Building Pipelines

On this page

Effect pipelines allow for the composition and sequencing of operations on values, enabling the transformation and manipulation of data in a concise and modular manner.

Why Pipelines are Good for Structuring Your Application

Pipelines are an excellent way to structure your application and handle data transformations in a concise and modular manner. They offer several benefits:

  1. Readability: Pipelines allow you to compose functions in a readable and sequential manner. You can clearly see the flow of data and the operations applied to it, making it easier to understand and maintain the code.

  2. Code Organization: With pipelines, you can break down complex operations into smaller, manageable functions. Each function performs a specific task, making your code more modular and easier to reason about.

  3. Reusability: Pipelines promote the reuse of functions. By breaking down operations into smaller functions, you can reuse them in different pipelines or contexts, improving code reuse and reducing duplication.

  4. Type Safety: By leveraging the type system, pipelines help catch errors at compile-time. Functions in a pipeline have well-defined input and output types, ensuring that the data flows correctly through the pipeline and minimizing runtime errors.

Now, let's delve into how to define pipelines and explore some of the key components:

pipe

The pipe function is a utility that allows us to compose functions in a readable and sequential manner. It takes the output of one function and passes it as the input to the next function in the pipeline. This enables us to build complex transformations by chaining multiple functions together.

The basic syntax of pipe is as follows:

ts
import { pipe } from "effect"
const result = pipe(input, func1, func2, ..., funcN)
ts
import { pipe } from "effect"
const result = pipe(input, func1, func2, ..., funcN)

In this syntax, input is the initial value, and func1, func2, ..., funcN are the functions to be applied in sequence. The result of each function becomes the input for the next function, and the final result is returned.

Here's an illustration of how pipe works:

Pipe

It's important to note that functions passed to pipe must have a single argument because they are only called with a single argument.

Let's see an example to better understand how pipe works:

ts
import { pipe } from "effect"
 
const increment = (x: number) => x + 1
const double = (x: number) => x * 2
const subtractTen = (x: number) => x - 10
 
const result = pipe(5, increment, double, subtractTen)
 
console.log(result) // Output: 2
ts
import { pipe } from "effect"
 
const increment = (x: number) => x + 1
const double = (x: number) => x * 2
const subtractTen = (x: number) => x - 10
 
const result = pipe(5, increment, double, subtractTen)
 
console.log(result) // Output: 2

In the above example, we start with an input value of 5. The increment function adds 1 to the initial value, resulting in 6. Then, the double function doubles the value, giving us 12. Finally, the subtractTen function subtracts 10 from 12, resulting in the final output of 2.

Functions vs Methods

In the Effect ecosystem, libraries often expose functions rather than methods. This design choice is important for two key reasons: tree shakeability and extendibility.

Tree Shakeability

Tree shakeability refers to the ability of a build system to eliminate unused code during the bundling process. Functions are tree shakeable, while methods are not.

When functions are used in the Effect ecosystem, only the functions that are actually imported and used in your application will be included in the final bundled code. Unused functions are automatically removed, resulting in a smaller bundle size and improved performance.

On the other hand, methods are attached to objects or prototypes, and they cannot be easily tree shaken. Even if you only use a subset of methods, all methods associated with an object or prototype will be included in the bundle, leading to unnecessary code bloat.

Extendibility

Another important advantage of using functions in the Effect ecosystem is the ease of extendibility. With methods, extending the functionality of an existing API often requires modifying the prototype of the object, which can be complex and error-prone.

In contrast, with functions, extending the functionality is much simpler. You can define your own "extension methods" as plain old functions without the need to modify the prototypes of objects. This promotes cleaner and more modular code, and it also allows for better compatibility with other libraries and modules.

The use of functions in the Effect ecosystem libraries is important for achieving tree shakeability and ensuring extendibility. Functions enable efficient bundling by eliminating unused code, and they provide a flexible and modular approach to extending the libraries' functionality.

Now let's explore some examples of APIs that can be used with the pipe function to build pipelines.

map

The Effect.map function is used to transform the value inside an Effect. It takes a function and applies it to the value contained within the Effect, creating a new Effect with the transformed value.

