Content API Scala Client

A Scala client for the Guardian's Content API.

Setup

Add the following line to your SBT build definition, and set the version number to be the latest from the releases page:

libraryDependencies += "com.gu" %% "content-api-client-default" % "x.y.z"

Please note;

as of version 7.0, the content api scala client no longer supports java 7.
as of version 17.24.0 the content api scala client only supports scala 2.12+

If you don't have an API key, go to open-platform.theguardian.com/access/ to get one. You will then need to create a new instance of the client and set the key:

val client = new GuardianContentClient("your-api-key")

Setup with custom Http layer

As of version 12.0, the core module does not provide an http implementation. This is to accomodate use cases where people want to use their existing infrastructure, rather than relying on an extra dependency on OkHttp (the client used in the default module above). First, add the following line to your SBT definition:

libraryDependencies += "com.gu" %% "content-api-client" % "x.y"

Then, create your own client by extending the ContentApiClient trait and implementing the get method, e.g. using Play's ScalaWS client library

Note that as of version 17.0, ContentApiClient no longer enforces backoff strategy. We have decoupled the client from the retry-backoff logic (More on this, later in the README) A sample implementation is shown below and more examples can be found later in this readme.

import play.api.libs.ws.WSClient

class ContentApiClient(ws: WSClient) extends ContentApiClient
  def get(url: String, headers: Map[String, String])(implicit context: ExecutionContext): Future[HttpResponse] =
    ws.url(url).withHttpHeaders(headers: _*).get.map(r => HttpResponse(r.bodyAsBytes, r.status, r.statusText))
}

Usage

There are then four different types of query that can be performed: for a single item, or to filter through content, tags, or sections. You make a request of the Content API by creating a query and then using the client to get a response, which will come back in a Future.

Use these imports for the following code samples (substituting your own execution context for real code):

import com.gu.contentapi.client.GuardianContentClient
import scala.concurrent.ExecutionContext.Implicits.global

Single item

Every item on http://www.theguardian.com/ can be retrieved on the same path at https://content.guardianapis.com/. They can be either content items, tags, or sections. For example:

// query for a single content item and print its web title
val itemQuery = ContentApiLogic.item("commentisfree/2013/jan/16/vegans-stomach-unpalatable-truth-quinoa")
client.getResponse(itemQuery).foreach { itemResponse =>
  println(itemResponse.content.get.webTitle)
}

// print web title for a tag
val tagQuery = ContentApiLogic.item("music/metal")
client.getResponse(tagQuery).foreach { tagResponse =>
  println(tagResponse.tag.get.webTitle)
}

// print web title for a section
val sectionQuery = ContentApiLogic.item("environment")
client.getResponse(sectionQuery).foreach { sectionResponse =>
  println(sectionResponse.section.get.webTitle)
}

Individual content items contain information not available from the /search endpoint described below. For example:

// print the body of a given content item
val itemBodyQuery = ContentApiLogic.item("politics/2014/sep/15/putin-bad-as-stalin-former-defence-secretary")
  .showFields("body")
client.getResponse(itemBodyQuery) map { response =>
  for (fields <- response.content.get.fields) println(fields.body)
}

// print the web title of every tag a content item has
val itemWebTitleQuery = ContentApiLogic.item("environment/2014/sep/14/invest-in-monitoring-and-tagging-sharks-to-prevent-attacks")
  .showTags("all")
client.getResponse(itemWebTitleQuery) map { response =>
  for (tag <- response.content.get.tags) println(tag.webTitle)
}

// print the web title of the most viewed content items from the world section
val mostViewedTitleQuery = ContentApiLogic.item("world").showMostViewed()
client.getResponse(mostViewedTitleQuery) map { response =>
  for (result <- response.mostViewed.get) println(result.webTitle)
}

Content

Filtering or searching for multiple content items happens at https://content.guardianapis.com/search. For example:

// print the total number of content items
val allContentSearch = ContentApiLogic.search
client.getResponse(allContentSearch) map { response =>
  println(response.total)
}

// print the web titles of the 15 most recent content items
val lastFifteenSearch = ContentApiLogic.search.pageSize(15)
client.getResponse(lastFifteenSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web titles of the 10 most recent content items matching a search term
val toastSearch = ContentApiLogic.search.q("cheese on toast")
client.getResponse(toastSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web titles of the 10 (default page size) most recent content items with certain tags
val tagSearch = ContentApiLogic.search.tag("lifeandstyle/cheese,type/gallery")
client.getResponse(tagSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web titles of the 10 most recent content items in the world section
val sectionSearch = ContentApiLogic.search.section("world")
client.getResponse(sectionSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web titles of the last 10 content items published a week ago
import java.time.temporal.ChronoUnit
import java.time.Instant
val timeSearch = ContentApiLogic.search.toDate(Instant.now().minus(7, ChronoUnit.DAYS))
client.getResponse(timeSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web titles of the last 10 content items published whose type is article
val typeSearch = ContentApiLogic.search.contentType("article")
client.getResponse(typeSearch) map { response =>
  for (result <- response.results) println(result.webTitle)
}

