lhc

LHC

LHC is an extended/advanced HTTP client. Implementing basic http-communication enhancements like interceptors, exception handling, format handling, accessing response data, configuring endpoints and placeholders and fully compatible, RFC-compliant URL-template support.

LHC uses typhoeus for low level http communication.

See LHS, if you are searching for something more high level that can query webservices easily and provides an ActiveRecord like interface.

Quick start guide

  gem install lhc

or add it to your Gemfile:

  gem 'lhc'

use it like:

  response = LHC.get('http://datastore/v2/feedbacks')
  response.data.items[0]
  response.data.items[0].recommended
  response.body
  response.headers

Table of contents

• LHC
  • Quick start guide
  • Table of contents
  • Basic methods
  • Request
    • Formats
      • Default format
      • Unformatted requests
        • Upload with LHC
    • Parallel requests
    • Follow redirects
    • Transfer data through the request body
    • Request parameters
      • Array Parameter Encoding
    • Request URL encoding
    • Request URL-Templates
    • Request timeout
    • Request Agent
  • Response
    • Accessing response data
  • Exceptions
    • Custom error handling
    • Ignore certain errors
  • Configuration
    • Configuring endpoints
    • Configuring placeholders
  • Interceptors
    • Quick start: Configure/Enable Interceptors
    • Interceptors on local request level
    • Core Interceptors
      • Authentication Interceptor
        • Bearer Authentication
        • Basic Authentication
        • Reauthenticate
        • Bearer Authentication with client access token
      • Caching Interceptor
        • Options
      • Default Timeout Interceptor
        • Overwrite defaults
      • Logging Interceptor
        • Installation
        • What and how it logs
        • Configure
      • Monitoring Interceptor
        • Installation
        • Environment
        • What it tracks
        • Configure
      • Prometheus Interceptor
      • Retry Interceptor
        • Limit the amount of retries while making the request
        • Change the default maximum of retries of the retry interceptor
        • Retry all requests
      • Rollbar Interceptor
        • Forward additional parameters
      • Throttle
      • Zipkin
    • Create an interceptor from scratch
      • Interceptor callbacks
      • Interceptor request/response
      • Provide a response replacement through an interceptor
  • Testing
  • License

Basic methods

Available are get, post, put & delete.

Other methods are available using LHC.request(options).

Request

The request class handles the http request, implements the interceptor pattern, loads configured endpoints, generates urls from url-templates and raises exceptions for any response code that is not indicating success (2xx).

  response = LHC.request(url: 'http://local.ch', method: :options)

  response.request.response #<LHC::Response> the associated response.

  response.request.options #<Hash> the options used for creating the request.

  response.request.params # access request params

  response.request.headers # access request headers

  response.request.url #<String> URL that is used for doing the request

  response.request.method #<Symbol> provides the used http-method

Formats

You can use any of the basic methods in combination with a format like json:

LHC.json.get(options)

Currently supported formats: json, multipart, plain (for no formatting)

If formats are used, headers for Content-Type and Accept are enforced by LHC, but also http bodies are translated by LHC, so you can pass bodies as ruby objects:

LHC.json.post('http://slack', body: { text: 'Hi there' })
# Content-Type: application/json
# Accept: application/json
# Translates body to "{\"text\":\"Hi there\"}" before sending

Default format

If you use LHC's basic methods LHC.get, LHC.post etc. without any explicit format, JSON will be chosen as the default format.

Unformatted requests

In case you need to send requests without LHC formatting headers or the body, use plain:

LHC.plain.post('http://endpoint', body: { weird: 'format%s2xX' })

Upload with LHC

If you want to upload data with LHC, it's recommended to use the multipart format:

response = LHC.multipart.post('http://upload', body: { file })
response.headers['Location']
# Content-Type: multipart/form-data
# Leaves body unformatted

Parallel requests

If you pass an array of requests to LHC.request, it will perform those requests in parallel. You will get back an array of LHC::Response objects in the same order of the passed requests.

  options = []
  options << { url: 'http://datastore/v2/feedbacks' }
  options << { url: 'http://datastore/v2/content-ads/123/feedbacks' }
  responses = LHC.request(options)

LHC.request([request1, request2, request3])
# returns [response1, response2, response3]

Follow redirects

LHC.get('http://local.ch', followlocation: true)

Transfer data through the request body

Data that is transfered using the HTTP request body is transfered using the selected format, or the default json, so you need to provide it as a ruby object.

