This package attempts to approach tn-models from a functional paradigm to avoid having issues with types in classes. Thus, preventing doing weird stuff with class static fields which are undiscoverable from TS perspective.
We also prevent runtime obfuscation of fields ( this happened with classes where a field was forgotten to be declared, runtime would have no clue that the field ever was returned from the API). With this in mind, this package warns users that there is a mismatch between what the model declared was vs what the api actually returned.
The package is based in zod to replace models and fields approach from previous version
- Getting started
- API reference
- Roadmap
- Contribution guide
npm i @thinknimble/tn-models-fpyarn add @thinknimble/tn-models-fppnpm i @thinknimble/tn-models-fp/**
*
* // creating a simple user api with login, registration, update
*
*/
import axios from "axios"
import { z } from "zod"
import { GetInferredFromRaw, createCustomServiceCall } from "@thinknimble/tn-models"
/**
* The entity is the default type that is used as an output shape to the prebuilt methods
*/
const userEntity = {
id: z.string().uuid(),
firstName: z.string(),
lastName: z.string(),
email: z.string(),
token: z.string(),
}
/**
* Defining a create shape because the registration api expects a different object
*/
const createShape = {
email: z.string(),
password: z.string(),
firstName: z.string(),
lastName: z.string(),
}
/**
* Since I know my update method can be partial and wont include all the same fields as a create I create an update shape
*/
const updateShape = {
firstName: z.string().optional(),
lastName: z.string().optional(),
}
/**
* login only requires an email and password
*/
const loginShape = {
email: z.string(),
password: z.string(),
}
/**
* Create your api
* Each api has a create, retrieve, list method by default
* They must be enabled by declaring an `entity` in the `models` field
*
* These methods are accessible through the api directly eg:
* userApi.create({})
* userApi.retrieve({})
* userApi.list()
*
*/
/**
* create additional methods using the createCustomServiceCall provider
* these methods are accesible with the shorthand name for csc
* e.g userApi.csc.login({})
*/
const customUpdate = createCustomServiceCall(
{
inputShape: partialUpdateShape,
outputShape: accountShape,
cb: async ({ client, slashEndingBaseUri, input, utils: { toApi, fromApi } }) => {
const { id, ...rest } = toApi(input)
const res = await client.patch(`${slashEndingBaseUri}${id}/`, rest)
return fromApi(res.data)
}
},
)
const login = createCustomServiceCall(
{
inputShape: loginShape,
outputShape: accountShape,
cb: async ({ client, slashEndingBaseUri, input, utils: { toApi, fromApi } }) => {
const data = toApi(input)
const res = await client.post(`api/login/`, rest)
return fromApi(res.data)
}
},
)
/**
* There is no need for an output shape in this case
*/
const deleteEntity = createCustomServiceCall(
{
inputShape: z.string().uuid(),
cb: async ({ client, slashEndingBaseUri, input }) => {
const res = await client.delete(`api/users/${input}/`)
return
}
},
)
const userApi = createApi({
client: axios.create(), // a client of your choice
baseUri: "api/users/", // a base URI to be used as a default
models: {
/**
* if I do not declare any overrides for the three default methods this will be used
*/
entity: accountShape,
/**
*
* Pass a `create` model if you want to override the default input type, otherwise, just passing an entity will generate a default create
* In order to customize the output shape of the default methods you must create a custom call (createCustomServiceCall). That would only be necessary if your declared entity shape type is not what the creation request responds with
*
* */
create: createShape,
},
customCalls:{
// Additional (aka custom calls) methods are declared here
login, update, deleteEntity
}
})
/**
* finally use your api with your favorite wrapper or directly
*/
/**
* This is a utility from TN-Models that is used to return a TS type from the zod shape
* The type can be used anywhere in the code and removes the need for creating one manually
*/
type User = GetInferredFromRaw<typeof scheduleRequestInputShape>
let user: User | null = null
try {
const user = userApi.create({ email: "test@test.com", password: "password", firstName: "first", lastName: "last" })
