Name: stores
Owner: Dojo
Description: :rocket: Dojo 2 - data stores.
Created: 2016-04-11 07:46:15.0
Updated: 2018-05-02 14:00:38.0
Pushed: 2018-05-02 14:00:36.0
Homepage: http://dojo.io
Size: 915
Language: TypeScript
GitHub Committers
User | Most Recent Commit | # Commits |
---|
Other Committers
User | Most Recent Commit | # Commits |
---|
This library provides an application store designed to complement @dojo/widgets and @dojo/widget-core or any other reactive application.
To use @dojo/stores
, install the package along with its required peer dependencies:
install @dojo/stores
er dependencies
install @dojo/core
install @dojo/has
install @dojo/shim
install @dojo/widget-core
Dojo 2 stores is a predictable, consistent state container for Javascript applications with inspiration from Redux and Flux architectures. However, the @dojo/stores
package aims to provide more built-in support for common patterns such as asynchronous behaviors, undo support and more!
Managing state can become difficult to coordinate when an application becomes complicated with multiple views, widgets, components, and models. With each of these attempting to update attributes of state at varying points within the application lifecycle things can get confusing. When state changes are hard to understand and/or non-deterministic it becomes increasingly difficult to identify and reproduce bugs or add new features.
The @dojo/stores
package provides a centralized store, designed to be the single source of truth for an application. It operates using uni-directional data flow. This means all application data follows the same lifecycle, ensuring the application logic is predictable and easy to understand.
Note: Do you need a centralized store? Lifting state up to parent widgets and using local state are likely to be sufficient in less complex applications.
To work with @dojo/stores
there are three core but simple concepts - Operations, Commands, and Processes.
Operation
Command
Process
Operations are the raw instructions the store uses to make modifications to the state. The operations are based on the JSON Patch and JSON Pointer specifications that have been customized specifically for Dojo 2 stores, primarily to prevent access to the state's root.
Dojo 2 stores support four of the six JSON Patch operations: “add”, “remove”, “replace”, and “test”. The “copy” and “move” operations are not currently supported. Each operation is a simple object which contains instructions with the OperationType
, path
and optionally the value
(depending on operation).
rt { OperationType } from '@dojo/stores/State/patch';
rt { Pointer } from '@dojo/stores/state/Pointer';
t operations = [{
op: OperationType.ADD,
path: new Pointer('/foo'),
value: 'foo'
op: OperationType.REPLACE,
path: new Pointer('/bar'),
value: 'bar'
op: OperationType.REMOVE,
path: new Pointer('/qux')
Dojo 2 stores provides a helper package that can generate PatchOperation
objects from @dojo/stores/state/operations
:
add
- Returns a PatchOperation
of type OperationType.ADD
for the path
and value
remove
- Returns a PatchOperation
of type OperationType.REMOVE
for the path
replace
- Returns a PatchOperation
of type OperationType.REPLACE
for the path
and value
test
- Returns a PatchOperation
of type OperationType.TEST
for the path
and value
These functions accept a Path
type. This is returned by the path
and at
methods on a store. More often this will be created using the path
and at
functions provided as part of the arguments to a command, which are described in more detail below. Rather than accepting a full string path, the path
and at
functions accept a series of arguments and provide type checking to verify that the path created actually exists on the state object.
Commands are simply functions which are called internally by the store when executing a Process
and return an array of PatchOperations
that tells the store
what state changes needs to be performed.
Each command is passed a CommandRequest
which provides path
and at
functions to generate Path
s in a typesafe way, a get
function for access to the store's state, and a payload
object for the argument that the process executor was called with.
