A library for Simple & Efficient data access in Scala and Scala.js
- Installation
- Remote data
- Define your data sources
- Creating a runtime
- Creating and running a fetch
- Batching
- Parallelism
- Deduplication & Caching
Add the following dependency to your project's build file.
For Scala 2.12.x through 3.x:
"com.47deg" %% "fetch" % "3.1.1"
Or, if using Scala.js (1.8.x):
"com.47deg" %%% "fetch" % "3.1.1"
Fetch is a library for making access to data both simple and efficient. Fetch is especially useful when querying data that has a latency cost, such as databases or web services.
To tell Fetch how to get the data you want, you must implement the DataSource
typeclass. Data sources have fetch
and batch
methods that define how to fetch such a piece of data.
Data Sources take two type parameters:
Identity
is a type that has enough information to fetch the dataResult
is the type of data we want to fetch
import cats.data.NonEmptyList
import cats.effect.Concurrent
trait DataSource[F[_], Identity, Result]{
def data: Data[Identity, Result]
def CF: Concurrent[F]
def fetch(id: Identity): F[Option[Result]]
def batch(ids: NonEmptyList[Identity]): F[Map[Identity, Result]]
}
Returning Concurrent
instances from the fetch methods allows us to specify if the fetch must run synchronously or asynchronously, and use all the goodies available in cats
and cats-effect
.
We'll implement a dummy data source that can convert integers to strings. For convenience, we define a fetchString
function that lifts identities (Int
in our dummy data source) to a Fetch
.
import cats._
import cats.data.NonEmptyList
import cats.effect._
import cats.implicits._
import fetch._
def latency[F[_] : Sync](milis: Long): F[Unit] =
Sync[F].delay(Thread.sleep(milis))
object ToString extends Data[Int, String] {
def name = "To String"
def source[F[_] : Async]: DataSource[F, Int, String] = new DataSource[F, Int, String]{
override def data = ToString
override def CF = Concurrent[F]
override def fetch(id: Int): F[Option[String]] = for {
_ <- CF.delay(println(s"--> [${Thread.currentThread.getId}] One ToString $id"))
_ <- latency(100)
_ <- CF.delay(println(s"<-- [${Thread.currentThread.getId}] One ToString $id"))
} yield Option(id.toString)
override def batch(ids: NonEmptyList[Int]): F[Map[Int, String]] = for {
_ <- CF.delay(println(s"--> [${Thread.currentThread.getId}] Batch ToString $ids"))
_ <- latency(100)
_ <- CF.delay(println(s"<-- [${Thread.currentThread.getId}] Batch ToString $ids"))
} yield ids.toList.map(i => (i, i.toString)).toMap
}
}
def fetchString[F[_] : Async](n: Int): Fetch[F, String] =
Fetch(n, ToString.source)
Since we'll use IO
from the cats-effect
library to execute our fetches, we'll need an IORuntime
for executing our IO
instances.
import cats.effect.unsafe.implicits.global //Gives us an IORuntime in places it is normally not provided
Normally, in your applications, this is provided by IOApp
, and you should not need to import this except in limited scenarios such as test environments that do not have Cats Effect integration.
For more information, and particularly on why you would usually not want to make one of these yourself, see this post by Daniel Spiewak
Now that we can convert Int
values to Fetch[F, String]
, let's try creating a fetch.
def fetchOne[F[_] : Async]: Fetch[F, String] =
fetchString(1)
Let's run it and wait for the fetch to complete. We'll use IO#unsafeRunTimed
for testing purposes, which will run an IO[A]
to Option[A]
and return None
if it didn't complete in time:
import scala.concurrent.duration._
Fetch.run[IO](fetchOne).unsafeRunTimed(5.seconds)
// --> [169] One ToString 1
// <-- [169] One ToString 1
// res0: Option[String] = Some(value = "1")
As you can see in the previous example, the ToStringSource
is queried once to get the value of 1.
Multiple fetches to the same data source are automatically batched. For illustrating this, we are going to compose three independent fetch results as a tuple.
def fetchThree[F[_] : Async]: Fetch[F, (String, String, String)] =
(fetchString(1), fetchString(2), fetchString(3)).tupled
When executing the above fetch, note how the three identities get batched, and the data source is only queried once.
Fetch.run[IO](fetchThree).unsafeRunTimed(5.seconds)
// --> [170] Batch ToString NonEmptyList(1, 2, 3)
// <-- [170] Batch ToString NonEmptyList(1, 2, 3)
// res1: Option[(String, String, String)] = Some(value = ("1", "2", "3"))
Note that the DataSource#batch
method is not mandatory. It will be implemented in terms of DataSource#fetch
if you don't provide an implementation.
object UnbatchedToString extends Data[Int, String] {
def name = "Unbatched to string"
def source[F[_]: Async] = new DataSource[F, Int, String] {
override def data = UnbatchedToString
override def CF = Concurrent[F]
override def fetch(id: Int): F[Option[String]] =
CF.delay(println(s"--> [${Thread.currentThread.getId}] One UnbatchedToString $id")) >>
latency(100) >>
CF.delay(println(s"<-- [${Thread.currentThread.getId}] One UnbatchedToString $id")) >>
CF.pure(Option(id.toString))
}
}
def unbatchedString[F[_]: Async](n: Int): Fetch[F, String] =
Fetch(n, UnbatchedToString.source)
Let's create a tuple of unbatched string requests.
def fetchUnbatchedThree[F[_] : Async]: Fetch[F, (String, String, String)] =
(unbatchedString(1), unbatchedString(2), unbatchedString(3)).tupled
When executing the above fetch, note how the three identities get requested in parallel. You can override batch
to execute queries sequentially if you need to.