const res = userApi.csc.login({ email: "random@random.com", password: "iamapassword" })
const userAfterLogin = res.data
} catch (e) {
console.log(e)
}You need a couple of things:
- An
AxiosInstance. Either create it on the fly or provide an existing one - A base uri for your api ( don't worry about trailing slashes we take care of that)
- Models (all optional):
- Model for resource entity: shape of the resource that will be returned by the api. Declaring this model will grant you access to built-in methods
- Model for the input of resource creation (optional, grants you access to
createbuilt-in method)
IG:
import axios from "axios"
import { z } from "zod"
const createShape = {
completed: z.boolean().default(false),
content: z.string().min(1),
completedDate: z.string().datetime().nullable(),
}
export const todoApi = createApi({
client: axios.create(),
baseUri: "api/todo",
models: {
create: createShape,
entity: { id: z.string().uuid(), ...createShape },
},
})import {todoApi} from './services'
import {useMutation} from '@tanstack/react-query'
import {useState} from 'react'
import { Pagination } from '@thinknimble/tn-models-fp'
const TodoManager = () => {
const [selectedTodoId,setSelectedTodoId] = useState()
/**
* Please note the use of TanStack is not required!
*/
const {data: selectedTodo} = useQuery({
queryKey: ['todo',selectedTodoId],
queryFn: () => todoApi.retrieve( selectedTodoId )
})
const {mutateAsync:create} = useMutation({
mutationFn: todoApi.create
})
const [pagination, setPagination] = useState( new Pagination({ page:1 }) )
const { data:currentList } = useQuery({
queryKey: ['todo-list',page],
queryFn: () => todoApi.list( { pagination } )
})
return (
//...
)
}We include a field: customCalls in createApi that can be passed so you can create your own calls.
To do this you pass an object with the service callbacks. These should be created with createCustomServiceCall method:
The createCustomServiceCall method takes the models for your input and output shapes of the call. Then there's a cb field callback that is powered up with multiple arguments that provide you with all the tools we think you need to make a type-safe call
const updatePartial = createCustomServiceCall(
{
inputShape: partialUpdateShape,
outputShape: entityShape,
cb: async ({ client, slashEndingBaseUri, input, utils: { toApi, fromApi } }) => {
const { id, ...rest } = toApi(input)
const res = await client.patch(`${slashEndingBaseUri}${id}`, rest)
return fromApi(res.data)
}
},
)This is the main entrypoint to tn-models.
An api handler can be created with or without custom service calls. Any custom call is provided with a set of utils accordingly to what they're told what the input-output is. These utils allow to convert camelCase->snake_case (toApi) as well as snake_case->camelCase (fromApi).
The result of the function call is an object that allows to do type-safe calls to the given api. It follows as closely as possible the same api as the class-based version of the library.
When passing an entity model to createApi parameter you get a couple of built-in methods as a result.
If you passed a create model you would get the input type resolved from that shape. Otherwise entity will be used as a default
Note that if entity shape is used to resolve the type of the input, any readonly fields and id itself will be stripped from that type. So
const entityShape = {
id: z.string().uuid(),
firstName: z.string(),
lastName: z.string(),
fullName: z.string().readonly(),
}
const api = createApi({
//...
models: {
entity: entityShape,
},
})Will result in the following call signature for the create method:
api.create({
//id: stripped because it is considered readonly regardless of whether you declared it readonly or not
firstName: "sample first name",
lastName: "sample last name",
//fullName: stripped because it is a readonly-declared field
})Returns the resolved type of entity from given models.
Note: Please check that your backend uses the ListMixin in the targetted resource, otherwise this method will not be useful for you.
This method translates as a GET call to the baseUri/ uri.
Returns a paginated version of the resolved type of entity from given models.
This method takes as parameter the resolved type of the entity from given models minus the declared readonly fields which are stripped to keep you from sending them in the request.
