A module to access JSONAPI data from an API, using a Pinia store, restructured to make life easier.
NOTE: This library is a replacement for jsonapi-vuex offering near-identical functionality. For those porting projects, please see PORTING.md for a quick-start list of changes. (The name 'reversal' is due to a package name clash in npm).
Documentation, including JSDoc-generated API reference, is available at: Pinia-Jsonapi Documentation
JSONAPI is an extremely useful format for clearly interacting with an API - however it is less useful for the end developer, who is generally more interested in the data contained in a record than the structure surrounding it.
In this module we 'reverse' the JSONAPI data into a form where data attributes become top-level keys, and JSONAPI-specific data is moved down under another key: _jv.
For example, the JSONAPI record:
{
"id": "1",
"type": "widget",
"attributes": {
"name": "sprocket",
"color": "black"
},
"meta": {}
}This would be accessed as record.attributes.color
This module would restructure this record to be:
{
"name": "sprocket",
"color": "black",
"_jv": {
"id": "1",
"type": "widget",
"meta": {}
}
}This would now be accessed as record.color
In cases where there are multiple records being returned in an object, the id is used as the key (though this is ignored in the code, and the 'real' id is always read from _jv):
{
"1": {
"name": "sprocket",
"color": "black"
},
"2": {
"name": "cog",
"color": "red"
}
}These are accessed as record.1.name or record.2.color, or if a list is needed, via Object.values(record).
The above structure is actually how records are maintained in the store, nested below the endpoint:
{
"widget": {
"1": {...},
"2": {...}
},
"doohickey": {
"20": {...}
}
}Relationships in JSONAPI are structured as either a data key containing one or more resource identifier objects under the relationship name, (or links which point to the related object in the API). For data entries, these are added to the 'root' of the object, where the key is the relationship name, and the value is a javascript getter that calls the getData pinia getter for that record. This allows related data to be handled as if it was an attribute. (The object is structured using the same 'id-as-key' system for multiple entries as for records as described above).
The value returned by the getter will therefore be the value of the related object (if in the store), or an empty object (if not). Updating the store will cause the retuern value to update automatically.
{
"id": "1",
"type": "widget",
"attributes": {
"name": "sprocket"
},
"relationships": {
"doohickeys": {
"data": [
{
"type": "doohickey",
"id": "20"
}
]
}
}
}This becomes:
{
"name": "sprocket",
"doohickeys": {
"20": » get('doohickey/20') // getter call
}
}
"_jv": {
"id": "1",
"type": "widget",
"relationships": {...}
}
}The attributes of the related object can then be accessed as e.g.: record.doohickeys.20.size
NOTE: See the examples/tgestapp directory for a simple vue3 app using this module.
Having created a Vue project, simply add the module to your store.js, passing it an axios-like instance:
import axios from 'axios'
import { createJsonapiStore } from 'pinia-jsonapi'
const api = axios.create({
baseURL: 'https://api.example.com/1/api/',
headers: {
'Content-Type': 'application/vnd.api+json',
},
})
const { jsonapiStore } = createJsonapiStore(api)
export { jsonapiStore }The store is then used in components as follows:
import { jsonapiStore } from '../store'
const store = jsonapiStore()NOTE: The store is created with the name/id jv by default - if you wish to instantiate multiple stores, you can pass a name param to createJsonapiStore as follows:
const { jsonapiStore } = createJsonapiStore(api, config, 'store-name')The running config, status tracker and utils are also exported by createJsonapiStore if needed:
const { jsonapiStore, config, status, utils } = createJsonapiStore(api)There are a number of features which are worth explaining in more detail. Many of these can be configured - see the Configuration section.
