Awesome

Stanga

The essential Cycling gear every Cyclist needs. Crafted with care. For easier rides.

Motivation

Cycle.js does great job when the application is simple. However, when the application complexity grows, trivial things become non-trivial very quickly. The tutorials and examples don't have any common patterns or best practices to deal with these complex situations and developers are by their own. The goal of this library is to provide some tried-and-tested, "battle-proven" utilities to solve these problems so that you can focus on building your Cycle application.

The origin of these utilities comes from CALM^2, but they're applicable to Cycle codebase easily (with minor modifications) because the architecture designs are so close to each other. Thanks to Vesa who is the main inventor of these patterns.

Tutorial

OBS! If you are searching for API reference, please see below.

Placing the state to Model driver and reading it

If you're already familiar with cycle-examples this section may confuse you. Traditionally in Cycle, the application state lives inside the main so that intents (like clicks) modify the state (= model) and the modified state is passed to the view that produces you a stream of virtual dom vdom$.

With stanga, you move your state from main to Model driver. You can give the initial state to the driver during the driver initialization.

import {run} from "@cycle/core"
import {makeDOMDriver} from "@cycle/dom"
import {Model} from "stanga"

run(main, {
  DOM: makeDOMDriver("#app"),
  M: Model(0)     // initial state = 0
})

Now the state lives in model driver M which is just an observable that you can use like any other observable in Cycle:

function main({DOM, M}) {
  const state$ = M
  return {
    DOM: state$.map(counter => h("h1", `Counter value is ${counter}`))
  }
}

Not bad, huh? Let's go forward!

Modifying the state

Now you know how to read the state. Let's take a look how to modify it. Like in Cycle's docs, write effects go to sinks - and Model driver is not an exception. You may've faced mod$ streams when looking at Cycle's examples. mod$ is a stream of functions currentState => newState, that perform the actual state modifications.

Model driver uses the same mod functions for state updates. However, they those functions must be "converted" into the format that the driver understands. That's why the driver provides a .mod function to do that conversion. Let's make the counter editable!

import {Observable as O} from "rx"

function main({DOM, M}) {
  const state$ = M
  const incMod$ = DOM.select(".inc")
    .events("click")
    .map(() => state => state + 1)
  const decMod$ = DOM.select(".dec")
    .events("click")
    .map(() => state => state - 1)
  
  // let's merge all mods from this component
  const mod$ = O.merge(incMod$, decMod$)
  
  const vdom$ = state$.map(counter => h("div", [
    h("h1", `Counter value is ${counter}`),
    h("button.inc", "++"),
    h("button.dec", "--")
  ]))
  
  return {
    DOM: vdom$,
    // and make the compatible with Model driver
    M: M.mod(mod$)
  }
}

Okay that looks very similar to Cycle's examples. But there is no (direct) connection between intents (mod$) and state$?! That's right, the connection is made inside the driver - driver changes the state based on the mods and emits the changed state back to the component via state$ stream. If you've used HTTP driver, the behaviour is exactly same. Got it?

If you want to completely override the state, Model driver provides .set method, which takes a stream of values (instead of mod functions!) and resets the state by using those values (i.e. shorthand to M.mod(value$.map(value => () => value))).

Accessing sub-states by using lenses

Nothing new? All of that could've done by using "traditional" Cycle approach, you might think. Yes, true. But let's go forward. How about if you must have two counters, print their total value and provide a way to reset both counters at once?

In the "traditional" approach, the solution might look something like this... Whoah! That's a lot of stuff with things like proxy subjects, stream switching, skipping and concatenation. Let's take a look how one would build the same app with stanga's model driver!

Model driver source has .lens(Lens) method that uses internally partial.lenses. In order to understand lenses better, you have to take a look at partial.lenses docs. For now you can just treat them like property getters const a = M.lens("a") where a is a model driver (containing exactly same methods .mod, .set and .lens) BUT so that it uses the original state's property .a - it is a stream that emits the changes only when property a changes. And modifications with a.mod(mod$) change only a's state. However, because a is a part of M, also M gets changed when a changes!!

Let's see how it looks like in the counter example:

function Counter({DOM, M}) {
  // NO CHANGES TO THE EXISTING CODE!
}

function main({DOM, M}) {
  const state$ = M
  const a$ = state$.lens("a")
  const b$ = state$.lens("b")
  // we can use "lensed" a$ and b$ as a driver for child components
  const a = isolate(Counter)({DOM, M: a$})
  const b = isolate(Counter)({DOM, M: b$})

  const resetMod$ = DOM.select(".reset").events("click").map(() => () => ({a: 0, b: 0}))
  const aMod$ = a.M
  const bMod$ = b.M

  const vdom$ = O.combineLatest(state$, a.DOM, b.DOM, (state, a, b) =>
    h("div", [
      a, b,
      h("hr"),
      h("h2", `Total: ${state.a + state.b}`),
      h("button.reset", "Reset")
    ]))
  
  return {
    DOM: vdom$,
    // a.M and b.M are already converted to Model driver's mods
    // in Counter component so we can merge them to parent's mods
    // directly
    M: O.merge(M.mod(resetMod$), aMod$, bMod$)
  }
}

run(main, {
  DOM: makeDOMDriver("#app"),
  M: Model({a: 0, b: 0})
})

As you can see, there is no changes to the original Counter component! And no switching, proxying or other "advanced" stuff. In the end, the problem is not advanced - just some simple stream processing!