The get
function returns back state for a given Path
, for example get(path('my', 'deep', 'state'))
or get(at(path('my', 'array', 'item'), 9))
.
tion addTodoCommand({ at, path, get, payload }: CommandRequest) {
const todosPath = path('todos');
const length = get(todosPath).length;
const operations = [
add(at(todosPath, length), payload)
];
return operations;
tion calculateCountsCommand({ get, path }: CommandRequest) {
const todos = get(path('todos'));
const completedTodos = todos.filter((todo: any) => todo.completed);
const operations = [
replace(path('activeCount'), todos.length - completedTodos.length),
replace(path('completedCount'), completedTodos.length)
];
return operations;
A Command
, or the CommandRequest
argument to it, can be provided with generics that indicate the type of the state of the store they are intended to target and the payload that will be passed. This will provide type checking for all calls to path
and at
and usages of payload
, ensuring that the operations will be targeting real properties of the store, and providing type inference for the return type of get
.
rface MyState {
id: string;
rface Payload {
id: string;
t command = (request: CommandRequest<MyState, Payload>) => {
return [
add(path('id'), request.payload.id);
];
In order to avoid typing each Command
explicitly, a CommandFactory
can be created that will pass its generic types onto all commands it creates. Creating commands using a factory is essentially the same as creating them without one. It is simply a convenience to avoid repeating the same typings for each command.
rface Todo {
completed: boolean;
label: string;
rface TodoState {
todos: Todo[];
activeCount: number;
completedCount: number;
t createCommand = createCommandFactory<TodoState>();
t addTodoCommand = createCommand(({ at, get, path, payload }) => {
const todos = get(path('todos'));
// const todos = get(path('todoes')); Fails to compile because `todoes` is not a property on the state
// const value = todos + 3; Fails to compile because todos is typed as inferred to be an array of `Todo` objects
const operations = [
// Using the utilities provided by the `operations` module ensures that the paths provided are valid,
// and that the values being added or replaced are of the appropriate type
add(at(path('todos'), todos.length), payload)
];
return operations;
t calculateCountsCommand = createCommand(({ get, path }) => {
const todos = get(path('todos'));
const completedTodos = todos.filter((todo: any) => todo.completed);
const operations = [
replace(path('activeCount'), todos.length - completedTodos.length),
replace(path('completedCount'), completedTodos.length)
];
return operations;
Important: Access to state root is not permitted and will throw an error, for example, get(path('/'))
. This applies to Operations
also, it is not possible to create an operation that will update the state root. Best practices with @dojo/stores mean touching the smallest part of the store as is necessary.
Commands support asynchronous behavior out of the box simply by returning a Promise<PatchOperation[]>
.
c function postTodoCommand({ get, path, payload: { id }}: CommandRequest): Promise<PatchOperation[]> {
const response = await fetch('/todos');
if (!response.ok) {
throw new Error('Unable to post todo');
}
const json = await response.json();
const todos = get(path('todos'));
const index = findIndex(todos, byId(id));
// success
return [
replace(at(path('todos'), index), {
...todos[index],
loading: false,
id: data.uuid
})
];
A Process
is the construct used to execute commands against a store
instance in order to make changes to the application state. Processes
are created using the createProcess
factory function that accepts an array of commands and an optional callback that can be used to manage errors thrown from a command. The optional callback receives an error
object and a result
object. The error
object contains the error
stack and the command that caused the error.
The result object contains the following:
payload
passed to the processundoOperations
to undo the process
process
.The array of Commands
are executed in sequence by the store until the last Command is completed or a Command
throws an error. These processes often represent an application behavior. For example, adding a todo in a simple todo application which will be made up with multiple discreet commands.
A simple process
to add a todo and recalculate the todo count:
t addTodoProcess = createProcess('add-todo', [ addTodoCommand, calculateCountCommand ]);
A callback
can be provided which will be called when an error occurs or the process is successfully completed:
tion addTodoProcessCallback(error, result) {
if (error) {
// do something with the error
// possibly undo the operations
result.store.apply(result.undoOperations);
}
// possible additional state changes by running another process using result.executor(otherProcess)
t addTodoProcess = createProcess('add-todo', [ addTodoCommand, calculateCountCommand ], addTodoProcessCallback);
The Process
creates a deferred executor by passing the store
instance addTodoProcess(store)
which can be executed immediately by passing the payload
, addTodoProcess(store)(payload)
. Or more often passed to your widgets and used to initiate state changes on user interactions. The payload
argument for the executor
is required and is passed to each of the Process
's commands in a payload
argument.