Sections

Filtering or searching for multiple sections happens at http://content.guardianapis.com/sections. For example:

// print the web title of each section
val allSectionsQuery = ContentApiLogic.sections
client.getResponse(allSectionsQuery) map { response =>
  for (result <- response.results) println(result.webTitle)
}

// print the web title of each section with 'network' in the title
val networkSectionsQuery = ContentApiLogic.sections.q("network")
client.getResponse(networkSectionsQuery) map { response =>
  for (result <- response.results) println(result.webTitle)
}

Editions

Filtering or searching for multiple Editions happens at http://content.guardianapis.com/editions. For example:

// print the apiUrl of each edition
val allEditionsQuery = ContentApiLogic.editions
client.getResponse(allEditionsQuery) map { response =>
  for (result <- response.results) println(result.apiUrl)
}

// print the webUrl of the edition with 'US' in edition field.
val usEditionsQuery = ContentApiLogic.editions.q("US")
client.getResponse(usEditionsQuery) map { response =>
  for (result <- response.results) println(result.webUrl)
}

Removed Content

This is removed in version 15.8

Pagination

The client allows you to paginate through results in the following ways:

paginate(query)(f) unfolds a query until there are no more page results to process. f is a pure function processing a CAPI response and paginate returns a list of processed responses (wrapped in a Future)
paginateAccum(query)(f, g) folds over the results and accumulates into a final result. f transforms a response into an accumulated result, g multiplies two results together
paginateFold(query)(f, m) folds over the results by accumulating a final result. f takes two parameters: a response and the accumulated result so far.

E.g. the following simply sums the number of results:

val result: Future[Int] = client.paginateFold(query)(0){ (r: SearchResponse, t: Int) => r.results.length + t }

Retrying recoverable errors (backoff strategies)

Sometimes the backend services that the client relies on can return HTTP failure results, and some of these are potentially recoverable within a relatively short period of time. Rather than immediately fail these requests by default as we have done previously, it is now possible to automatically retry those failures that may yield a successful result on a subsequent attempt. As of version 17.0 of the client you can apply a retry and a backoff strategy when instantiating a ContentApiClient instance by mixing in RetryableContentApiClient trait and providing a backoff strategy.

The following strategies are available;

  ContentApiBackoff.doublingStrategy                               
  ContentApiBackoff.exponentialStrategy
  ContentApiBackoff.multiplierStrategy 
  ContentApiBackoff.constantStrategy

To use these, it's simply a case of defining a duration and a number of retries you intend to make. And in the case of the multiplierStrategy, a double defining the factor by which to multiply the waiting time.

Some examples:

class ApiClient extends ContentApiClient {
  def get(url: String, headers: Map[String, String])(implicit context: ExecutionContext): Future[HttpResponse] = {
    // your implemenetation  
  }
}

val apiClient = new ApiClient with RetryableContentApiClient {
  override implicit val executor = ScheduledExecutor()  // or apply your own preferred executor

  // create a doubling backoff and retry strategy
  // we will wait for 250ms, 500ms, 1000ms, 2000ms  and then 4000ms while recoverable errors are encountered
  private val retryDuration = Duration(250L, TimeUnit.MILLISECONDS)
  private val retryAttempts = 5
  override val backoffStrategy = ContentApiBackoff.doublingStrategy(retryDuration, retryAttempts)

}

val apiClient = new ApiClient with RetryableContentApiClient {
  override implicit val executor = ScheduledExecutor()  // or apply your own preferred executor

  // create a constant backoff and retry strategy
  // we will allow up to three attempts, each separated by one second
  val retryDuration = Duration(1000L, TimeUnit.MILLISECONDS)
  val retryAttempts = 3
  override val backoffStrategy = ContentApiBackoff.constantStrategy(retryDuration, retryAttempts)
}

val apiClient = new ApiClient with RetryableContentApiClient {
  override implicit val executor = ScheduledExecutor()  // or apply your own preferred executor

  // create an exponential backoff strategy
  // the duration is multiplied by the next power of two on each attempt, so: 200ms, 400ms, 800ms, 1600ms and 3200ms
  // this means a maximum waiting time of 6.2 seconds if the error doesn't clear before then
  val retrySchedule: Duration = Duration(100L, TimeUnit.MILLISECONDS)
  val retryCount: Int = 5
  override val backoffStrategy = ContentApiBackoff.exponentialStrategy(retrySchedule, retryCount)
}

val apiClient = new ApiClient with RetryableContentApiClient {
  override implicit val executor = ScheduledExecutor()  // or apply your own preferred executor

