@gadgetinc/react
is a set of React hooks for connecting to a Gadget backend application from a React application. There's a few key features:
- Rule-obeying hooks for reading and writing data from a backend that handle all request lifecycle and auth like
useFindOne
anduseAction
- A full featured, GraphQL-powered nested object selection system using the
select
option - Full type safety for inputs and outputs driven by each Gadget app's backend schema, including over dynamic selections
- Data hydrations that return useful objects like
Date
s
This library wraps urql to communicate with a Gadget-generated API, so it benefits from all of
the same features as urql
as well!
@gadgetinc/react
is a companion package to the JavaScript client package generated for your Gadget app, so you must install the JS client for your app, and then install this package.
To install the JS client for your app, you must set up the Gadget NPM registry, and then install the client:
npm config set @gadget-client:registry https://registry.gadget.dev/npm
# then
yarn add @gadget-client/example-app-slug
# or
npm install @gadget-client/example-app-slug
Full installation instructions can be found in the Gadget docs at https://docs.gadget.dev/api/<my-app-slug>/external-api-calls/installing
.
Once you have your JS client installed, you can install the React hooks library with yarn or npm:
yarn add @gadgetinc/react react
# or
npm install --save @gadgetinc/react react
And finally, you must instantiate an instance of your API client, and then wrap your application in the Provider
component exported from this library. In your root-most React component, you can add a Provider
around the other components of your application:
// import the API client for your specific application from your client package, be sure to replace this package name with your own
import { Client } from "@gadget-client/example-app-slug";
// import the required Provider object and some example hooks from this package
import { Provider } from "@gadgetinc/react";
// instantiate the API client for our app
const api = new Client({ authenticationMode: { browserSession: true } });
export const MyApp = (props) => {
// wrapp the application in the <Provider> so the hooks can find the current client
return <Provider api={api}>{props.children}</Provider>;
};
// import the API client for your specific application from your client package, see your app's installing instructions
import { Client } from "@gadget-client/my-gadget-app";
// import the required Provider object and some example hooks from this package
import { Provider, useAction, useFindMany } from "@gadgetinc/react";
// instantiate the API client for our app
const api = new Client({ authenticationMode: { browserSession: true } });
export function MyComponent() {
// ensure any components which use the @gadgetinc/react hooks are wrapped with the provider, and passed the current api client
return (
<Provider api={api}>
<WidgetDeleter />
</Provider>
);
}
function WidgetDeleter() {
// `useFindMany` executes a backend fetch to get a list of widgets from the backend. we select only the id and name fields of each widget to demonstrate type-safe sub selection
const [result, refresh] = useFindMany(api.widget, {
select: {
id: true,
name: true,
},
});
// `useAction` sets up a mutation we can run to delete a specific widget when a user clicks a button
const [_, deleteWidget] = useAction(api.widget.delete);
// handle all the different states the data fetch can be in
if (result.error) return <>Error: {result.error.toString()}</>;
if (result.fetching && !result.data) return <>Fetching...</>;
if (!result.data) return <>No widgets found</>;
return (
<>
{result.data.map((widget) => (
<button
onClick={(event) => {
event.preventDefault();
void deleteWidget({ id: widget.id }).then(() => refresh());
}}
>
Delete {widget.name}
</button>
))}
</>
);
}
Find more examples in the https://github.com/gadget-inc/examples repo.
Under the hood, your Gadget app's API client and @gadgetinc/react
use a powerful, production-grade GraphQL client called urql. urql has a great client-side data caching feature built-in called Document Caching which allows React components issuing GraphQL requests for the same data to de-duplicate requests and share client-side state. @gadgetinc/react
enables this functionality by default.
@gadgetinc/react
runs urql's Document Caching with a default requestPolicy
of cache-and-network
, which means your React hooks will re-render data with any cached results from the in-memory store, and then make an underlying HTTP request to fetch the most up to date data.
If you want to change the default requestPolicy
that your Gadget API client and React hooks use, you can pass the requestPolicy
option to your API client constructor.
// instantiate the API client for our app that will make network calls for every query, regardless of cache state
const api = new Client({
requestPolicy: "network-only",
});
There are four different request policies that you can use:
cache-first
prefers cached results and falls back to sending an API request when no prior result is cached.cache-and-network
(the default) returns cached results but also always sends an API request, which is perfect for displaying data quickly while keeping it up-to-date.network-only
will always send an API request and will ignore cached results.cache-only
will always return cached results or null.
For more information on urql
's built-in client-side caching, see urql
's docs.