Also consider setting the http header for content-type or use one of the provided formats, like LHC.json.

  LHC.post('http://datastore/v2/feedbacks',
    body: feedback,
    headers: { 'Content-Type' => 'application/json' }
  )

Request parameters

When using LHC, try to pass params via params option. It's not recommended to build a url and attach the parameters yourself:

DO

LHC.get('http://local.ch', params: { q: 'Restaurant' })

DON'T

LHC.get('http://local.ch?q=Restaurant')

Array Parameter Encoding

LHC can encode array parameters in URLs in two ways. The default is :rack which generates URL parameters compatible with Rack and Rails.

LHC.get('http://local.ch', params: { q: [1, 2] })
# http://local.ch?q[]=1&q[]=2

Some Java-based apps expect their arrays in the :multi format:

LHC.get('http://local.ch', params: { q: [1, 2] }, params_encoding: :multi)
# http://local.ch?q=1&q=2

Request URL encoding

LHC, by default, encodes urls:

LHC.get('http://local.ch?q=some space')
# http://local.ch?q=some%20space

LHC.get('http://local.ch', params: { q: 'some space' })
# http://local.ch?q=some%20space

which can be disabled:

LHC.get('http://local.ch?q=some space', url_encoding: false)
# http://local.ch?q=some space

Request URL-Templates

Instead of using concrete urls you can also use url-templates that contain placeholders. This is especially handy for configuring an endpoint once and generate the url from the params when doing the request. Since version 7.0 url templates follow the RFC 6750.

  LHC.get('http://datastore/v2/feedbacks/{id}', params:{ id: 123 })
  # GET http://datastore/v2/feedbacks/123

You can also use URL templates, when configuring endpoints:

  LHC.configure do |c|
    c.endpoint(:find_feedback, 'http://datastore/v2/feedbacks/{id}')
  end
  
  LHC.get(:find_feedback, params:{ id: 123 }) # GET http://datastore/v2/feedbacks/123

If you miss to provide a parameter that is part of the url-template, it will raise an exception.

Request timeout

Working and configuring timeouts is important, to ensure your app stays alive when services you depend on start to get really slow...

LHC forwards two timeout options directly to typhoeus:

timeout (in seconds) - The maximum time in seconds that you allow the libcurl transfer operation to take. Normally, name lookups can take a considerable time and limiting operations to less than a few seconds risk aborting perfectly normal operations. This option may cause libcurl to use the SIGALRM signal to timeout system calls. connecttimeout (in seconds) - It should contain the maximum time in seconds that you allow the connection phase to the server to take. This only limits the connection phase, it has no impact once it has connected. Set to zero to switch to the default built-in connection timeout - 300 seconds.

LHC.get('http://local.ch', timeout: 5, connecttimeout: 1)

LHC provides a timeout interceptor that lets you apply default timeout values to all the requests that you are performig in your application.

Request Agent

LHC identifies itself towards outher services, using the User-Agent header.

User-Agent LHC (9.4.2) [https://github.com/local-ch/lhc]

If LHC is used in an Rails Application context, also the application name is added to the User-Agent header.

User-Agent LHC (9.4.2; MyRailsApplicationName) [https://github.com/local-ch/lhc]

Response

  response.request #<LHC::Request> the associated request.

  response.data #<OpenStruct> in case response body contains parsable JSON.
  response.data.something.nested

  response.body #<String>

  response.code #<Fixnum>

  response.headers #<Hash>

  response.time #<Fixnum> Provides response time in ms.

  response.timeout? #true|false

Accessing response data

The response data can be access with dot-notation and square-bracket notation. You can convert response data to open structs or json (if the response format is json).

  response = LHC.request(url: 'http://datastore/entry/1')
  response.data.as_open_struct #<OpenStruct name='local.ch'>
  response.data.as_json # { name: 'local.ch' }
  response.data.name # 'local.ch'
  response.data[:name] # 'local.ch'

You can also access response data directly through the response object (with square bracket notation only):

  LHC.json.get(url: 'http://datastore/entry/1')[:name]

Exceptions

Anything but a response code indicating success (2xx) raises an exception.