  // create a multiplier backoff strategy
  // here, the duration is multiplied by the factor between each retry
  // with a factor of 3.0, and 3 retries we will wait: 250ms, 750ms, 2250ms
  val retrySchedule: Duration = Duration(250L, TimeUnit.MILLISECONDS)
  val retryCount: Int = 3
  val retryFactor: Double = 3.0
  override val backoffStrategy: Backoff = Backoff.multiplierStrategy(retrySchedule, retryCount, retryFactor)
}

Note that there are some minimum values that are enforced to prevent using excessively short waits, or attempts to circumvent the backoff strategy entirely by forcing immediate retries.

exponential requires a minimum delay of 100ms

doubling, multiple and constant strategies requires a minimum delay of 250ms

All strategies will ensure at least one retry attempt is configured.

No upper limits are enforced for retries, delays or multipliers, so exercise caution in case you accidentally configure your client to wait for a very long time.

Which response codes will be retried by the backoff strategy

Initially we have specified the following codes as potentially recoverable:

408 - Request Timeout
429 - Too Many Requests
503 - Service Unavailable
504 - Gateway Timeout
509 - Bandwidth Limit Exceeded

This list may be subject to change over time.

Default client - GuardianContentClient

We have provided a default client GuardianContentClient (okHttp as underlying HTTP layer) which has multi-variants for specifying backoff strategies

val client = new GuardianContentClient("YOUR API KEY HERE") // default vanilla client with no retry mechanism
val client = GuardianContentClient("YOUR API KEY HERE", yourBackoffstrategy) // default client with injectable backoff strategy
val client = GuardianContentClient("YOUR API KEY HERE") // default client with hardcoded doubling backoff strategy

Explore in the REPL

One easy way to get started with the client is to try it in the Scala REPL.

First clone this repo, then run sbt console from the client directory. This will start a REPL with a few useful things imported for you, so you can get started quickly:

scala> val client = GuardianContentClient("YOUR API KEY HERE")
client: com.gu.contentapi.client.GuardianContentClient = com.gu.contentapi.client.GuardianContentClient@3eb2a60

scala> val query = ContentApiLogic.search.showTags("all")
query: com.gu.contentapi.client.model.SearchQuery = SearchQuery(/search?show-tags=all)

scala> val response = Await.result(client.getResponse(query), 5.seconds)
SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".
SLF4J: Defaulting to no-operation (NOP) logger implementation
SLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.
response: com.gu.contentapi.client.model.v1.SearchResponse =
SearchResponse(ok,developer,1853997,1,10,1,185400,newest,List(Content(politics/blog/live/2016/may/17/corbyn-more-popular-than-ever-with-labour-members-poll-suggests-politics-live,Liveblog,Some(politics),Some(Politics),Some(CapiDateTime(1463487146000,2016-05-17T12:12:26.000Z)),EU referendum: Boris Johnson accuses Cameron of making UK look like 'banana republic' - Politics live,https://www.theguardian.com/politics/blog/live/2016/may/17/corbyn-more-popular-than-ever-with-labour-members-poll-suggests-politics-live,https://content.guardianapis.com/politics/blog/live/2016/may/17/corbyn-more-popular-than-ever-with-labour-members-poll-suggests-politics-live,None,List(Tag(politics/series/politics-live-with-andrew-sparrow,Series,Some(po...

scala> client.shutdown()

Serialisation / Deserialisation

To serialise the models returned from the client, you can use the JSON encoder in content-api-models. You will need to import circe as a dependency in build.sbt –

val circeVersion = "0.12.3"

libraryDependencies ++= Seq(
  "io.circe" %% "circe-core",
  "io.circe" %% "circe-generic",
  "io.circe" %% "circe-parser"
).map(_ % circeVersion)

... and then call the relevant method on the model.

import com.gu.contentapi.json.CirceEncoders._
import io.circe.syntax._

// ... later ...

val query = ContentApiClient.search.q("An example query")
val str = client.getResponse(query).map(_.asJson.toString)

Running Tests

Some tests require access to the API. See Setup for details on how to get one.

The key needs to be passed to the tests either as a system property or an environment variable.

$ sbt -DCAPI_TEST_KEY=your_api_key test
# or
$ env CAPI_TEST_KEY=your_api_key sbt test

Troubleshooting

If you have any problems you can speak to other developers at the [Guardian API talk group] (http://groups.google.com/group/guardian-api-talk).

guardian / content-api-scala-client 32.0.0