-
Includes - If the JSONAPI record contains an
includessection, the data in this will be added to the store alongside the 'main' records. (If includes are not used, then you will need to usegetRelatedto fetch relationships). -
Follow relationships -
Relationshipsspecified asdataresources in the JSONAPI data will be added alongside the attributes in the restructured data 'root' as a getter property. Querying this key will return the record from the store, if present. Additionally, helper methods will be added to_jvto make dealing with these easier (see Helper functions) -
Recursive Relationships -
Relationshipscan be recursive - e.g.author => article => blog => author. This can cause infinite recursion problems with anything walking the object (such asJSON.stringify). By default, recursion is detected and stopped when following relationships, with the recursive relationship replaced with a (restructured) resource identifier. -
Preserve JSON - The original JSONAPI record(s) can optionally be preserved in
_jvif needed - for example if you need access tometaor other sections. To avoid duplication, thedatasection (attributes,relationshipsetc) is removed. -
Clean Patches - by default, data passed to the
patchaction is used as-is. IfcleanPatchis enabled, then the patch object is compared to the record in the store (if it exists), and any attributes with identical values are removed. This means that the finalpatchwill only contain new or modified attributes, which is safer and more efficient, as it avoids sending unnecessary or 'stale' data. Additionally, unwanted properties in_jv(links, meta, relationships) can be removed. -
Merging - By default, data returned from the API overwrites records already in the store. However, this may lead to inconsistencies if using Sparse fieldsets or otherwise obtaining only a subset of data from the API. If merging is enabled, then new data will be merged onto existing data. this does however mean that you are responsible for explicitly calling the
deleteRecordaction in cases where attributes have been removed in the API, as they will never be removed from the store, only added to. -
Clear on update - If enabled, then each new set of records is considered to be definitive for that
type, and any other records of thattypein the store will be removed. This option is useful for cases where you expect the API response to contain the full set of records from the server, as it avoids the need for manual cache expiry. The code will first apply the new records to the store, and then for eachtypewhich has had new records added, remove old ones. This is designed to be more efficient in terms of updating computed properties and UI redraws than emptying then repopulating the store. (see Configuration) -
Endpoints - by default this module assumes that object types and API endpoints (item and collection) all share the same name. however, some APIs use plurals or other variations on the endpoint names. You can override the endpoint name via the
axiosurlconfig option or thelinks.selfattribute (see Endpoints) -
JSONPath - the
getDatagetter takes a second (optional) argument which is a JSONPath. This is used to filter the results being returned from the store. (seegetData) -
Searching - The API can be searched without any changes being propagated to the store. This is useful for AJAX-style queries. (see
search) -
Action Status Tracking - The state of any in-flight action can be checked to discover if it has completed yet (successfully or with an error). (See Action Status)
The 2 function types available in Pinia are used as follows:
-
Actions - These are used to query and modify the API, returning the results. Actions are asynchronous.
-
Getters - These are used to directly query the store without contacting the API. Getters are synchronous.
The usual way to use this module is to use actions wherever possible. All actions are asynchronous, and both query the API and update the store, then return data in a normalized form. Actions can be handled using the then/catch methods of promises, or using async/await.
There are 4 actions (with aliases): get (fetch), post (create), patch (update), and delete which correspond to RESTful methods.
Actions are functiosn called from the store, taking 1 or 2 arguments:
The first argument is an object containing restructured data. Actions which take no data argument apart from the record (get and delete) can also accept a URL to fetch, relative to the value of axios baseURL (if set). The leading slash is optional. This means you don't need to create an 'empty' restructured data object to get or delete a record.
The second argument is an (optional) axios config object. This is used to configure axios, most commonly used for adding like headers, URL parameters etc.
Note - The return value of the get action differs in that it returns the results of the action, rather than querying the store for the requested item/collection. This is because the get may be a partial or filtered request, returning only a subset of the item/collection. This means that if you use these results, later updates to the stores will not be reflected. If you want to query the store, then use the get getter once the action has returned.