Fetch.run[IO](fetchUnbatchedThree).unsafeRunTimed(5.seconds)
// --> [169] One UnbatchedToString 1
// --> [170] One UnbatchedToString 2
// <-- [169] One UnbatchedToString 1
// <-- [170] One UnbatchedToString 2
// --> [169] One UnbatchedToString 3
// <-- [169] One UnbatchedToString 3
// res2: Option[(String, String, String)] = Some(value = ("1", "2", "3"))
If we combine two independent fetches from different data sources, the fetches can be run in parallel. First, let's add a data source that fetches a string's size.
object Length extends Data[String, Int] {
def name = "Length"
def source[F[_] : Async] = new DataSource[F, String, Int] {
override def data = Length
override def CF = Concurrent[F]
override def fetch(id: String): F[Option[Int]] = for {
_ <- CF.delay(println(s"--> [${Thread.currentThread.getId}] One Length $id"))
_ <- latency(100)
_ <- CF.delay(println(s"<-- [${Thread.currentThread.getId}] One Length $id"))
} yield Option(id.size)
override def batch(ids: NonEmptyList[String]): F[Map[String, Int]] = for {
_ <- CF.delay(println(s"--> [${Thread.currentThread.getId}] Batch Length $ids"))
_ <- latency(100)
_ <- CF.delay(println(s"<-- [${Thread.currentThread.getId}] Batch Length $ids"))
} yield ids.toList.map(i => (i, i.size)).toMap
}
}
def fetchLength[F[_] : Async](s: String): Fetch[F, Int] =
Fetch(s, Length.source)
And now we can easily receive data from the two sources in a single fetch.
def fetchMulti[F[_] : Async]: Fetch[F, (String, Int)] =
(fetchString(1), fetchLength("one")).tupled
Note how the two independent data fetches run in parallel, minimizing the latency cost of querying the two data sources.
Fetch.run[IO](fetchMulti).unsafeRunTimed(5.seconds)
// --> [169] One Length one
// --> [170] One ToString 1
// <-- [170] One ToString 1
// <-- [169] One Length one
// res3: Option[(String, Int)] = Some(value = ("1", 3))
The Fetch library supports deduplication and optional caching. By default, fetches that are chained together will share the same cache backend, providing some deduplication.
When fetching an identity twice within the same Fetch
, such as a batch of fetches or when you flatMap
one fetch into another, subsequent fetches for the same identity are cached.
Let's try creating a fetch that asks for the same identity twice, by using flatMap
(in a for-comprehension) to chain the requests together:
def fetchTwice[F[_] : Async]: Fetch[F, (String, String)] = for {
one <- fetchString(1)
two <- fetchString(1)
} yield (one, two)
While running it, notice that the data source is only queried once. The next time the identity is requested, it's served from the internal cache.
val runFetchTwice = Fetch.run[IO](fetchTwice)
runFetchTwice.unsafeRunTimed(5.seconds)
// --> [170] One ToString 1
// <-- [170] One ToString 1
// res4: Option[(String, String)] = Some(value = ("1", "1"))
This will still fetch the data again, however, if we call it once more:
runFetchTwice.unsafeRunTimed(5.seconds)
// --> [170] One ToString 1
// <-- [170] One ToString 1
// res5: Option[(String, String)] = Some(value = ("1", "1"))
If we want to cache between multiple individual fetches, you should use Fetch.runCache
or Fetch.runAll
to return the cache for reusing later.
Here is an example where we fetch four separate times, and explicitly share the cache to keep the deduplication functionality:
//We get the cache from the first run and pass it to all subsequent fetches
val runFetchFourTimesSharedCache = for {
(cache, one) <- Fetch.runCache[IO](fetchString(1))
two <- Fetch.run[IO](fetchString(1), cache)
three <- Fetch.run[IO](fetchString(1), cache)
four <- Fetch.run[IO](fetchString(1), cache)
} yield (one, two, three, four)
runFetchFourTimesSharedCache.unsafeRunTimed(5.seconds)
// --> [169] One ToString 1
// <-- [169] One ToString 1
// res6: Option[(String, String, String, String)] = Some(
// value = ("1", "1", "1", "1")
// )
As you can see above, the cache will now work between calls and can be used to deduplicate requests over a period of time. Note that this does not support any kind of automatic cache invalidation, so you will need to keep track of which values you want to re-fetch if you plan on sharing the cache.
For more in-depth information, take a look at our documentation.
Fetch is designed and developed by 47 Degrees
Copyright (C) 2016-2023 47 Degrees. http://47deg.com