It's important to note that Effects are immutable, meaning that when you use map on an Effect, it doesn't modify the original data type. Instead, it returns a new copy of the Effect with the transformed value.

The syntax for map is as follows:

ts
import { pipe, Effect } from "effect"
const mappedEffect = pipe(effect, Effect.map(func))
ts
import { pipe, Effect } from "effect"
const mappedEffect = pipe(effect, Effect.map(func))

In the code above, func represents the transformation function to be applied, and effect is the input Effect.

Let's see an example to better understand how map works:

ts
import { pipe, Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const mappedEffect = pipe(
Effect.succeed(5),
Effect.map((x) => increment(x))
)
 
Effect.runPromise(mappedEffect).then(console.log) // Output: 6
ts
import { pipe, Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const mappedEffect = pipe(
Effect.succeed(5),
Effect.map((x) => increment(x))
)
 
Effect.runPromise(mappedEffect).then(console.log) // Output: 6

In the example above:

  • We define a simple increment function that adds 1 to its input.
  • We create an Effect using Effect.succeed with an initial value of 5.
  • We use map to apply the increment function to the value inside the effect.

The resulting Effect contains the transformed value of 6.

If you simply want to replace the value with a constant, you can use Effect.as:

ts
import { pipe, Effect } from "effect"
 
const program = pipe(Effect.succeed(5), Effect.as("new value"))
 
Effect.runPromise(program).then(console.log) // Output: "new value"
ts
import { pipe, Effect } from "effect"
 
const program = pipe(Effect.succeed(5), Effect.as("new value"))
 
Effect.runPromise(program).then(console.log) // Output: "new value"

flatMap

In the previous example, we saw how to transform an Effect using the map function when the transformation function returns a regular value. But what if we want to transform an Effect using a function that itself returns another Effect?

Many JavaScript / TypeScript engineers are familiar with the term flatMap due to it's presence as a method on the Array prototype. However, flatMap can actually be used to describe a more generic data transformation. While this section will focus specifically on explaining how one can use flatMap from the Effect module, we provide a more detailed explanation of flatMap in the FAQ.

If you're just looking for how to flatten a nested Arrays within an Effect:

Flattening a Nested Array

If you have an Effect where result type is a nested Array, for example:

ts
Effect<Array<Array<A>>>
ts
Effect<Array<Array<A>>>

you can easily flatten the array using either the ReadonlyArray module exported from effect:

ts
import { pipe, Effect, ReadonlyArray } from "effect"
 
const flattened = pipe(
Effect.succeed([
[1, 2],
[3, 4]
]),
Effect.map((nested) => ReadonlyArray.flatten(nested))
)
ts
import { pipe, Effect, ReadonlyArray } from "effect"
 
const flattened = pipe(
Effect.succeed([
[1, 2],
[3, 4]
]),
Effect.map((nested) => ReadonlyArray.flatten(nested))
)

or using the standard Array.prototype.flat() method.

The Effect.flatMap function allows us to chain computations that produce Effect values. It takes a transformation function that produces a new Effect and then "flattens" the nested Effect structure.

It's important to note that Effects are immutable, meaning that when you use flatMap on an Effect, it doesn't modify the original data type. Instead, it returns a new copy of the Effect with the transformed value.

The syntax for flatMap is as follows:

ts
import { pipe, Effect } from "effect"
const flatMappedEffect = pipe(effect, Effect.flatMap(func))
ts
import { pipe, Effect } from "effect"
const flatMappedEffect = pipe(effect, Effect.flatMap(func))

In the code above, func represents the transformation function that returns a new Effect, and effect is the input Effect that we want to transform.

Let's see an example to better understand how flatMap works:

ts
import { pipe, Effect } from "effect"
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const flatMappedEffect = pipe(
Effect.succeed([10, 2]),
Effect.flatMap(([a, b]) => divide(a, b))
)
 
Effect.runPromise(flatMappedEffect).then(console.log) // Output: 5
ts
import { pipe, Effect } from "effect"
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const flatMappedEffect = pipe(
Effect.succeed([10, 2]),
Effect.flatMap(([a, b]) => divide(a, b))
)
 
Effect.runPromise(flatMappedEffect).then(console.log) // Output: 5

In the example above:

  • We define a function called divide that takes two numbers a and b and returns an Effect that can potentially succeed with the result of the division or fail with an error if b is zero.
  • We create an Effect using Effect.succeed with a tuple containing the values 10 and 2.
  • We use flatMap to apply the divide function to the values inside the effect.