@gadgetinc/react
contains a variety of React hooks for working with your Gadget application's API from a React application. Specific code examples for your application's API can be found within your application's docs at https://docs.gadget.dev/api/
Your React application must be wrapped in the Provider
component from this library for the hooks to function properly. No other wrappers (like urql
's) are necessary.
useFindOne(manager: ModelFinder, id: string, options: SingleFinderOptions = {}): [{data, fetching, error}, refetch]
useFindOne
fetches one record from your Gadget database with a given id
. useFindOne
expects a record with the given id to be found in the backend database, and will return an error in the error
property if no record with this id is found. useFindOne(api.widget, 42)
is the React equivalent of await api.widget.findOne(42)
. Options:
manager
: The model manager you for the model you want to find a record of. The value always comes from an instance of your application's API client instantiated elsewhere. Required. Example:api.widget
, orapi.shopifyProduct
id
: The backend id of the record you want to find. Required.options
: Options for making the call to the backend. Not required, and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useFindOne
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a refetch
function to trigger a refresh of the hook's data.
data: GadgetRecord | null
: The record fetched from the backend. Isnull
while the data is being loaded, or if the record wasn't found.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the request. Will contain an error if the record isn't found byid
. See the Errors section.
Without any options, useFindOne
will fetch the record and cause your component to rerender as the fetch happens and when the data or error arrives:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const ShowWidgetName = (props: { id: string }) => {
const [{ data, fetching, error }, _refetch] = useFindOne(api.widget, props.id);
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return <span className="widget-name">{data.name}</span>;
};
useFindOne
can take options which allow the customization of which fields are returned:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const OnlySomeWidgetFields = (props: { id: string }) => {
// fetch only the widget id and name fields
const [{ data, fetching, error }, _refetch] = useFindOne(api.widget, props.id, { select: { id: true, name: true } });
return (
<span className="widget-name">
{data?.id}: {data?.name}
</span>
);
};
See the refetch
function docs for more information on refreshing the returned server side value.
useMaybeFindOne(manager: ModelFinder, id: string, options: SingleFinderOptions = {}): [{data, fetching, error}, refetch]
useMaybeFindOne
fetches one record from your Gadget database with a given id
. useMaybeFindOne
will return data: null
and error: null
if no record with the given id
is found in the backend database. useMaybeFindOne
otherwise behaves identically to useFindOne
, and accepts the same options.
useMaybeFindOne(api.widget, 42)
is the React equivalent of await api.widget.maybeFindOne(42)
.
useFindMany(manager: ModelFinder, options: ManyFinderOptions = {}): [{data, fetching, error}, refetch]
useFindMany
fetches a page of records from your Gadget database, optionally sorted, filtered, or searched. useFindMany(api.widget, { ... })
is the React equivalent of await api.widget.findMany({ ... })
. Options:
manager
: The model manager you for the model you want to find a page of records for. The value always comes from an instance of your application's API client instantiated elsewhere. Required. Example:api.widget
, orapi.shopifyProduct
options
: Options for making the call to the backend. Not required and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs.filter
: A list of filters to limit the set of returned records. Optional. See the Model Filtering section in your application's API documentation to see the available filters for your models.search
: A search string to match backend records against. Optional. See the Model Searching section in your application's API documentation to see the available search syntax.sort
: A sort order to return backend records by. Optional. See the Sorting section in your application's API documentation for more info.first
&after
: Pagination arguments to pass to fetch a subsequent page of records from the backend.first
should hold a record count andafter
should hold a string cursor retrieved from thepageInfo
of the previous page of results. See the Model Pagination section in your application's API documentation for more info.last
&before
: Pagination arguments to pass to fetch a subsequent page of records from the backend.last
should hold a record count andbefore
should hold a string cursor retrieved from thepageInfo
of the previous page of results. See the Model Pagination section in your application's API documentation for more info.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useFindMany
returns two values: a result object with the data
, fetching
, and error
keys for use in your React component's output, and a refetch
function to trigger a refresh of the hook's data.
data: GadgetRecordList | null
: The resulting page of records fetched from the backend for your model, once they've arrivedfetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the request. See the Errors section.