Note that you can pass lensed sub-states directly to sub-component as a model driver and merge the mod$ sinks to parent's mod$'es like any like you'd merge any other stream. And this is the core pattern of stanga that gets repeated everywhere. Once you get it, you'll be able to create complex apps like they were as simple as the previous counter app.

Processing list states by using lenses (again)

Lifting list observable to observable of sinks

Lists are a little bit more complicated thing. If you don't believe, just take a look at cycle's advanced list example... just kidding! With stanga, list processing is almost as easy as processing props.

What is "list state"? It's an observable emitting events that contain arrays with arbitrary number of items. If those items have an unique key (e.g. id), stanga provides liftListById, flatCombine and flatMerge that make the list processing extremely easy.

liftListById is most powerful of those functions - it conceals many performance optimizations, caching, cold->hot observable conversions and event replaying that you'd normally need to do by yourself. Now all you need is just an id in your list items! Conceptually liftListById is almost like (but bumped with steroids :muscle:):

const liftListById = fn => list$.map(items => items.map(item => fn(item.id, item)))

The transformer function that is passed to liftListById should invoke some (child) component function and return the sinks from the child component. The transformer function receives item id as item observable as a second parameter

Model implements liftListBy and liftListById so that you can use them directly by using dot notation M.liftListById((id, item$) => ...). In addition when using liftListById with model, the second parameter is lensed item model so you can use it like you'd use the parent model!

The code is far more simpler than the explanation:

let ID = 0

function main({DOM, M}) {
  const counters$ = M
  const childSinks$ = counters$.liftListById((id, counter$) =>
    isolate(Counter, `counter-${id}`)({DOM, M: counter$.lens("val")}))
  // ...
}

run(main, {
  DOM: makeDOMDriver("#app"),
  M: Model([{id: ID++, val: 0}, {id: ID++, val: 0}])    // two counters initially
})

Note that counter$ is a model that can be passed to child component directly as a model driver. Counter component expects the model to be an integer (counter's value) so we need to get the value by using lens (should be nothing new here, huh?).

Extracting values from the lifted sinks

Well.. now you have an observable that emits events of list of sinks and you should "extract" values from those child sinks somehow. Basically there are two different kind of strategies: combining and merging.

In combine strategy, you pick some of the sinks and combine their latest values by using Observable.combineLatest. As a result, you get a list observable containing the latest values from child sinks. Usually you want to apply this extraction strategy to DOM sinks so that you get a nice array of child virtual-dom trees that you can post-process in your parent component (maybe add more virtual-dom around children?).

To use the combine strategy, import flatCombine from stanga and define the sinks you want to "extract and combine" - the returned value is an "sink-like" object containing the extracted sinks as keys and their list observables as values. Again, the actual code is simpler than the explanation:

function main({DOM, M}) {
  const counters$ = M
  const childSinks$ = counters$.liftListById((id, counter$) =>
    isolate(Counter, `counter-${id}`)({DOM, M: counter$.lens("val")}))
  // you can also extract multiple keys, e.g. flatCombine(childSinks$, "DOM", "foo", "bar")
  const children = flatCombine(childSinks$, "DOM")
  // extracted children.DOM is now a normal observable containing an array of 
  // children vtrees
  const vdom$ = children.DOM.map(childVTrees =>
    h("ul", childVTrees.map(child => h("li", [child]))))
  ...
}

Merge strategy behaves almost like combine strategy but instead of combining an array from child sinks, the child sinks are merged by using Observable.merge. This is ideal for sinks containing actions (like HTTP requests or M mods):

function main({DOM, M}) {
  const counters$ = M
  const childSinks$ = counters$.liftListById((id, counter$) =>
    isolate(Counter, `counter-${id}`)({DOM, M: counter$.lens("val")}))
  const vdom$ = ...  
  const childSinks = flatMerge(childSinks$, "M")
  return {
    DOM: vdom$,
    M: childSinks.M 
  }
}