t addTodoExecutor = addTodoProcess(store);
odoExecutor({
foo: 'arguments',
bar: 'get',
baz: 'passed',
qux: 'here'
Initial state can be defined on store creation by executing a Process
after the store has been instantiated.
ommand that creates the basic initial state
t initialStateCommand = createCommand({ path }) => {
return [
add(path('todos'), []),
add(path('currentTodo'), ''),
add(path('activeCount'), 0),
add(path('completedCount'), 0)
]);
t initialStateProcess = createProcess('initial', [ initialStateCommand ]);
reates the store, initializes the state and runs the `getTodosProcess` shown below.
t store = createStore();
ialStateProcess(store)();
f a process contains an async command, like fetching initial data from a remote service the return promise can be used
o control the flow.
odosProcess(store)().then(() => {
// do things once the todos have been fetched.
The payload
argument for the process executor can be specified as the second generic type when using createProcess
t process = createProcess<any, { foo: string }>('foo', [ command ]);
t processExecutor = process(store);
he executor will require an argument that satisfies `{ foo: string }`
essExecutor({ foo: 'bar' });
essExecutor({ foo: 1 }); // Compile error
The process executor's payload
type will also be inferred by the payload
type of the commands if not specified explicitly, however, the payload
type for all the commands must be assignable when this is not the case the payload generic type needs to be explicitly passed.
t createCommandOne = createCommandFactory<any, { foo: string }>();
t createCommandTwo = createCommandFactory<any, { bar: string }>();
t commandOne = createCommandOne(({ get, path, payload }) => []);
t commandTwo = createCommandTwo(({ get, path, payload }) => []);
t processOne = createProcess('one', [commandOne]);
t executorOne = processOne(store); // payload for executor inferred based on `commandOne`
utorOne({ foo: 'foo' });
utorOne({ bar: 'foo' }); // compile error
ompile error as payload types for commandOne and commandTwo are not assignable
t processTwo = createProcess('two', [commandOne, commandTwo]);
xplicitly passing a generic that satisfies all the command payload types enables payload type widening
t processTwo = createProcess<any, { foo: string, bar: string }>('two', [commandOne, commandTwo]);
t executorTwo = processTwo(store);
utorTwo({ foo: 'foo' }); // compile error, as requires both `bar` and `foo`
utorTwo({ foo: 'foo', bar: 'bar' }); // Yay, valid
Alternatively, the payload can be typed at command creation
t createCommandOne = createCommandFactory<MyState>();
t commandOne = createCommandOne<{ foo: string }>(({ get, path, payload }) => []);
Although Dojo 2 stores is a big atom state store, you never get access to the entire state object. To access the sections of state that are needed we use pointers to return the slice of state that is needed i.e. path('path', 'to', 'state')
. State is never directly updated by the user, with state changes only being processed by the operations returned by commands.
There is no concept of reducers
, meaning that there is no confusion about where logic needs to reside between reducers
and actions
. Commands
are the only place that state logic resides and return operations
that dictate what state
changes are required and processed internally by the store
.
Additionally, this means that there is no need to coordinate actions
and reducers
using a string action key. Commands are simple function references that can be reused in multiple processes
.
To be notified of changes in the store, use the onChange()
function, which takes a path
or an array of path
's and a callback for when that portion of state changes, for example:
e.onChange(store.path('foo', 'bar'), () => {
// do something when the state at foo/bar has been updated.
or
e.onChange([
store.path('foo', 'bar'),
store.path('baz')
) => {
// do something when the state at /foo/bar or /baz has been updated.
In order to be notified when any changes occur within the store's state, simply register to the stores .on()
method for a type of invalidate
passing the function to be called.
e.on('invalidate', () => {
// do something when the store's state has been updated.
