional checks to confirm
that it can use this charset. When the wildcard is used then Cowboy
will immediately call `charsets_provided` to confirm the charset
is acceptable. If the callback is undefined Cowboy assumes any
charset is acceptable. When the wildcard is not used and the charset
given in the accept header matches one of the configured media
types Cowboy will use that charset and skip the `charsets_provided`
step entirely.

Cowboy will add the negotiated `media_type` to the Req object
after this step completes:

[source,erlang]
----
req() :: #{
    media_type => ParsedMime
}
----

// @todo Case sensitivity of parsed mime content?

Cowboy may also add the negotiated `charset` to the Req object
after this step completes:

[source,erlang]
----
req() :: #{
    charset => binary()  %% lowercase; case insensitive
}
----

=== delete_completed

[source,erlang]
----
delete_completed(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether the resource has been fully deleted from the
system, including from any internal cache.

Returning `false` will result in a '202 Accepted' response
being sent instead of a '200 OK' or '204 No Content'.

=== delete_resource

[source,erlang]
----
delete_resource(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Delete the resource.

Cowboy will send an error response when this function
returns `false`.

=== expires

[source,erlang]
----
expires(Req, State) -> {Result, Req, State}

Result  :: calendar:datetime() | binary() | undefined
Default :: undefined
----

Return the resource's expiration date.

=== forbidden

[source,erlang]
----
forbidden(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether access to the resource is forbidden.

A '403 Forbidden' response will be sent if this
function returns `true`. This status code means that
access is forbidden regardless of authentication,
and that the request shouldn't be repeated.

=== generate_etag

[source,erlang]
----
generate_etag(Req, State) -> {Result, Req, State}

Result  :: binary() | {weak | strong, binary()}
Default  - no etag value
----

Return the entity tag of the resource.

When a binary is returned, the value is automatically
parsed to a tuple. The binary must be in the same
format as the etag header, including quotes.

=== is_authorized

[source,erlang]
----
is_authorized(Req, State) -> {Result, Req, State}

Result  :: true | {false, AuthHeader :: iodata()}
Default  - true
----

Return whether the user is authorized to perform the action.

This function should be used to perform any necessary
authentication of the user before attempting to perform
any action on the resource.

When authentication fails, the `AuthHeader` value will
be sent in the www-authenticate header for the
'401 Unauthorized' response.

=== is_conflict

[source,erlang]
----
is_conflict(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether the PUT request results in a conflict.

A '409 Conflict' response is sent when `true`.

=== known_methods

[source,erlang]
----
known_methods(Req, State) -> {Result, Req, State}

Result  :: [binary()]  %% case sensitive
Default :: [<<"GET">>, <<"HEAD">>, <<"POST">>, <<"PUT">>,
            <<"PATCH">>, <<"DELETE">>, <<"OPTIONS">>]
----

Return the list of known methods.

The full list of methods known by the server should be
returned, regardless of their use in the resource.

The default value lists the methods Cowboy knows and
implement in `cowboy_rest`.

=== languages_provided

[source,erlang]
----
languages_provided(Req, State) -> {Result, Req, State}

Result  :: [binary()]  %% lowercase; case insensitive
Default  - skip this step
----

Return the list of languages the resource provides in order
of preference.

During content negotiation Cowboy will pick the most
appropriate language for the client. The client advertises
languages it prefers with the accept-language header. When
that header is missing, Cowboy picks the first language
from the resource.

// @todo We should explain precisely how languages are picked.

Cowboy will add the negotiated `language` to the Req object
after this step completes:

[source,erlang]
----
req() :: #{
    language => binary()  %% lowercase; case insensitive
}
----

=== last_modified

[source,erlang]
----
last_modified(Req, State) -> {Result, Req, State}

Result  :: calendar:datetime()
Default  - no last modified value
----

Return the resource's last modification date.

This date will be used to test against the if-modified-since
and if-unmodified-since headers, and sent as the last-modified
header in the response to GET and HEAD requests.

=== malformed_request

[source,erlang]
----
malformed_request(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether the request is malformed.

A request is malformed when a component required by the
resource is invalid. This may include the query string
or individual headers. They should be parsed and validated
in this function. The body should not be read at this point.

=== moved_permanently

[source,erlang]
----
moved_permanently(Req, State) -> {Result, Req, State}

Result  :: {true, URI :: iodata()} | false
Default :: false
----

Return whether the resource was permanently moved, and
what its new location is.

=== moved_temporarily

[source,erlang]
----
moved_temporarily(Req, State) -> {Result, Req, State}

Result  :: {true, URI :: iodata()} | false
Default :: false
----

Return whether the resource was temporarily moved, and
what its new location is.

=== multiple_choices

[source,erlang]
----
multiple_choices(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether the client should engage in reactive
negotiation.

Return `true` when the server has multiple representations
of a resource, each with their specific identifier, but is
unable to determine which is best for the client. For
example an image might have different sizes and the server
is unable to determine the capabilities of the client.

When returning `true` the server should send a body with
links to the different representations. If the server has
a preferred representation it can send its link inside a
location header.

=== options

[source,erlang]
----
options(Req, State) -> {ok, Req, State}
----

Respond to an OPTIONS request.

The response should inform the client the communication
options available for this resource. By default Cowboy
will send a '200 OK' response with the allow header set.

=== previously_existed

[source,erlang]
----
previously_existed(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether the resource existed previously.

=== ProvideCallback

[source,erlang]
----
ProvideCallback(Req, State) -> {Result, Req, State}

Result  :: cowboy_req:resp_body()
Default  - crash
----

Return the response body.

The response body can be provided either as the actual data
to be sent or a tuple indicating which file to send.

This function is called for both GET and HEAD requests. For
the latter the body is not sent, however.

// @todo Perhaps we can optimize HEAD requests and just
// allow calculating the length instead of returning the
// whole thing.

Note that there used to be a way to stream the response body.
It was temporarily removed and will be added back in a later
release.

// @todo Add a way to switch to loop handler for streaming the body.

=== resource_exists

[source,erlang]
----
resource_exists(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether the resource exists.

=== service_available

[source,erlang]
----
service_available(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether the service is available.

A '503 Service Unavailable' response will be sent when this
function returns `false`.

=== uri_too_long

[source,erlang]
----
uri_too_long(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: false
----

Return whether the requested URI is too long.

This function can be used to further restrict the length
of the URI for this specific resource.

=== valid_content_headers

[source,erlang]
----
valid_content_headers(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether the content headers are valid.

This callback can be used to reject requests that have
invalid content header values, for example an unsupported
content-encoding.

=== valid_entity_length

[source,erlang]
----
valid_entity_length(Req, State) -> {Result, Req, State}

Result  :: boolean()
Default :: true
----

Return whether the request body length is within acceptable boundaries.

A '413 Request Entity Too Large' response will be sent if this
function returns `false`.

=== variances

[source,erlang]
----
variances(Req, State) -> {Result, Req, State}

Result  :: [binary()]  %% case insensitive
Default :: []
----

Return the list of request headers that affect the
representation of the resource.

Cowboy automatically adds the accept, accept-charset and
accept-language headers when necessary. It's also useful
to note that some standard headers also do not need to be
listed here, like the authorization header.

== Changelog

* *2.1*: The `switch_handler` return value was added.
* *1.0*: Behavior introduced.

== See also

link:man:cowboy(7)[cowboy(7)],
link:man:cowboy_handler(3)[cowboy_handler(3)]