The resulting Effect contains the transformed value of 5.

When using Effect.flatMap, it is crucial to ensure that all effects are properly considered in the computation. If you ignore an effect returned by a computation within flatMap, it can result in unexpected behavior and incorrect program logic.

Let's take a look at the following code snippet:

ts
Effect.flatMap(([a, b]) => {
Effect.sync(() => console.log(`Performing division: ${a} / ${b}`))
return divide(a, b)
})
ts
Effect.flatMap(([a, b]) => {
Effect.sync(() => console.log(`Performing division: ${a} / ${b}`))
return divide(a, b)
})

The problem here is that the effect returned by Effect.sync will be ignored and not included in the resulting computation. It will not be chained or combined with the effect returned bydivide(a, b). As a result, you may encounter unexpected program behavior and obtain incorrect results.

To ensure correct behavior, make sure that any effect you want to include in the computation is explicitly chained using flatMap or combined with other Effect values using the appropriate functions provided by the Effect library.

In the next section, we will explore how to address this problem using the tap API.

tap

In the previous section, we discussed the problem of ignoring effects in a computation. To address this issue, we can use the tap API, which allows us to execute side effects without affecting the computation flow.

The tap API has a similar signature to flatMap, but the result of the transformation function is ignored. This means that the value returned by the previous computation (i.e. [a, b]) will still be available for the next computation (i.e. divide).

Let's take a look at an updated code snippet that incorporates the tap API to fix the problem:

ts
const program = pipe(
Effect.succeed([10, 2]),
Effect.tap(([a, b]) =>
Effect.sync(() => console.log(`Performing division: ${a} / ${b}`))
),
// [a, b] is still available!
Effect.flatMap(([a, b]) => divide(a, b))
)
 
Effect.runPromise(program).then(console.log)
/*
Output:
Performing division: 10 / 2
5
*/
ts
const program = pipe(
Effect.succeed([10, 2]),
Effect.tap(([a, b]) =>
Effect.sync(() => console.log(`Performing division: ${a} / ${b}`))
),
// [a, b] is still available!
Effect.flatMap(([a, b]) => divide(a, b))
)
 
Effect.runPromise(program).then(console.log)
/*
Output:
Performing division: 10 / 2
5
*/

By using Effect.tap, we include the effect of Effect.sync in the computation. This ensures that the logging action is performed, providing us with the desired output and maintaining correct program logic. The resulting output is still 5, as expected.

Using tap allows us to execute side effects during the computation without altering the result. This can be useful for logging, performing additional actions, or observing the intermediate values without interfering with the main computation flow.

all

The Effect.all function is a powerful utility provided by Effect that allows you to combine multiple effects into a single effect that produces a tuple of results.

The syntax for all is as follows:

ts
import { Effect } from "effect"
const combinedEffect = Effect.all(effects)
ts
import { Effect } from "effect"
const combinedEffect = Effect.all(effects)

Here, effects represents multiple effects that you want to combine.

The all function will execute all these effects in sequence (to explore options for managing concurrency and controlling how these effects are executed, you can refer to the Concurrency Options documentation).

It will return a new effect that produces a tuple containing the results of each individual effect. Keep in mind that the order of the results corresponds to the order of the original effects passed to all.

Let's see an example to better understand how all works:

ts
import { Effect } from "effect"
 
const foo = Effect.succeed(42)
const bar = Effect.succeed("Hello")
 
const combinedEffect = Effect.all([foo, bar])
 
Effect.runPromise(combinedEffect).then(console.log) // Output: [42, "Hello"]
ts
import { Effect } from "effect"
 
const foo = Effect.succeed(42)
const bar = Effect.succeed("Hello")
 
const combinedEffect = Effect.all([foo, bar])
 
Effect.runPromise(combinedEffect).then(console.log) // Output: [42, "Hello"]

In this example, foo and bar are two separate effects. Using Effect.all, we combine them into a single effect combinedEffect. When we run combinedEffect, it produces a tuple [42, "Hello"] containing the results of both effects.