Store data can be connected to widgets within your application using the Containers & Injectors Pattern supported by @dojo/widget-core
. The @dojo/stores
package provides a specialized injector that invalidates store containers on two conditions:
paths
on container creation to ensure invalidation will only occur when state you are interested in changes.paths
are defined for the container, it will invalidate when any data changes in the store.rt { WidgetBase } from '@dojo/widget-core/WidgetBase';
rt { Store } from '@dojo/stores/Stores';
rt { StoreContainer } from '@dojo/stores/StoreInjector';
rface State {
foo: string;
bar: {
baz: string;
}
ill only invalidate when the foo property is changed
t Container = StoreContainer<State>(WidgetBase, 'state', { paths: [ [ 'foo' ] ], getProperties(store: Store<State>) {
return {
foo: store.get(store.path('foo'))
}
atch all, will invalidate if _any_ state changes in the store even if the container is not interested in the changes
t Container = StoreContainer<State>(WidgetBase, 'state', { getProperties(store: Store<State>) {
return {
foo: store.get(store.path('foo'))
}
Instead of typing StoreContainer
for every usage, it is possible to create a pre-typed StoreContainer that can be used across your application. To do this use createStoreContainer
from @dojo/stores/StoreInjector
passing the stores state interface as the generic argument. The result of this can be exported and used across your application.
rface State {
foo: string;
rt const StoreContainer = createStoreContainer<State>();
An optional transformer
can be passed to a process that is used to transform the executor
's payload to the command
payload type. The return type of the transformer
must match the command
payload
type of the process
. The argument type of the executor
is inferred from transformers payload
type.
rface CommandPayload {
bar: number;
foo: number;
CommandPayload` types the command payload and the argument of the process executor
t createCommand = createCommandFactory<any, CommandPayload>();
payload` is typed to `CommandPayload`
t command = createCommand(({ get, path, payload }) => {
t process = createProcess('example', [command]);
rface TransformerPayload {
foo: string;
he transformer return type must match the original `CommandPayload`
t transformer = (payload: TransformerPayload): CommandPayload => {
return {
bar: 1,
foo: 2
};
hen no transformer passed to the process the executor `payload` is typed to `CommandPayload`
t processExecutor = process(store);
orks
essExecutor({ bar: 1, foo: 2 });
hese shouldn't work as `payload` does not match `CommandPayload` interface
essExecutor({ bar: '', foo: 2 });
essExecutor({ foo: '' });
hen a `transformer` passed to the process the `transformer` return type must match `CommandPayload`
nd the executor `payload` type becomes `payload` type of the `transformer`
t processExecutor = process(store, transformer);
orks as the `transformer` payload type is `{ foo: string }`
essExecutor({ foo: '' });
houldn't work as `payload` does not match the `TransformerPayload` type
essExecutor({ bar: 1, foo: 2 });
essExecutor({ bar: 1, foo: '' });
Each Command
will be passed the result of the transformer as the payload
for example: { bar: 1, foo: 2 }
Optimistic updating can be used to build a responsive UI despite interactions that might take some time to respond, for example saving to a remote resource.
In the case of adding a todo item for instance, with optimistic updating, we can immediately add the todo before we even make a request to the server and avoid having an unnatural waiting period or loading indicator. When the server responds we can then reconcile the outcome based on whether it is successful or not.
In the success scenario, we might need to update the added Todo item with an id that was provided in the response from the server and change the color of the Todo item to green to indicate it was successfully saved.
In the error scenario, it might be that we want to show a notification to say the request failed and turn the Todo item red, with a “retry” button. It's even possible to revert/undo the adding of the Todo item or anything else that happened in the process.
t handleAddTodoErrorProcess = createProcess('error', [ () => [ add(path('failed'), true) ]; ]);
tion addTodoCallback(error, result) {
if (error) {
result.store.apply(result.undoOperations);
result.executor(handleAddTodoErrorProcess);
}
t addTodoProcess = createProcess('add-todo', [
addTodoCommand,
calculateCountsCommand,
postTodoCommand,
calculateCountsCommand
],
addTodoCallback);
addTodoCommand
: Adds the new todo into the application statecalculateCountsCommand
: Recalculates the count of completed and active todo itemspostTodoCommand
: posts the todo item to a remote service and using the process callback we can make changes if there is a failurefailed
state field is set to true
id
field with the value received from the remote servicecalculateCountsCommand
: Runs again after the success of postTodoCommand
To support “pessimistic” updates to the application state, i.e. wait until a remote service call has been completed before changing the application state simply put the async command before the application store update. This can be useful when performing a deletion of a resource, when it can be surprising if an item is removed from the UI “optimistically” only for it to reappear back if the remote service call fails.