Some examples:
// To get all items in a collection, using a string path:
store.get('widget').then((data) => {
console.log(data)
})
// axios request query params (JSONAPI options, auth tokens etc)
const params = {
token: 'abcdef123456',
}
// Get a specific record from the 'widget' endpoint, passing parameters to axios:
store.get('widget/1', { params: params }).then((data) => {
console.log(data)
})
// Restructured representation of a record
const newWidget = {
name: 'sprocket',
color: 'black',
_jv: {
type: 'widget',
},
}
// Create a new widget in the API, using a restructured object, and passing parameters to axios:
store.post(newWidget, { params: params }).then((data) => {
console.log(data)
})
// Update a widget in the API
const widgetColor = {
color: 'red',
_jv: {
type: 'widget',
id: '1',
},
}
store.patch(widgetColor, { params: params })
// Fetch, then update a widget in the API
store.get('widget/1', { params: params }).then((widget1) => {
widget1['color'] = 'red'
store.patch(widget1, { params: params })
})
// Delete a widget from the API
store.delete('widget/1', { params: params })The search action is the same as the get action, except that it does not result in any updates to the store. This action exists for efficiency purposes - for example to do 'search-as-you-type' AJAX-style queries without continually updating the store with all the results.
const widgetSearch = (text) => {
const params = { 'filter[text_contains]': text }
store.search('widget', { params: params }).then((data) => {
return data
})
}There are also 4 'relationship' actions: getRelated, postRelated, patchRelated and deleteRelated which modify relationships via an object's relationship URL.
Note - in many cases you may prefer to use the jsonapi server-side include option to get data on relationships included in your original query. (See Relationships).
Like the RESTful actions, this takes 2 arguments - the string or object to be acted on, and an axios config object. It returns a deeply nested restructured tree - relationship -> type -> id -> data.
Note - getRelated only works on specific items, not collections.
By default this action will fetch the record specified from the API, then work out its relationships and also fetch those.
If the argument is a string, it can optionally take a 3rd argument, e.g. type/id/relationship to cause only the named relationship to be followed.
If the argument is an object, then if the object contains a _jv/relationships section, then only these relationships will are followed. If the relationships section contains keys (i.e relationship names) but no values (i.e. resource linkage) then these will be fetched from the API.
// Assuming the API holds the following data
jsonapi = {
data: {
type: 'widget',
id: '1',
},
relationships: {
widgets: {
data: {
type: 'widget',
id: '2',
},
},
doohickeys: {
data: {
type: 'doohickey',
id: '10',
},
},
},
}
// Get all of widget 1's related items (widgets and doohickeys)
store.getRelated('widget/1').then((data) => {
console.log(data)
})
// Get only the items in the 'widgets' relationship
store.getRelated('widget/1/widgets').then((data) => {
console.log(data)
})
// Equivalent, using object instead of string argument
const customRels = {
_jv: {
type: 'widget',
id: '1',
relationships: {
widgets: {
data: {
type: 'widget',
id: '2',
},
},
},
},
}
// Equivalent, but 'doohickeys' resource linkage will be fetched from the server
// i.e. { data: { type: 'doohickey', id: '10' }}
const customRelsNoData = {
_jv: {
type: 'widget',
id: '1',
relationships: {
doohickeys: undefined,
},
},
}
store.getRelated(customRels).then((data) => {
console.log(data)
})The other 3 methods are all for 'writing' to the relationships of an object. they use the relationship URLs of an object, rather than writing to the object itself.
post- adds relationships to an item.delete- removes relationships from an item.patch- replace all relationships for an item.
All methods return the updated item from the API, and also update the store (by internally calling the get action).
These methods take a single argument - an object representing the item, with the '_jv section containing relationships that are to be acted on. For example:
const rels = {
_jv: {
type: 'widget',
id: '1',
relationships: {
widgets: {
data: {
type: 'widget',
id: '2',
},
},
doohickeys: {
data: {
type: 'doohickeys',
id: '10',
},
},
},
},
}
// Adds 'widget/2' and 'doohickey/10' relationships to 'widgets' and 'doohickeys' on 'widget/1'
store.postRelated(rels).then((data) => {
console.log(data)
})
// Removes 'widget/2' and 'doohickey/10' relationships from 'widgets' and 'doohickeys' on 'widget/1'
store.deleteRelated(rels).then((data) => {
console.log(data)
})
// Replaces 'widgets' and 'doohickeys' relationships with just 'widget/1' and 'doohickey/10'
store.patchRelated(rels).then((data) => {
console.log(data)
})There are several actions which modify store state. They are called from the 'API' actions, but can also be called directly.