Without any options, useFindMany
will fetch the first page of backend records sorted by id. As the fetch happens, the hook will cause your component to rerender when the data or error arrives:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const ShowWidgetNames = () => {
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget);
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<ul>
{data.map((widget) => (
<li key={widget.id}>{widget.name}</li>
))}
</ul>
);
};
useFindMany
accepts the select
option to allow customization of which fields are returned:
export const OnlySomeWidgetFields = (props: { id: string }) => {
// fetch only the widget id and name fields
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, { select: { id: true, name: true } });
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<ul>
{data.map((widget) => (
<li key={widget.id}>{widget.name}</li>
))}
</ul>
);
};
useFindMany
accepts a filter
option to limit which records are returned from the backend. For example, we can filter to return only widgets created since the start of 2022:
// fetch only the widgets created recently
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, {
filter: {
createdAt: { greaterThan: new Date(2022, 01, 01) },
},
});
See your application's API documentation for more information on which filters are available on what models.
useFindMany
accepts a sort
option to change the order of the records that are returned. For example, we can sort returned widgets by the createdAt
field:
// return the most recently created widgets first
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, {
sort: {
createdAt: "Descending",
},
});
useFindMany
accepts a search
option to limit the fetched records to only those matching a given search query. For example, we can search all the backend widgets for those matching the string "penny" in any searchable field:
// return the most recently created widgets first
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, {
search: "penny",
});
See your application's API documentation for more information on the search query syntax and which fields are searchable.
useFindMany
accepts pagination arguments for getting the second, third, etc page of results from the backend beyond just the first page. Gadget applications use Relay Cursor style GraphQL pagination, where a second page is fetched by asking for the next x many results after a cursor returned with the first page.
// return the first 10 results after some cursor from somewhere else
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, {
first: 10,
after: "some-cursor-value",
});
// data is a GadgetRecordList object, which has extra properties for inquiring about the pagination state
// the current page's start and end cursor are available for use to then make later requests for different pages
data.endCursor; // => string, used for forward pagination, pass to the `after:` variable
data.startCursor; // => string, used for backwards pagination, pass to the `before:` variable
// data also reports if there are more pages for fetching
data.hasNextPage; // => boolean, true if there is another page to fetch after the `endCursor`
data.hasPreviousPage; // => boolean, true if there is another page to fetch before the `startCursor`
An easy way to do pagination is using React state, or for a better user experience, using the URL with whatever router system works for your application. We use React state to demonstrate pagination in this example:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const WidgetPaginator = () => {
// store the current cursor in the backend data
const [cursor, setCursor] = useState(null);
// pass the current cursor to the `after:` variable, telling the backend to return data after this cursor
const [{ data, fetching, error }, _refetch] = useFindMany(api.widget, { first: 20, after: cursor });
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<>
<ul>
{data.map((widget) => (
<li key={widget.id}>{widget.name}</li>
))}
</ul>
<button
onClick={() => {
// update the cursor, which will trigger a refetch of the next page and rerender with a new `data` object
setCursor(data.endCursor);
}}
disabled={!data.hasNextPage}
>
Next page
</button>
</>
);
};
useFindFirst(manager: ModelFinder, options: FindManyOptions = {}): [{data, fetching, error}, refetch]
useFindFirst
fetches the first record from your backend Gadget database that matches the given filter
, sort
, and search
parameters. useFindFirst(api.widget, { ... })
is the React equivalent of await api.widget.findFirst({ ... })
. useFindFirst
expects a record to be found in the backend, so if no record is found matching the conditions, then the returned result object will data: null
and a MissingDataError
for the error
object. useFindFirst
accepts the following arguments:
manager
: The model manager you for the model you want to find a page of records for. The value always comes from an instance of your application's API client instantiated elsewhere. Required. Example:api.widget
, orapi.shopifyProduct
options
: Options for making the call to the backend. Not required and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs. Optional.filter
: A list of filters to find a record matching. Optional. See the Model Filtering section in your application's API documentation to see the available filters for your models.search
: A search string to find a record matching. Optional. See the Model Searching section in your application's API documentation to see the available search syntax.sort
: A sort order to order the backend records by.useFindFirst
will only return the first record matching the givensearch
andfilter
, sosort
can be used to break ties and select a specific record. Optional. See the Sorting section in your application's API documentation for more info.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useFindFirst
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a refetch
function to trigger a refresh of the hook's data.
data: GadgetRecord | null
: The record fetched from the backend. Isnull
while the data is being loaded, or if a matching record wasn't found.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the request. Will contain an error if the first record isn't found. See the Errors section.