  LHC.get('localhost') # UnknownError: 0
  LHC.get('http://localhost:3000') # LHC::Timeout: 0

You can access the response object that was causing the error.

LHC.get('local.ch')
rescue => e
  e.response #<LHC:Response>
  e.response.code # 403
  e.response.timeout? # false
  Rails.logger.error e
  # LHC::UnknownError: get http://local.cac
  # Params: {:url=>"http://local.cac", :method=>:get}
  # Response Code: 0
  # <Response Body>

All errors that are raise by LHC inherit from LHC::Error. They are divided into LHC::ClientError, LHC::ServerError, LHC::Timeout and LHC::UnkownError and mapped according to the following status code.

400 => LHC::BadRequest
401 => LHC::Unauthorized
402 => LHC::PaymentRequired
403 => LHC::Forbidden
403 => LHC::Forbidden
404 => LHC::NotFound
405 => LHC::MethodNotAllowed
406 => LHC::NotAcceptable
407 => LHC::ProxyAuthenticationRequired
408 => LHC::RequestTimeout
409 => LHC::Conflict
410 => LHC::Gone
411 => LHC::LengthRequired
412 => LHC::PreconditionFailed
413 => LHC::RequestEntityTooLarge
414 => LHC::RequestUriToLong
415 => LHC::UnsupportedMediaType
416 => LHC::RequestedRangeNotSatisfiable
417 => LHC::ExpectationFailed
422 => LHC::UnprocessableEntity
423 => LHC::Locked
424 => LHC::FailedDependency
426 => LHC::UpgradeRequired

500 => LHC::InternalServerError
501 => LHC::NotImplemented
502 => LHC::BadGateway
503 => LHC::ServiceUnavailable
504 => LHC::GatewayTimeout
505 => LHC::HttpVersionNotSupported
507 => LHC::InsufficientStorage
510 => LHC::NotExtended

timeout? => LHC::Timeout

anything_else => LHC::UnknownError

Custom error handling

You can provide custom error handlers to handle errors happening during the request.

If a error handler is provided nothing is raised.

If your error handler returns anything else but nil it replaces the response body.

handler = ->(response){ do_something_with_response; return {name: 'unknown'} }
response = LHC.get('http://something', error_handler: handler)
response.data.name # 'unknown'

Ignore certain errors

As it's discouraged to rescue errors and then don't handle them (ruby styleguide)[https://github.com/bbatsov/ruby-style-guide#dont-hide-exceptions], but you often want to continue working with nil, LHC provides the ignored_errors option.

Errors listed in this option will not be raised and will leave the response.body and response.data to stay nil.

response = LHC.get('http://something', ignored_errors: [LHC::NotFound])

response.body # nil
response.data # nil
response.error_ignored? # true
response.request.error_ignored? # true

Configuration

If you want to configure LHC, do it on initialization (like in a Rails initializer, environment.rb or application.rb), otherwise you could run into the problem that certain configurations can only be set once.

You can use LHC.configure to prevent the initialization problem. Take care that you only use LHC.configure once, because it is actually reseting previously made configurations and applies the new once.

  LHC.configure do |c|
    c.placeholder :datastore, 'http://datastore'
    c.endpoint :feedbacks, '{+datastore}/feedbacks', params: { has_reviews: true }
    c.interceptors = [CachingInterceptor, MonitorInterceptor, TrackingIdInterceptor]
  end

Configuring endpoints

You can configure endpoints, for later use, by giving them a name, a url and some parameters (optional).

  LHC.configure do |c|
    c.endpoint(:feedbacks, 'http://datastore/v2/feedbacks', params: { has_reviews: true })
    c.endpoint(:find_feedback, 'http://datastore/v2/feedbacks/{id}')
  end

  LHC.get(:feedbacks) # GET http://datastore/v2/feedbacks
  LHC.get(:find_feedback, params:{ id: 123 }) # GET http://datastore/v2/feedbacks/123

Explicit request options override configured options.

  LHC.get(:feedbacks, params: { has_reviews: false }) # Overrides configured params

Configuring placeholders

You can configure global placeholders, that are used when generating urls from url-templates.