These actions all take normalised data as an argument.
Deletes a single record from the store.
store.deleteRecord({ _jv: { type: 'widget', id: '1' } })Updates records in the store. Replaces or merges with existing records, depending on the value of the mergeRecords configuration option.
addRecords takes a normalised data object as an argument.
// Update a single record in ths store
store.addRecords({
name: 'sprocket',
color: 'black',
_jv: {
id: '1',
type: 'widget',
}
})
// Update multiple records
store.addRecords({
10: {
name: 'sprocket',
color: 'black',
_jv: {
id: '10',
type: 'widget',
}
},
20: { ... }
})As addRecords, but explicitly replaces existing records.
As addRecords, but explicitly merges onto existing records.
Will remove all records from the store (of a given type) which aren't contained in a given response. Can be set as the default behaviour on updates - see clearOnUpdate.
// Remove all records of type 'widget' from the store
store.clearRecords({ _jv: { type: 'widget' } })Most errors are likely to be those raised by the API in response to the request. These will take the form of an Axios Error Handling object, containing an JSONAPI Error object.
To handle errors with pinia-jsonapi using then/catch methods on the promise:
store.get('/widget/99')
.then((res) => {
// request is successful - normalised jsonapi
console.log(res)
})
.catch((errs) => {
if (errs.response) {
// API error
console.log('HTTP Error Code:', errs.response.status)
// Work with each error from the JSONAPI 'errors' array
for (let err of errs.response.data.errors) {
console.log(err.detail)
}
} else {
// Some other type of error
console.log(errors.message)
}
})Otherwise if you are using async/await:
try {
let res = await store.get('/widget/99')
console.log(res)
} catch(errs) {
<... handle errors here ...>
}The status of actions can be monitored using the status wrapper function, imported from pinia-jsonapi.
status takes as an argument an action dispatch function (or any function which returns a promise). It then calls and tracks the state of that function.
It returns the promise created by the function, with an ID added (_statusID). This ID can be used to get the status of the function via the status.status object:
import { status } from 'pinia-jsonapi'
// Capture the returned promise
let myAction = status.run(() => store.get('widget/1'))
// Make a reference to the ID and status value
let myID = myAction._statusID
let myStatus = status.status[myID]
console.log('myAction Status is now:', myStatus) // Pending
// Handle the promise
myAction
.then((result) => {
console.log('myAction Status is now:', myStatus) // Success
console.log(result)
})
.catch((error) => {
console.log('myAction Status is now:', myStatus) // Error
console.log(error)
})The value for the ID in status.status will be set to one of:
- 0 - Action is Pending
- 1 - Action is Complete (Success)
- -1 - Action is Complete (Error)
These values can be easily overridden if you wish to use the value directly:
// Change the status values at the start
status.PENDING = 'Please wait...'
status.SUCCESS = 'Task completed successfully'
status.ERROR = 'There was an error'You can now easily track status in your UI:
<!-- Displayed once action completes (success or error) -->
<span v-if="myStatus">{{ result }}</span>
<!-- Display only on error -->
<span v-if="myStatus === -1">Error!</span>
<!-- Display the status value directly -->
<span>{{ myStatus }}</span>Note - By default action IDs will always increment. If you have concerns about status.status growing too large, and wish to limit this, see maxStatusID in Configuration
There are 2 getters available: getData and getRelatedData.
NOTE: There is a single namespace for actions and getters in Pinia - so the getter names must be different to the action names.
getData returns information directly from the store for previously cached records. This is useful for performance reasons, or for use in computed properties.
getData returns an object with getters that point to the data in the store. This means that updates to the store will be dynamically reflected in the results object. However it also means that it is not possible to modify this object as getters aren't writeable.