Without any options, useFindFirst
will fetch the first matching record and cause your component to rerender as the fetch happens and when the data or error arrives:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const MostRecentPublishedPost = (props: { id: string }) => {
// request the first record from the backend where publishedAt is set and with the most recent publishedAt value
const [{ data, fetching, error }, _refetch] = useFindFirst(api.blogPost, {
filter: { publishedAt: { isSet: true } },
sort: { publishedAt: "Descending" },
});
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return <span className="banner">Check out our most recent blog post titled {data.title}</span>;
};
useFindFirst
can take options which allow the customization of which fields are returned:
const [_result, _refetch] = useFindOne(api.widget, props.id, {
// fetch only the widget id and name fields
select: {
id: true,
name: true,
},
});
useFindBy(findFunction: ModelFindFunction, fieldValue: any, options: FindByOptions = {}): [{data, fetching, error}, refetch]
useFindBy
fetches one record from your backend looked up by a specific field and value. useFindBy
uses the existing specific-field-finder functions generated within your application's API client, like api.user.findByEmail
or similar. These functions are generated for Gadget fields that have Unique Validations added to them.
useFindBy
expects to be passed a field-specific finder function as the first argument, and then the value of the field to find by as the second argument, and then optionally accepts a third options object. If no record is found matching the conditions, then the returned object will have null
for the data
. useFindBy(api.widget.findByEmail, "[email protected]")
is the React equivalent of api.widget.findByEmail("[email protected]")
. Options:
findFunction
: The model finder function from your application's API client for finding records by a specific field. Gadget generates these finder functions for the fields where they are available. Changes to your Gadget backend schema may be required to get these to exist. Required. Example:api.widget.findBySlug
, orapi.user.findByEmail
.fieldValue
: The value of the field to search for a record using. This is which slug or email you'd pass toapi.widget.findBySlug
orapi.user.findByEmail
.options
: Options for making the call to the backend. Not required and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useFindBy
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a refetch
function to trigger a refresh of the hook's data.
data: GadgetRecord | null
: The record fetched from the backend. Isnull
while the data is being loaded, or if a matching record wasn't found for the givenfieldValue
.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the request. Will contain an error if a matching record isn't found. See the Errors section.
Without any options, useFindBy
will fetch the record with the given field value, and cause your component to rerender as the fetch happens and when the data or error arrives:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
// get a slug from the URL or similar, and look up a post record by this slug
export const PostBySlug = (props: { slug: string }) => {
const [{ data, fetching, error }, _refetch] = useFindBy(api.blogPost.findBySlug, props.slug);
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<>
<h2>{data.title}</h2>
<p>{data.body}</p>
</>
);
};
useFindBy
can take options which allow the customization of which fields are returned:
// fetch only a post id and title fields
const [{ data, fetching, error }, _refetch] = useFindBy(api.blogPost.findBySlug, "some-slug", { select: { id: true, title: true } });
The refetch
function returned as the second element can be executed in order to trigger a refetch of the most up to date data from the backend. See urql
's docs on re-executing queries for more information.
useMaybeFindFirst(manager: ModelFinder, options: FindManyOptions = {}): [{data, fetching, error}, refetch]
useMaybeFindFirst
fetches the first record from your backend Gadget database that matches the given filter
, sort
, and search
parameters. useMaybeFindFirst(api.widget, { ... })
is the React equivalent of await api.widget.maybeFindFirst({ ... })
. useMaybeFindFirst
returns data: null
if no record is found in the backend database, and otherwise works identically to useFindFirst
. See useFindFirst
for more details on the options useMaybeFindFirst
accepts.
useAction(actionFunction: ModelActionFunction, options: UseActionOptions = {}): [{data, fetching, error}, refetch]
useAction
is a hook for running a backend action on one record of a Gadget model. useAction(api.widget.create)
is the React equivalent of await api.widget.create({...})
. useAction
doesn't immediately dispatch a request to run an action server side, but instead returns a result object and a function which runs the action, similar to urql
's useMutation
hook. useAction
must be passed an action function from an instance of your application's generated API client. Options:
actionFunction
: The model action function from your application's API client for acting on records. Gadget generates these action functions for each action defined on backend Gadget models. Required. Example:api.widget.create
, orapi.user.update
orapi.blogPost.publish
.options
: Options for making the call to the backend. Not required and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useAction
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a act
function to actually run the backend action. useAction
is a rule-following React hook that wraps action execution, which means it doesn't just run the action as soon as the hook is invoked. Instead, useAction
returns a configured function that will actually run the action, which you need to call in response to some user event. The act
function accepts the action inputs as arguments -- not useAction
itself. useAction
's result will return the data
, fetching
, and error
details for the most recent execution of the action.
data: GadgetRecord | null
: The record fetched from the backend after a mutation. Isnull
while before the mutation is run and while it is currently ongoing.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the mutation. Will contain an error if the client passed invalid data, if server failed to complete the mutation, or if a network error was encountered. See the Errors section.