There are a couple of flavors of this method to your convenience:
update(partialResource)does a patch request with a partial bodyupdate.replace(fullResource)does a put request with a full bodyupdate.replace.asPartial(partialResource)does a put request with a partial body
zod works as a validation library but it also provides a good set of type utilities that can be used to narrow, infer or define typescript types.
This library sees zod as a library that bridges the gap between typescript world and javascript world. In other words, compile-time and run-time. For this reason it was determined that it would fit perfectly for fulfilling the role of models in this new approach.
Zod is going to be used both as the core tool for our type inference and as a validator parser (for snake_casing requests and camelCasing responses as well as checking whether the type received from api is the same as expected).
What we're using in this approach (and what we would require users to use) are zod raw shapes. Which in plain words are objects which values are
ZodTypeAny: pretty much anything that you can create with zod'sz
Sample models:
import { z } from "zod"
const createShape = {
//ZodRawShape
completed: z.boolean().default(false), //ZodBoolean
content: z.string().min(1), // ZodString
completedDate: z.string().datetime().nullable(), // ZodString
extraInformation: z.object({
// ZodObject
developerUserId: z.string().uuid(), //...
reviewerUserId: z.string().uuid(),
qaUserId: z.string().uuid(),
prDetails: z.object({
url: z.string().url(),
}),
}),
}
const entityShape = { ...createZodRaw, id: z.number() }Usually when using zod we directly create a zod schema (in any of their forms) but here we would like to be a step before the schema itself.
The reason for this decision was based on the fact that we're going to need to convert our schemas from/to camel/snake case. If we were to create a zod schema (object) we would render the shape inaccessible which would deter us from being able to swap its keys to another casing style.
IG:
const myZodObject = z.object({
// zod schema
dateOfBirth: z.string().datetime(),
email: z.string(),
age: z.number(),
}) // after declaration, the shape cannot be retrieved
const myZodShape = {
//zod shape
dateOfBirth: z.string().datetime(),
email: z.string(),
age: z.number(),
} // asking for the shape allow us to do what we please with its keys and later simply call `z.object` internally when we need the zod schemaPS: In a not-that-far future we want to accept zod objects as well as shapes. Or even eradicate shapes altogether, since ZodObjects are the go-to for creating schemas with zod.
You can mark fields as readonly with the built-in zod's .readonly() function. This will use a z.ZodReadonly for your field which will allow the library to identify it as a readonly field, thus preventing those fields to be included in models for creation and update.
const entityShape = {
firstName: z.string(),
lastName: z.string(),
fullName: z.string().readonly(),
}This function is used as a complement to createApi and allows us to create custom service calls attached to the api.
We provided multiple overloads for it to be fully type-safe and properly infer the parameters for you.
Without this function, you cannot add custom service calls. This was designed as to enforce the type safety of the custom calls. If you're looking for a way to self-host one of these calls please check standAlone calls
Example
// from tn-models-client sample repo
const deleteTodo = createCustomServiceCall(
{
inputShape: z.number(), //define your input shape (in this case is a ZodPrimitive)
cb: async ({ input, client, slashEndingBaseUri }) => {
//you get your parsed input, the axios client and the base uri you defined in `createApi`
await client.delete(`${slashEndingBaseUri}${input}`)
}
},
)
const updatePartial = createCustomServiceCall(
{
inputShape: partialUpdateZodRaw, //you can also pass `ZodRawShape`s
outputShape: entityZodRaw,
cb: async ({ client, slashEndingBaseUri, input, utils: { toApi, fromApi } }) => {
// we provide util methods to convert from and to api within your custom call so have you them in handy to use here.
const { id, ...rest } = toApi(input)
const res = await client.patch(`${slashEndingBaseUri}${id}`, rest)
return fromApi(res.data)
}
},
)To add these custom calls to your created api you simply pass them as object to the customCalls field
IG (same as first createApi example but with custom calls)
export const todoApi = createApi(
{
client,
baseUri,
models: {
create: createZodRaw,
entity: entityZodRaw,
},
customCalls: {
// object with declared custom service calls
deleteTodo,
updatePartial,
}
}
)We also added a csc alias in case you feel customServiceCall is too long.