  LHC.configure do |c|
    c.placeholder(:datastore, 'http://datastore')
    c.endpoint(:feedbacks, '{+datastore}/feedbacks', { params: { has_reviews: true } })
  end
  
  LHC.get(:feedbacks) # http://datastore/v2/feedbacks

Interceptors

To monitor and manipulate the HTTP communication done with LHC, you can define interceptors that follow the (Inteceptor Pattern)[https://en.wikipedia.org/wiki/Interceptor_pattern]. There are some interceptors that are part of LHC already, so called Core Interceptors, that cover some basic usecases.

Quick start: Configure/Enable Interceptors

LHC.configure do |c|
  c.interceptors = [LHC::Auth, LHC::Caching, LHC::DefaultTimeout, LHC::Logging, LHC::Monitoring, LHC::Prometheus, LHC::Retry, LHC::Rollbar, LHC::Zipkin]
end

You can only set the list of global interceptors once and you can not alter it after you set it.

Interceptors on local request level

You can override the global list of interceptors on local request level:

  interceptors = LHC.config.interceptors
  interceptors -= [LHC::Caching] # remove caching
  interceptors += [LHC::Retry] # add retry
  LHC.request({url: 'http://local.ch', retry: 2, interceptors: interceptors})

  LHC.request({url: 'http://local.ch', interceptors: []}) # no interceptor for this request at all

Core Interceptors

Authentication Interceptor

Add the auth interceptor to your basic set of LHC interceptors.

  LHC.configure do |c|
    c.interceptors = [LHC::Auth]
  end

Bearer Authentication

  LHC.get('http://local.ch', auth: { bearer: -> { access_token } })

Adds the following header to the request:

  'Authorization': 'Bearer 123456'

Assuming the method access_token responds on runtime of the request with 123456.

Basic Authentication

  LHC.get('http://local.ch', auth: { basic: { username: 'steve', password: 'can' } })

Adds the following header to the request:

  'Authorization': 'Basic c3RldmU6Y2Fu'

Which is the base64 encoded credentials "username:password".

Reauthenticate

The current implementation can only offer reauthenticate for client access tokens. For this to work the following has to be given:

• You have configured and implemented LHC::Auth.refresh_client_token = -> { TokenRefreshUtil.client_access_token(true) } which when called will force a refresh of the token and return the new value. It is also expected that this implementation will handle invalidating caches if necessary.
• Your interceptors contain LHC::Auth and LHC::Retry, whereas LHC::Retry comes after LHC::Auth in the chain.

Bearer Authentication with client access token

Reauthentication will be initiated if:

• setup is correct
• response.success? is false and an LHC::Unauthorized was observed
• reauthentication wasn't already attempted once

If this is the case, this happens:

• refresh the client token, by calling refresh_client_token
• the authentication header will be updated with the new token
• LHC::Retry will be triggered by adding retry: { max: 1 } to the request options

Caching Interceptor

Add the cache interceptor to your basic set of LHC interceptors.

  LHC.configure do |c|
    c.interceptors = [LHC::Caching]
  end

You can configure your own cache (default Rails.cache) and logger (default Rails.logger):

  LHC::Caching.cache = ActiveSupport::Cache::MemoryStore.new
  LHC::Caching.logger = Logger.new(STDOUT)

Caching is not enabled by default, although you added it to your basic set of interceptors. If you want to have requests served/stored and stored in/from cache, you have to enable it by request.

  LHC.get('http://local.ch', cache: true)

You can also enable caching when configuring an endpoint in LHS.

  class Feedbacks < LHS::Service
    endpoint '{+datastore}/v2/feedbacks', cache: true
  end

Only GET requests are cached by default. If you want to cache any other request method, just configure it:

  LHC.get('http://local.ch', cache: { methods: [:get] })

Responses served from cache are marked as served from cache:

  response = LHC.get('http://local.ch', cache: true)
  response.from_cache? # true

Options

  LHC.get('http://local.ch', cache: { key: 'key' expires_in: 1.day, race_condition_ttl: 15.seconds, use: ActiveSupport::Cache::MemoryStore.new })

expires_in - lets the cache expires every X seconds.

key - Set the key that is used for caching by using the option. Every key is prefixed with LHC_CACHE(v1): .

race_condition_ttl - very useful in situations where a cache entry is used very frequently and is under heavy load. If a cache expires and due to heavy load several different processes will try to read data natively and then they all will try to write to cache. To avoid that case the first process to find an expired cache entry will bump the cache expiration time by the value set in cache_race_condition_ttl.

use - Set an explicit cache to be used for this request. If this option is missing LHC::Caching.cache is used.