If you wish to modify the results object (e.g. for patching) then you should use the utils.deepCopy method on the object to make a copy that is safe to modify. This deep copies the object, while preserving the Helper Functions.
computed: {
...mapGetters({
// Map 'jv/get' as a computed property 'get'
get: 'jv/get',
}),
// Create a computed property that calls the getter with normalized data
getWidget: function() {
return store.getData({ _jv: { type: 'Widget' } })
},
},Like actions, getData takes an object or string indicating the desired resources. This can be an empty string, type, or type and id, to return the whole store, a collection, or an item.
getData takes an optional 2nd argument - a jsonpath string to filter the record(s) which are being retrieved. See the project page for JSONPath Syntax
// Assuming the store is as follows:
store = {
widget: {
'1': {
name: 'sprocket',
color: 'black',
},
'2': {
name: 'cog',
color: 'red',
},
},
}
// Get all records (of any type) with id = 10 (useful if your API has globally unique UUIDs)
store.getData('', '$.*.10')
// Get all widgets that are red:
store.getData('widget', '$[?(@.color=="red")]')
// Note that filters can create impossible conditions
// The following will return empty, as widget 1 is not red
store.getData('widget/1', '$[?(@.color=="red")]')getRelatedData returns the relations of the specified resource. The resource is specified by either a string, or by a normalized resource object (as in getData). The getter returns an object with each of the resource's relationships as a key. The resources inside the objects relationships JSON-API key are mapped to a getData getter. This means that the resources can be retrieved from the result of getRelatedData once they are loaded into the store (with the getRelated action). If the resources are not loaded into the store yet, only the keys of the related resources will be returned.
For example, to get all widgets related to the widget with id 1:
store.getRelatedData('widget/1')['widgets']Distinguishing between the attributes and relationships in the 'root' is simplified by a number of 'helper' functions which are provided in the _jv (jvtag) object:
-
attrs- a getter property which returns an object containing all attributes. -
rels- a getter property which returns an object containing all relationships. -
isAttr- a function which returns True/False for a given name. -
isRel- a function which returns True/False for a given name. -
isData- this property is created (set totrue) on items returned in 'direct response' to the request. This is used to distinguish between 'direct' and 'included' records. -
isIncluded- this property is created (set totrue) on items returned as anincludedrecord. This is used to distinguish between 'direct' and 'included' records.
These are particularly useful in Vue templates. For example to iterate over an item, picking out just the attributes:
<li v-for="(val, key) in widget._jv.attrs">{{ key }} {{ val }}</li>
<!-- Or -->
<li v-for="(val, key) in widget" v-if="widget._jv.isAttr(key)">{{ key }} {{ val }}</li>Utility Functions API Reference
Some functions are potentially useful for data manipulation etc outside the normal code flow. These functions are exported as utils, i.e:
import { utils } from `pinia-jsonapi`The current utility functions are:
Adds the 'helper' functions/properties to _jv in a restructured object.
addJvHelpers takes a restructured object as its argument, and returns (and modifies in-place) the object to include the helper methods (see Helper functions)
If you wish to clean patches on a per-patch basis, then set the cleanPatch configuration option to false, and instead use this method on your patch record prior to passing it to the action.
cleanPatch takes 3 arguments - the patch data, the state to be compared to, and an array of _jv properties to be preserved (see cleanPatchProps config option).
Makes a deep copy of a normalised object, and adds/updates the Helper functions. This is done because walking the object will normally cause the helper functions to be called, resulting in static (and out-of-date) results.
This function is designed for situations where you wish to modify the results of a getter call, which will throw an error if any of its data is modified.
Note - Be aware that this copy will be a 'static' version of the original object - if the store is subsequently updated, the copied object will no longer reflect this.
Returns an array containing the type, id and rels for a given restructured object (if defined).
Returns the self.links url, or constructs a path from the type and id.
getUrl takes 2 arguments, the restructured object, and optional post boolean (defaults to false). If post is true, then the constructed path will not contain an id.
Convert a JSONAPI object to a restructured object.
Convert a restructured object to a JSONAPI object.
Convert a restructured object to its store form.