The all function not only combines tuples but also works seamlessly with iterables, structs, and records. To explore the full potential of all head over to the Introduction to Effect's Control Flow Operators documentation.

Build your first pipeline

Now, let's combine all four functions (pipe, map, flatMap and all) to build a pipeline that performs a series of transformations:

ts
import { pipe, Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const task1 = Effect.promise(() => Promise.resolve(10))
 
const task2 = Effect.promise(() => Promise.resolve(2))
 
const program = pipe(
Effect.all([task1, task2]),
Effect.flatMap(([a, b]) => divide(a, b)),
Effect.map((n1) => increment(n1)),
Effect.map((n2) => `Result is: ${n2}`)
)
 
Effect.runPromise(program).then(console.log) // Output: "Result is: 6"
ts
import { pipe, Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const task1 = Effect.promise(() => Promise.resolve(10))
 
const task2 = Effect.promise(() => Promise.resolve(2))
 
const program = pipe(
Effect.all([task1, task2]),
Effect.flatMap(([a, b]) => divide(a, b)),
Effect.map((n1) => increment(n1)),
Effect.map((n2) => `Result is: ${n2}`)
)
 
Effect.runPromise(program).then(console.log) // Output: "Result is: 6"

The pipe method

Effect provides a pipe method that works similarly to the pipe method found in rxjs. This method allows you to chain multiple operations together, making your code more concise and readable.

Here's how the pipe method works:

ts
const result = effect.pipe(func1, func2, ..., funcN)
ts
const result = effect.pipe(func1, func2, ..., funcN)

This is equivalent to using the pipe function like this:

ts
const result = pipe(effect, func1, func2, ..., funcN)
ts
const result = pipe(effect, func1, func2, ..., funcN)

The pipe method is available on all effects and many other data types, eliminating the need to import the pipe function from the Function module and saving you some keystrokes.

Let's rewrite the previous example using the pipe method:

ts
import { Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const task1 = Effect.promise(() => Promise.resolve(10))
 
const task2 = Effect.promise(() => Promise.resolve(2))
 
const program = Effect.all([task1, task2]).pipe(
Effect.flatMap(([a, b]) => divide(a, b)),
Effect.map((n1) => increment(n1)),
Effect.map((n2) => `Result is: ${n2}`)
)
 
Effect.runPromise(program).then(console.log) // Output: "Result is: 6"
ts
import { Effect } from "effect"
 
const increment = (x: number) => x + 1
 
const divide = (a: number, b: number): Effect.Effect<number, Error> =>
b === 0
? Effect.fail(new Error("Cannot divide by zero"))
: Effect.succeed(a / b)
 
const task1 = Effect.promise(() => Promise.resolve(10))
 
const task2 = Effect.promise(() => Promise.resolve(2))
 
const program = Effect.all([task1, task2]).pipe(
Effect.flatMap(([a, b]) => divide(a, b)),
Effect.map((n1) => increment(n1)),
Effect.map((n2) => `Result is: ${n2}`)
)
 
Effect.runPromise(program).then(console.log) // Output: "Result is: 6"

Cheatsheet

Let's summarize the transformation functions we have seen so far:

FunctionInputOutput
mapEffect<A, E, R>, A => BEffect<B, E, R>
flatMapEffect<A, E, R>, A => Effect<B, E, R>Effect<B, E, R>
tapEffect<A, E, R>, A => Effect<B, E, R>Effect<A, E, R>
all[Effect<A, E, R>, Effect<B, E, R>, ...]Effect<[A, B, ...], E, R>

These functions are powerful tools for transforming and chaining Effect computations. They allow you to apply functions to values inside Effect and build complex pipelines of computations.

Remember to choose the appropriate function based on the type of transformation you want to perform. If you have a function that directly returns a value without wrapping it in an Effect, you can use map. If you have a function that returns an Effect itself, you can use flatMap to chain computations.