To make calls that include query params a filtersShape object has to be added in the first parameter of createCustomServiceCall. This enables the resulting service call function to include a filter parameter to have readily available the filters to pass them as parameter or parse them in your own way:
Example
const callWithFilter = createCustomServiceCall(
{
inputShape,
outputShape,
filtersShape: {
testFilter: z.string(),
testArrayFilter: z.string().array(),
}
// `parsedFilters` contains your filters ready to be used in the uri. They're snake cased so expect `test_filter` and `test_array_filter`,
cb: async ({ client, slashEndingBaseUri, parsedFilters }) => {
const result = await client.get(slashEndingBaseUri, { params: parsedFilters })
return result.data
}
},
)
const api = createApi(
{
client,
baseUri,
customCalls:{
callWithFilter,
}
}
)
await api.csc.callWithFilter({
input: { testInput: "testInput" },
filters: {
testFilter: "test",
testArrayFilter: [1, 22],
},
})There could be situations where you don't want to attach a call to an api. Probably a one-off request or an rpc-like request which is not attached to a specific resource.
For this case we can use createCustomServiceCall.standAlone which is a function that gets fed a client and is a self-contained version of regular custom service calls.
Here's an example
const standAloneCall = createCustomServiceCall.standAlone({
client: axios,
models: {
outputShape: {
testData: z.string(),
},
inputShape: {
testInput: z.number(),
},
},
name: "standAlone",
cb: async ({
client,
utils,
input,
//!! You don't have the uri available here since this is self hosted there's no other context than this, you'll have to provide the full uri in your api request
// slashEndingBaseUri,
}) => {
const res = await client.post("/api/my-endpoint/", utils.toApi(input))
return utils.fromApi(res.data)
},
})We provide a set of parameters in the custom service callback:
client: a type-wrapped axios instance that makes sure you call the apis with slash ending uris.
For this client to consume your uri strings you should either cast them as const or define them as template strings directly in the call
client.get(`${slashEndingBaseUri}`) // slashEndingBaseUri is an `as const` variable
client.get(`${slashEndingBaseUri}/ending/`) // ✅ define the template string directly in the function call
const uriSampleOutsideOfCall =`${slashEndingBaseUri}my-uri/not-const/`
client.get(uriSample)// ❌ this does not check, you'll get error, template string is already evaluated outside so it is considered `string`
const uriSampleOutsideOfCallAsConst = `${slashEndingBaseUri}my-uri/not-const/` as const
client get(uriSampleOutsideOfCallAsConst)//✅ was cast as const outside of the callslashEndingBaseUri: gives you a reference to the base uri you passed when you created the api so you can use it within the callbackinput:the parsed input based on theinputShapeyou passedutils: set of utilities to convert from and to api models (handles object casing)fromApi: convert a response object from the api (coming in snake casing) to its camelCase versiontoApi: convert an input into an snake_cased object so that you can feed it to the api.
parsedFilters: only available if you provided afiltersShapewhen constructing the custom call. This yields the filters ready-to-go(parsed to string and snake cased!) into theparamsfield of the opts parameter of axios
Allows users to create paginated calls that are not directly related with the GET baseUri/ endpoint of their resource. Such as when an endpoint retrieves a paginated list of things that are not exactly the resource ig: a search.