Default Timeout Interceptor

Applies default timeout values to all requests made in an application, that uses LHC.

  LHC.configure do |c|
    c.interceptors = [LHC::DefaultTimeout]
  end

timeout default: 15 seconds connecttimeout default: 2 seconds

Overwrite defaults

LHC::DefaultTimeout.timeout = 5 # seconds
LHC::DefaultTimeout.connecttimeout = 3 # seconds

Logging Interceptor

The logging interceptor logs all requests done with LHC to the Rails logs.

Installation

  LHC.configure do |c|
    c.interceptors = [LHC::Logging]
  end

  LHC::Logging.logger = Rails.logger

What and how it logs

The logging Interceptor logs basic information about the request and the response:

LHC.get('http://local.ch')
# Before LHC request<70128730317500> GET http://local.ch at 2018-05-23T07:53:19+02:00 Params={} Headers={\"User-Agent\"=>\"Typhoeus - https://github.com/typhoeus/typhoeus\", \"Expect\"=>\"\"}
# After LHC response for request<70128730317500>: GET http://local.ch at 2018-05-23T07:53:28+02:00 Time=0ms URL=http://local.ch:80/

Configure

You can configure the logger beeing used by the logging interceptor:

LHC::Logging.logger = Another::Logger

Monitoring Interceptor

The monitoring interceptor reports all requests done with LHC to a given StatsD instance.

Installation

  LHC.configure do |c|
    c.interceptors = [LHC::Monitoring]
  end

You also have to configure statsd in order to have the monitoring interceptor report.

  LHC::Monitoring.statsd = <your-instance-of-statsd>

Environment

By default, the monitoring interceptor uses Rails.env to determine the environment. In case you want to configure that, use:

LHC::Monitoring.env = ENV['DEPLOYMENT_TYPE'] || Rails.env

What it tracks

It tracks request attempts with before_request and after_request (counts).

In case your workers/processes are getting killed due limited time constraints, you are able to detect deltas with relying on "before_request", and "after_request" counts:

  "lhc.<app_name>.<env>.<host>.<http_method>.before_request", 1
  "lhc.<app_name>.<env>.<host>.<http_method>.after_request", 1

In case of a successful response it reports the response code with a count and the response time with a gauge value.

  LHC.get('http://local.ch')

  "lhc.<app_name>.<env>.<host>.<http_method>.count", 1
  "lhc.<app_name>.<env>.<host>.<http_method>.200", 1
  "lhc.<app_name>.<env>.<host>.<http_method>.time", 43

Timeouts are also reported:

  "lhc.<app_name>.<env>.<host>.<http_method>.timeout", 1

All the dots in the host are getting replaced with underscore, because dot is the default separator in graphite.

Configure

It is possible to set the key for Monitoring Interceptor on per request basis:

  LHC.get('http://local.ch', monitoring_key: 'local_website')

  "local_website.count", 1
  "local_website.200", 1
  "local_website.time", 43

If you use this approach you need to add all namespaces (app, environment etc.) to the key on your own.

Prometheus Interceptor

Logs basic request/response information to prometheus.

  require 'prometheus/client'

  LHC.configure do |c|
    c.interceptors = [LHC::Prometheus]
  end
    
  LHC::Prometheus.client = Prometheus::Client
  LHC::Prometheus.namespace = 'web_location_app'

  LHC.get('http://local.ch')

• Creates a prometheus counter that receives additional meta information for: :code, :success and :timeout.

• Creates a prometheus histogram for response times in milliseconds.

Retry Interceptor

If you enable the retry interceptor, you can have LHC retry requests for you:

  LHC.configure do |c|
    c.interceptors = [LHC::Retry]
  end
  
  response = LHC.get('http://local.ch', retry: true)

It will try to retry the request up to 3 times (default) internally, before it passes the last response back, or raises an error for the last response.

Consider, that all other interceptors will run for every single retry.