A config object can be passed to createJsonapiStore when instantiating. It will override the default options:
const config = { jvtag: '_splat' }
const { jsonapiStore } = createJsonapiStore(api, config)For many of these options, more information is provided in the Usage section.
jvtag- The tag in restructured objects to hold object metadata (defaults to_jv)followRelationshipsData- Whether to follow and expand relationships and store them alongside attributes in the item 'root' (defaults totrue).preserveJson- Whether actions should return the API response json (minusdata) in_jv/json(for access tometaetc) (defaults tofalse)mergeRecords- Whether new records should be merged onto existing records in the store, or just replace them (defaults tofalse).clearOnUpdate- Whether the store should clear old records and only keep new records when updating from a 'collection' get. Applies to thetype(s)associated with the new records. (defaults to false).cleanPatch- If enabled, patch object is compared to the record in the store, and only unique or modified attributes are kept in the patch. (defaults to false).cleanPatchProps- If cleanPatch is enabled, an array of_jvproperties that should be preserved -links,meta, and/orrelationships. (defaults to[]).recurseRelationships- Iffalse, replaces recursive relationships with a normalised resource identifier (i.e{ _jv: { type: 'x', id: 'y' } }). (defaults tofalse).maxStatusID- Sets the highest status ID that will be used instatusbefore rolling round to 1 again. (defaults to-1- no limit).relatedIncludes- When returning the original object frompatch|post|deleteRelatedmethods, also include related objects (defaults totrue).
By default pinia-jsonapi assumes that object type and API endpoint are the same. For example, type: person would have endpoint URLs of /person and /person/1 for collections and single items.
When performing request on an already known single item (like an update), pinia-jsonapi will use the links.self attribute of an item to determine the API endpoint, if it is present.
However many APIs vary how endpoints are named - for example plurals (e.g. type:person, /person/1 and /people), or cases where the endpoint doesn't match the type (e.g. type: person /author and /author/1) or even a combination (e.g. type: person /author/1 and /authors)
To solve this it is possible to override the endpoint for a request by explicitly setting the axios url configuration option:
data = { _jv: { type: 'person' } }
// Default behaviour - will always treat type = itemEP = collectionEP
store.get(data)
// GETs /person
// Explicitly override the endpoint url
store.get(data, { url: '/people' })
store.get(data, { url: '/author/1' })
// Override using a dynamic function
const customUrl = (data) => {
if (data.hasOwnProperty('id')) {
// single item (singular)
return `person/${data.id}`
} else {
// collection (plural)
return '/people'
}
}
store.get({ _jv: { type: 'widget' } }, { url: customUrl(data) })Note - If provided the url option is used as-is - you are responsible for setting a valid collection or item url (with id) as appropriate.
Any bugs, enhancements or questions welcome as Issues and PRs.
Development is currently being done with yarn - NPM should work, but if you hit unexpected issues, please try yarn before filing a bug.
Having cloned this repository, simply run:
yarn
This should pull in all dependencies and development dependencies.
There are several scripts set up in package.json:
yarn unit - Run the unit tests (uses karma, mocha, chai, sinon)
yarn e2e - Run the e2e tests (uses nightwatch)
yarn dev - Runs the example testapp used in e2e testing to allow interactive testing/debugging in a browser.
yarn fakeapiserver - Runs a fake JSONAPI server used by the testapp for interactive testing/debugging.
Note - All code is pre-processed with babel and eslint when testing for backwards compatability and linting.
Please follow these guidelines when writing and submitting code:
-
eslint - This is run over both the main code and the test suite during tests. See
.eslint.rc.jsfor changes to the default rules. -
>= ES6 - Please try to use ES6 and newer methods (matching the policy that
Vuehas). -
Tests - This project aspires to test-driven development. Please submit unit tests (and ideally e2e tests) with all PRs (unless there's a good reason not to).
-
Versioning - Semantic versioning should be used, see https://semver.org for details.
-
Continuous Integration - The project uses Github actions to run tests against all submissions - PRs that are not passing will not be accepted (without good reason).
-
Specific Commits - Please make all commits/PRs as atomic and specific as possible.