Skip to content

Type-safe left-to-right functional pipeline composition

Notifications You must be signed in to change notification settings

lorefnon/fluent-piper

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

fluent-piper

Have you see code like dispatch(intUserEvent(await findActiveUser(flatten(ids.map((it) => it.split(",")))))) and felt: wow, that's ugly.

Such code is hard to read is because to understand it you need to first unwrap it inside out. The FP world has had a solution for this for quite some time: functional pipelines.

This library brings a similar approach to typescript: You can now refactor the above code to be:

await pipe(ids)
    .thru(ids => ids.map(it => it.split(",")))
    .thru(flatten)
    .thru(findActiveUser)
    .await()
    .thru(initUserEvent)
    .thru(dispatch)
    .value

This is longer, but also easier to read because we can follow the logic top -> down, left -> right in natural reading order.

Also, unlike many similar javascript libraries, this is fully type-safe and does not impose any limitations on the number of steps your chain can have.

Installation

pnpm (recommended):

pnpm install fluent-piper

npm:

npm install fluent-piper

Usage - Eager evaluation

We pass an initial value to pipe and chain steps using .thru and finally call .value to get the result.

import {pipe} from "fluent-piper";

const result = pipe(10).thru((i: number) => i + 1).thru((i: number) => `Result: ${i}`).value;
//                   |           ^            |          ^
//                   |___________|            |__________|----- Types of Subsequent output -> input pairs must match
//                         |
//                         Initial input must match input of first step
//
// result: string = "Result: 11"
//         ^        ^
//         |        |__ Result of left-to-right composition
//         |
//         |__ Output type inferred from output type of last step
//

In above usage, steps are eagerly evaluated - every step passed to .thru gets immediately executed.

With async steps

We can use .await() to ensure that next step receives resolved value instead of a promise

import {pipe} from "fluent-piper";

const result = pipe(10)
    .thru(async (i: number) => i + 1)
    .await()
    .thru((i: number) => `Result: ${i}`)
        // ^--- Not a promise
    .value;
// result: Promise<string>

Advanced usage - Lazy Pipelines

There is a lazy API that offers few more features at the cost of higher overhead. The lazy API builds up a chain of thunks that get executed when the final .value is called. Until .value is invoked, nothing gets executed.

Catching exceptions

If some of the steps can throw, we can use .catch to handle them within the pipeline and chain subsequent steps

import {pipe} from "fluent-piper"

const departmentName = await pipe.lazy({ id: 1})
    .thru(fetchUser)
    .await()
    .thru(fetchDepartment)
    .await()
    .catch(e => {
        // Catch errors thrown from previous steps (sync/async)
        console.error(e);
        return null;
    })
    .thru(dept => dept?.name)
    .value;

Bailing early

A more railway oriented approach to handle early exits in a type-safe manner is available through .bailIf:

import {pipe} from "fluent-piper"

const departmentInfo = await pipe.lazy({ id: 1})
    .thru(fetchUser)
    .await()
    .bailIf(
        user => !user.departmentId, // Decide if to bail
        user => ({ type: "unassigned" as const }) // Value to bail with
    )
    .thru(fetchDepartment) // We will reach here only if we didn't bail above
    .await()
    .bailIf(
        isPublic, // (dept) => boolean
        dept => ({ type: "classified" as const })
    )
    .thru(dept => ({ type: "public" as const, name: dept.name })) // Only if dept is public
    .value; // Promise<{ type: "unassigned" } | { type: "classified" } | { type: "public", name: string }>

License

MIT

About

Type-safe left-to-right functional pipeline composition

Resources

Stars

Watchers

Forks

Packages

No packages published