Limit the amount of retries while making the request

  LHC.get('http://local.ch', retry: { max: 1 })

Change the default maximum of retries of the retry interceptor

  LHC::Retry.max = 3

Retry all requests

If you want to retry all requests made from your application, you just need to configure it globally:

  LHC::Retry.all = true
  configuration.interceptors = [LHC::Retry]

Do not retry certain response codes

If you do not want to retry based on certain response codes, use retry in combination with explicit ignore_errors:

  LHC.get('http://local.ch', ignore_errors: [LHC::NotFound], retry: { max: 1 })

Or if you use LHC::Retry.all:

LHC.get('http://local.ch', ignore_errors: [LHC::NotFound])

Rollbar Interceptor

Forward errors to rollbar when exceptions occur during http requests.

  LHC.configure do |c|
    c.interceptors = [LHC::Rollbar]
  end

  LHC.get('http://local.ch')

If it raises, it forwards the request and response object to rollbar, which contain all necessary data.

Forward additional parameters

  LHC.get('http://local.ch', rollbar: { tracking_key: 'this particular request' })

Throttle

The throttle interceptor allows you to raise an exception if a predefined quota of a provider request limit is reached in advance.

  LHC.configure do |c|
    c.interceptors = [LHC::Throttle]
  end

options = { 
  throttle: {
    track: true, # enables tracking of current limit/remaining requests of rate-limiting
    break: '80%', # quota in percent after which errors are raised. Percentage symbol is optional, values will be converted to integer (e.g. '23.5' will become 23)
    provider: 'local.ch', # name of the provider under which throttling tracking is aggregated,
    limit: { header: 'Rate-Limit-Limit' }, # either a hard-coded integer, or a hash pointing at the response header containing the limit value
    remaining: { header: 'Rate-Limit-Remaining' }, # a hash pointing at the response header containing the current amount of remaining requests
    expires: { header: 'Rate-Limit-Reset' } # a hash pointing at the response header containing the timestamp when the quota will reset
  } 
}

LHC.get('http://local.ch', options)
# { headers: { 'Rate-Limit-Limit' => 100, 'Rate-Limit-Remaining' => 19 } }

LHC.get('http://local.ch', options)
# raises LHC::Throttle::OutOfQuota: Reached predefined quota for local.ch

Zipkin

** Zipkin 0.33 breaks our current implementation of the Zipkin interceptor **

Zipkin is a distributed tracing system. It helps gather timing data needed to troubleshoot latency problems in microservice architectures Zipkin Distributed Tracing.

Add the zipkin interceptor to your basic set of LHC interceptors.

  LHC.configure do |c|
    c.interceptors = [LHC::Zipkin]
  end

The following configuration needs to happen in the application that wants to run this interceptor:

1. Add gem 'zipkin-tracer', '< 0.33.0' to your Gemfile.
2. Add the necessary Rack middleware and configuration

config.middleware.use ZipkinTracer::RackHandler, {
  service_name: 'service-name', # name your service will be known as in zipkin
  service_port: 80, # the port information that is sent along the trace
  json_api_host: 'http://zipkin-collector', # the zipkin endpoint
  sample_rate: 1 # sample rate, where 1 = 100% of all requests, and 0.1 is 10% of all requests
}

Create an interceptor from scratch

  class TrackingIdInterceptor < LHC::Interceptor

    def before_request
      request.params[:tid] = 123
    end
  end

  LHC.configure do |c|
    c.interceptors = [TrackingIdInterceptor]
  end

Interceptor callbacks

before_raw_request is called before the raw typhoeus request is prepared/created.

before_request is called when the request is prepared and about to be executed.

after_request is called after request was started.

before_response is called when response started to arrive.

after_response is called after the response arrived completely.

Interceptor request/response

Every interceptor can directly access their instance request or response.

Provide a response replacement through an interceptor

Inside an interceptor, you are able to provide a response, rather then doing a real request. This is useful for implementing e.g. caching.

class LHC::Cache < LHC::Interceptor

  def before_request(request)
    cached_response = Rails.cache.fetch(request.url)
    return LHC::Response.new(cached_response) if cached_response
  end
end

Take care that having more than one interceptor trying to return a response will cause an exception. You can access the request.response to identify if a response was already provided by another interceptor.

  class RemoteCacheInterceptor < LHC::Interceptor

    def before_request(request)
      return unless request.response.nil?
      return LHC::Response.new(remote_cache)
    end
  end

Testing

When writing tests for your application when using LHC, please make sure you require the lhc rspec test helper:

# spec/spec_helper.rb

require 'lhc/rspec'

License

GNU General Public License Version 3.