API BEE
API Bee is a small client / spec for a particular style of JSON API.
These APIs must
- Expose resource collections as paginated lists, with entries and paging properties.
- Resource entities in collections must include an 'href' property pointing to the individual resource, so as to provide a degree of discoverability.
A single resource might look like:
{
'name': 'Foo',
'description': 'foo resoruce',
'href': 'http://api.myservice.com/resources/foo'
}
A resource collection looks like:
{
'href': 'http://api.myservice.com/resources',
'total_entries': 100,
'page': 1,
'per_page': 10,
'entries': [
{
'name': 'Foo',
'description': 'foo resoruce',
'href': 'http://api.myservice.com/resources/foo'
},
{
'name': 'Bar',
'description': 'bar resoruce',
'href': 'http://api.myservice.com/resources/bar'
},
...
]
}
Collection resources must include the fields 'href', 'total_entries', 'page', 'per_page' and 'entries'. This allows clients to paginate and fetch more pages.
Adapters
It is up to individual adapters to talk to different services, handle auth, etc. An adapter must at least implement 'get' for read-only APIs.
class ApiBee::Adapters::Special
def get(path, = {})
# fetch JSON from remote API, passing pagination options if available
end
end
ApiBee wraps your adapter and makes results lazily-loaded. That is, actual requests won't be made until accessing attributes or iterating the data set.
Use it:
api = ApiBee.setup :special, optional_custom_data
# No actual request made
resources = api.get('/my/resources')
# Requests once under the hood so you can iterate
resources.each do |r|
r[:name]
end
Lazy-loading
If an object in a response has a 'href' attribute, it will be used to fetch more data if you ask for an attribute currently not in the object.
# JSON Dataset
{
'user': {
'name': 'Ismael',
'href': 'http://api.com/users/ismael'
}
}
# Instantiate object
user = api.get('/user')
# No extra request made
user[:name]
# Extra request to http://api.com/users/ismael to fetch more user data
user[:last_name]
This works for objects in collections too.
# JSON collection
{
'users': {
'total_entries': 100,
'page': 1,
'per_page': 2,
'href': 'http://api.com/users',
'entries': [
{
'name': 'Ismael',
'href': 'http://api.com/users/ismael'
},
{
'name': 'John',
'href': 'http://api.com/users/john'
}
]
}
}
# Instantiate collection
users = api.get('/users') # no request yet
users.total_entries # => 100 # request made
users.size # => 2, current page
users.each ... # iterate current page
users.current_page # => 1
users.has_next_page? # => true
users.next_page # => 2
# Access entry. No request made
ismael = users.first
ismael[:name] # => 'ismael'
ismael[:last_name'] #=> request made to http://api.com/users/ismael
Per instance configuration
You can configure some variables on a per-instance basis. To configure the attribute name used to access a resource's URL (defaults to :href):
api = ApiBee.setup(MyCustomAdapter) do |config|
config.uri_property_name = :uri # use :uri instead
end
Adapter-wide configuration
You can declare configuration in the adapter definition itself, too. Just define the config_api_bee class method in your adapter:
class MyCustomAdapter
def get(path, )
# ...
end
def self.config_api_bee(config)
config.uri_property_name = :uri
end
end
api = ApiBee.setup(MyCustomAdapter) # instance is configured correctly
Delegate to adapter
Lazy-loading and paginating resources is great for GET requests, but you might want to still use your adapter's other methods.
api = ApiBee.setup(MyCustomAdapter) do |config|
config.expose :delete, :post
end
# This still wraps your adapter's get() method and adds lazy-loading and pagination
api.get('/products').first[:title]
# This delegates directly to MyCustomAdapter#post()
api.post('/products', :title => 'Foo', :price => 100.0)
finding a single resource
There's a special get_one method that you can call on lists. It delegates to the adapter and it's useful for finding a single resource in the context of a paginated list.
resources = api.get('/my/resources')
resource = resources.get_one('foobar')
That delegates to Adapter#get_one passing 2 arguments: the list's href and the passed name or identifier, so:
class ApiBee::Adapters::Special
# ...
def get_one(href, id)
get "#{href}/#{id}"
end
end
Hash adapter
ApiBee ships with an in-memory Hash adapter so it can be use with test/local data (for example a YAML file).
api = ApiBee.setup(:hash, YAML.load_file('./my_data.yml'))
products = api.get('/products') # => ApiBee::Node::List
products.first # => ApiBee::Node::Single
products.each() # iterate current page
products.current_page # => 1
products.paginate(:page => 2, :per_page => 4) # => ApiBee::Node::List # Next page
products.has_next_page? # => false
products.has_prev_page? # => true
products.prev_page # => 1
Examples
See examples/github_api.rb for an adapter that paginates Github's API by decorating it's results with ApiBee's required pagination properties