building rest api-server with cuba
MIT License
this is a (hopefully) small collections of plugin, rack-middleware and datamapper extensions to build rest API servers with cuba. most will work without datamapper - the Cuba::Rest::Rack::ErrorRack
needs ERROR_2_STATUS
to be setup as needed. probably the dependency to datamapper is still there.
get!
from datamapper in case the object does not existoptimistic_get/optimistic_get!
from ixtlan-datamappererrors
and is not emptyCuba::Rest::Response
instance per thread, i.e. even with composition you will have only ONE response object (unlike with upstream Cuba
)Cuba
)Cuba
)just require 'cuba_rest'
. in case you want to customize or cherry pick things have a look at lib/cuba/rest.rb to see how.
the Cuba::Rest::Plugin
has two methods for this read( FilterClass )
and write( object )
/write( object, context )
/write( object, SerializerClass )
.
the FilterClass
must of type Ixtlan::Babel::HashFilter
, an instance receives the payload from the client and ensures you get only the data you need for your business logic. the Content-Type on the request must be set correctly to parse the given payload.
the write
method uses the Ixtlan::Babel::Factory
to produces a serializer. if the object
responds to errors
and those errors
are not empty then *precondition failed status is used along with payload of the serialized errors (Hash).
a POST request will use status 201 (created) along with a Location header (object must respond to id
)
for a DELETE request a status 204 (no content) is used.
there are some shortcuts to write up the cuba part:
on_path :id do |id|
on get do
...
end
end
this is the same as
on :id, empty_path do |id|
on get do
...
end
res.status 406 # method not allowed
end
where the empty_path is a matcher that all path was consumed.
to help there is a method modified_since
in Cuba
to retrieve the request header IF_MODIFIED_SINCE and a method last_modified
in Cuba::Rest::Response
(comes as default with cuba-rest) to set it for the response.
beside the extra matcher empty_path
, the on_path
method and the modifed_since
method, there are a couple of convienent methods to parse request parameters
there are a list of methods setting cache control headers on the response
expired by ArticleHttpCaching. more about caching is in increasing-application-performance-with-http-cache-headers.
other convenient methods are
it also default the response status to 404 unless there is some actual content. it will respect manual set response status.
per default Cuba::Rest::Rack::MimeRack
consumes application/json and produces application/json and text/x-yaml as requested by the request. other media types (like XML) need to be configured and Hash
and Array
need to have respective method (like to_xml
) to work.
the request either can set the HTTP_ACCEPT to the desired media type or use the file extension .json or .yaml to trigger the right media. note that the file extension will NOT be passed on the Cuba
matchers !
the main focus is JSON for input and output as well YAML as output since it produces nice on screen presentation.
this rack middleware is meant to rescue from some exceptions/errors and produces given http status codes. per default it catches
DataMapper::ObjectNotFoundError
gives 404 (not found)Ixtlan::DataMapper::StaleObjectException
gives 409 (conflict)just add your exceptions and status to Cuba::Rest::Rack::ErrorRack::ERROR_2_STATUS
it also logs a simple statement to console for each caught error. to produces a backtrace on the console log switch the logger to
require 'cuba/rest/rack/error_rack'
CubaRest::Rack::Logger = CubaRest::Rack::DebugErrorLogger
this must happen before requiring cuba-rest. NOTE bundler with its auto-require feature has to be off for cuba-rest
it is fully working example with test and part of the test suite of cuba-rest.
the example uses composition of Cuba
objects to have a similar directory structure as the API directory. i.e. all the 'users' stuff is in /users
and all the 'groups' in /groups
.
the actual application (business logic) is inside the models (lib/models.rb
). the cuba and filter/serializers deals with HTTP layer of the application. when starting the lib/console.rb
only the application gets loaded - no HTTP stuff.
it comes with Procfile
and Gemfile
for heroku. dito the database setup works with heroku as well.
there are also a few convinient commands in example/bin to start the server, seed the database, start an irb session with the loaded models, or run the (cu)test. all those scripts use bundler to setup the load_path to use. bundler is only used when the bin/load_path is out of date.
note: this is basically my personal template for starting a new app ;)
git checkout -b my-new-feature
)git commit -am 'Added some feature'
)git push origin my-new-feature
)enjoy :)