IG
const getMatches = createPaginatedServiceCall(
{
// inputShape: someInputShape (optional)
outputShape: entityZodShape,
opts: {
uri: "get-matches",
// httpMethod: 'post' (optional, default get)
}
}
)
const api = createApi(
//...models
customCalls:{
getMatches
}
)Add filters to calls by passing a filtersShape in the first parameter. This will allow to pass filters in the resulting service call function
Example
const paginatedCallWithFilters = createPaginatedServiceCall({
outputShape,
filtersShape: {
myExtraFilter: z.string(),
anotherExtraFilter: z.number(),
},
})
const api = createApi(
{
baseUri,
client,
customCalls:{
paginatedCallWithFilters,
}
}
)
const pagination = new Pagination({ page: 1, size: 20 })
const myExtraFilter = "test"
//act
await api.csc.paginatedCallWithFilters({
input: {
pagination,
},
filters: {
myExtraFilter, // besides the pagination qparams this will also pass my_extra_filter as a query param
},
})You can create paginated calls and have their uri to be dynamic.
Previous example simply showed a static uri but adding a urlParams to the inputShape would result in uri to be a builder function:
Example
const callWithUrlParams = createPaginatedServiceCall(
{
inputShape: {
urlParams: z.object({
someId: z.string(),
}),
},
outputShape,
opts: {
uri: ({ someId }) => `myUri/${someId}`,
}
}
)
const api = createApi(
{
baseUri,
client,
customCalls:{
callWithUrlParams,
}
}
)
const pagination = new Pagination({ page: 1, size: 20 })
const randomId = faker.datatype.uuid()
await api.csc.callWithUrlParams({ pagination, urlParams: { someId: randomId } }) // requests myUri/${randomId}Before using this please see createCustomServiceCall.standAlone
This is mostly an internal function and we would recommend using custom calls that have this util included under the hood. If custom calls don't suit your need then this is exported for your convenience
This util allows to create the utils independently without the need of creating the api.
This is useful especially for creating remote procedure calls where no resource is strictly attached and an action is being triggered ig: call to send an email
Creates a collection manager (intended to be) equivalent to the existing class CollectionManager util.
The only required parameter is the fetchList field, which expects a reference from your list function in the created api.
const api = createApi({
//...
})
const collectionManager = createCollectionManager({
fetchList: api.list,
list: [], // your feed list, type-inferred from api.list
pagination: feedPagination, // your pagination object
filters: feedFilters, // inferred from api.list
})Check out the Issues tab for incoming features and feature/request.
Submit your own if you feel there's something we're missing!
We're using pnpm while developing this library. You can easily get setup with it by doing
npm i -g pnpmWe always make sure that our PRs pass all current tests and it is highly recommended to at least add a set of tests that are related with the work of the given PR so that we get a minimum level of confidence that things are working fine.
To run tests we have a dev command which will scan the whole repo and run the tests it finds.
pnpm test:devWe can also isolate tests so that we tackle one at a time by adding .only calls to the test suite (describe) and/or the test itself (it). There's also the chance to run a single file as test
Run a single test file
pnpm test:dev ./src/api/tests/create-api.test.tsIsolate a test suite/test with only:
//some.test.ts
describe.only("some test suite", () => {
it.only("tests some functionality", () => {
//...
})
//...
})If you want to debug this library from an external application, you can run the watch command (below) so any changes would generate a new build, that is it will generate the lib files in /dist
pnpm run dev:watch
These build files enable us to refer to this package in an external app's package.json as a file:
// package.json of external application
//...
dependencies:{
//...
"@thinknimble/tn-models":"file:../path/to/tn-models"
//...
}
//...If you don't have an app and want a very simple one where you can test out tn-models you can check out this one which we set up for this sole purpose: tn-models-script-app. Please follow the README in there to get the app set up.
To make our life easier with package versioning and releases we're using changeset.
If a PR for a feature conveys a release with it OR you want to release a version after some PRs have been merged.
From the root of the project you have to
pnpm changeset
Follow the prompts:
- Give it a patch/minor/major depending on the release you mean to publish.
- Add a description of what the release contains.
Commit those changes into the PR or create a PR for it and merge it.
What this will do is create a Version release PR that will allow you to confirm the release. Once that PR is merged, the github action will reach out to npm and publish the package for you.