tion byId(id: string) {
return (item: any) => id === item.id;
c function deleteTodoCommand({ get, payload: { id } }: CommandRequest) {
const { todo, index } = find(get('/todos'), byId(id))
await fetch(`/todo/${todo.id}`, { method: 'DELETE' } );
return [ remove(path('todos', index)) ];
t deleteTodoProcess = createProcess('delete', [ deleteTodoCommand, calculateCountsCommand ]);
Note: The process requires the counts to be recalculated after successfully deleting a todo, the process above shows how easily commands can be shared and reused.
A Process
supports concurrent execution of multiple commands by specifying the commands in an array when creating the process:
t myProcess = createProcess('my-process', [ commandOne, [ concurrentCommandOne, concurrentCommandTwo ], commandTwo ]);
In this example, commandOne
is executed, then both concurrentCommandOne
and concurrentCommandTwo
are executed concurrently. Once all of the concurrent commands are completed the results are applied in order before continuing with the process and executing commandTwo
.
Note: Concurrent commands are always assumed to be asynchronous and resolved using Promise.all
.
The Process
callback provides a hook to apply generic/global functionality across multiple or all processes used within an application. This is done using higher order functions that wrap the process' local callback
using the error and result payload to decorate or perform an action for all processes it is used for.
callback
decorators can be composed together to combine multiple units of functionality, such that in the example below myProcess
would run the error
and result
through the collector
, logger
and then snapshot
callbacks.
t myProcess = createProcess('my-process', [ commandOne, commandTwo ], collector(logger(snapshot())));
Specifying a callback
decorator on an individual process explicitly works for targeted behavior but can become cumbersome when the decorator needs to be applied to multiple processes throughout the application.
The createProcessWith
higher order function can be used to specify callback
decorators that need to be applied across multiple processes
. The function accepts an array of callback
decorators and returns a new createProcess
factory function that will automatically apply the decorators to any process that it creates.
t customCreateProcess = createProcessWith([ logger ]);
myProcess` will automatically be decorated with the `logger` callback decorator.
t myProcess = customCreateProcess('my-process', [ commandOne, commandTwo ]);
An additional helper function createCallbackDecorator
can be used to turn a simple ProcessCallback
into a decorator that ensures the passed callback is executed after the decorating callback
has been run.
t myCallback = (error: ProcessError, result: ProcessResult) => {
// do things with the outcome of the process
urns the callback into a callback decorator
t myCallbackDecorator = createCallbackDecorator(myCallback);
se the callback decorator as normal
t myProcess = createProcess('my-process', [ commandOne ], myCallbackDecorator());
We appreciate your interest! Please see the Dojo 2 Meta Repository for the Contributing Guidelines.
This repository uses prettier
for code styling rules and formatting. A pre-commit hook is installed automatically and configured to run prettier
against all staged files as per the configuration in the project's package.json
.
An additional npm script to run prettier
(with write set to true
) against all src
and test
project files is available by running:
run prettier
To start working with this package, clone the repository and run npm install
.
In order to build the project run grunt dev
or grunt dist
.
Test cases MUST be written using Intern using the Object test interface and Assert assertion interface.
90% branch coverage MUST be provided for all code submitted to this repository, as reported by istanbul?s combined coverage results for all supported platforms.
To test locally in node run:
grunt test
To test against browsers with a local selenium server run:
grunt test:local
To test against BrowserStack or Sauce Labs run:
grunt test:browserstack
or
grunt test:saucelabs
© 2018 JS Foundation. New BSD license.