Let's combine those and create the "counter list" component:

let ID = 0

function Counter({DOM, M}) {
  // NOT CHANGED!
}

function main({DOM, M}) {
  const counters$ = M
  const childSinks$ = counters$.liftListById((id, counter$) =>
    isolate(Counter, `counter-${id}`)({DOM, M: counter$.lens("val")}))
  
  const childVTrees$ = flatCombine(childSinks$, "DOM").DOM
  const childMods$ = flatMerge(childSinks$, "M").M 
  
  const resetMod$ = DOM.select(".reset").events("click")
    .map(() => counters => counters.map(c => ({...c, val: 0})))
  const appendMod$ = DOM.select(".add").events("click")
    .map(() => counters => [...counters, {id: ID++, val: 0}])
  
  const vdom$ = O.combineLatest(counters$, childVTrees$, (counters, children) =>
    h("div", [
      h("ul", children.map(child => h("li", [child]))),
      h("hr"),
      h("h2", `Avg: ${avg(counters.map(c => c.val)).toFixed(2)}`),
      h("button.reset", "Reset"),
      h("button.add", "Add counter")
    ]))

  return {
    DOM: vdom$,
    M: O.merge(M.mod(resetMod$), M.mod(appendMod$), childMods$)
  }
}

function avg(list) {
  return list.length ? list.reduce((x, y) => x + y, 0) / list.length : 0
}

run(main, {
  DOM: makeDOMDriver("#app"),
  M: Model([{id: ID++, val: 0}, {id: ID++, val: 0}])    // two counters initially
})

Congrats! Now you know how to create complex apps with Cycle and stanga. The rest is just composing and combining the basic cases we just covered. If you didn't get it now, don't worry - read this tutorial again (and again) and take a look at the examples. It might take some time to learn these all new things like lenses and modifications but it's definitely worth it!

API Reference

All utilities can be imported from stanga package by using CommonJS compatible bundler (like Browserify or Webpack), e.g.

import {Model} from "stanga"

Model

Initializes new model (driver) that can be used with run

Model :: (initialState, opts) => ModelDriver

Initial state can be anything, usually it should be a valid JSON datatype (object, array, number, bool, string...). Valid options are:

• logging enable state logging (default false)
• warn override console warning function (default console.warn)
• info override console info function (default console.info)

L

Just a reference to the underlying partial.lenses implementation.

liftListById

liftListById :: (Observable [A{id, ...}], (id => {string: Observable B}), replay = ["DOM"]) => Observable [{string: Observable B}]
liftListById :: (Lens [A{id, ..}], (id, Lens A => {string: Observable B}), replay = ["DOM"]) => Observable [{string: Observable B}]

Takes a list observable (whose items have id property) and mapper function, applies mapper function to each list item and returns a list observable by using the return values from the mapper function (conceptually same as list$.map(items => items.map(...))).

• Item ids must be unique within the list.
• Mapper function must return "sink-like" objects (JSON object having only Observable values)

If the list observable is a lensed observable (got by using Model.lens(...)), then the mapper function receives also second parameter which is the lensed item associated to the id (first parameter).

By default, DOM sinks are replayed and other sinks are multicasted. If you want to add more replayed sinks, you can override the third parameter which is an array of strings indicating sink keys that should be replayed.

ATTENTION: mapper function is applied only once per item (by id), although the list observable emits multiple values. This enables some heavy performance optimizations to the list processing like duplicate detection, cold->hot observable conversion and caching.

listListBy

liftListById :: ((A => identity), Observable [A], (identity => {string: Observable B}), replay = ["DOM"]) => Observable [{string: Observable B}]
liftListById :: ((A => identity), Lens [A], (identity, Lens A => {string: Observable B}), replay = ["DOM"]) => Observable [{string: Observable B}]

Same as liftListById but allows user to define custom identity function instead of using id property. This function is curried so you can create your own liftListByX by passing only the first parameter

const liftListByTsers = liftListBy(item => item.tsers)

flatCombine

flatCombine :: (Observable [{string: Observable A}], ...string) => {string: Observable [A]}

Takes an list observable of "sink-like" objects (JSON object having Observable values), plucks sinks by using using the given keys and combines the plucked sinks by using Observable.combineLatest

const out = flatCombine(sinks$$, "DOM") // out = {DOM: Observable [vdom1, vdom2, ....]} 
out.DOM.map(childVTrees => h("div", ...)))

flatMerge

flatMerge :: (Observable [{string: Observable A}], ...string) => {string: Observable A}

Same as flatCombine but uses Observable.merge instead of combineLatest, thus resulting an observable of values instead of an observable of lists

const out = flatMerge(sinks$$, "HTTP", "M")
// => {HTTP: Observable req, M: Observable mod}

mergeByKeys

mergeByKeys :: (...{string: Observable}) => {string: Observable}

Takes 1..n sink-like objects and merges the sink observables having same key. The result object contains keys that appear in any of the given sink objects

const a = {HTTP: O.just(req), DOM: O.just(vdom)}
const b = {DOM: O.just(vdom), M: O.just(mod)}
const merged = mergeByKey(a, b)
// => {HTTP: Observable req, DOM: Observable vdom, M: Observable mod}

Migration guide

0.x -> 1.x

• Use partial.lenses@3.x instead of partial.lenses@1.x. Migration guide from here
• Instead of liftListById(model$, ....args) use model$.liftListById(...args)

1.0 -> 2.0

• Replace import {R} from "stanga" with import R from "ramda"
• Install ramda dependency if you were using such imports

License

MIT