For example, we can create a button which creates a post when clicked:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const CreatePost = () => {
// invoke the `useAction` hook, getting back a result object and an action runner function every render
const [{ data, fetching, error }, createBlogPost] = useAction(api.blogPost.create);
// deal with all the states of the result object
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
if (!data) {
return (
<button
onClick={() => {
// run the action when the button is clicked
// the action runner function accepts action inputs in the same format as api.blogPost.create, and the GraphQL API
createBlogPost({
title: "New post created from React",
});
}}
>
Create a post
</button>
);
} else {
return (
<>
<h2>{data.title}</h2>
<p>{data.body}</p>
</>
);
}
};
We can also run actions on existing models by passing the id:
in with the action parameters:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const UpdatePost = (props: { id: string }) => {
// invoke the `useAction` hook, getting back a result object and an action runner function every render
const [{ data, fetching, error }, updateBlogPost] = useAction(api.blogPost.update);
const [title, setTitle] = useState("");
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<form
onSubmit={() => {
// run the action with the action runner function when the form is submitted
// pass the id of the blog post we're updating as one parameter, and the new post attributes as another
updateBlogPost({
id: props.id,
title,
});
}}
>
<label>Title</label>
<input type="text" value={title} onChange={(event) => setTitle(event.target.value)} />
<input type="submit">Submit</input>
</form>
);
};
useAction
returns an error
object that is an ErrorWrapper
. The error
object can have a .networkError
, a list of .executionErrors
, or a list of .validationErrors
which is useful for building client side forms driven by server side validations.
useAction
can take options which allow the customization of which fields are returned:
// fetch only a post id and title fields
const [{ data, fetching, error }, _refetch] = useFindBy(api.blogPost.findBySlug, "some-slug", { select: { id: true, title: true } });
useActionForm(actionFunction: ModelActionFunction | GlobalActionFunction, options: UseActionFormOptions = {}): UseActionFormResult
useActionForm
manages form state for calling actions in your Gadget backend. useActionForm
can fetch the record for editing, manage the state of the fields as the user changes them in a form, track validations and errors, and then call the action with the form data when the user submits the form. useActionForm
can call Actions on models as well as Global Actions. See the Gadget Reference Docs for more information.
actionFunction
: The model action or action function from your application's API client. Gadget generates these action functions for each action or model action defined in your project. Required. Example:api.widget.create
, orapi.user.update
orapi.blogPost.publish
.options
: options for controlling the behavior of the form not an exhaustive list, for additonal properties see gadget reference docs for more information.defaultValues
: Default values to seed the form inputs with. Can be omitted. Mutually exclusive with thefindBy
. Optional.findBy
: Details for automatically finding a record to seed the form values with. When passed as a string, will look up a record with thatid
. When passed an object, will call afindBy<Field>
function on the api object to retrieve a record by that field.select
: Which fields to select from the backend when retrieving initial data withfindBy
. See docs on the select option for more information. Optional.send
: Which fields to send from the form values to the backend for the action. Useful if you want to include fields in your form state for driving UI that shouldn't be sent with the submission. Optional.onSubmit
: Callback called right before data is submitted to the backend action. Optional.onSuccess
: Callback called after a successful submission to the backend action. Passed the action result, which is the object with{data, error, fetching}
keys. Optional.onError
: Callback called after an error occurs finding the initial record or during submission to the backend action. Passed the error, which can be a transport error from a broken network, or a list of validation errors returned by the backend. Optional.
useActionForm
returns an object with properties from react-hook-form
(see docs) and additional properties for handling the action see gadget docs for an exhaustive list.
Example:
import { useActionForm } from "@gadgetinc/react";
import { api } from "../api";
const PostForm = () => {
const { register, submit } = useActionForm(api.post.create);
return (
<form onSubmit={submit}>
<label htmlFor="title">Title</label>
<input id="title" type="text" {...register("post.title")} />
<label htmlFor="content">Content</label>
<textarea id="content" {...register("post.content")} />
<input type="submit" />
</form>
);
};
useGlobalAction(actionFunction: GlobalActionFunction, options: UseGlobalActionOptions = {}): [{data, fetching, error}, refetch]
useGlobalAction
is a hook for running a backend Global Action. useGlobalAction(api.someGlobalAction)
is the React equivalent of await api.someGlobalAction({...})
. useGlobalAction
doesn't immediately dispatch a request to run an action server side, but instead returns a result object and a function which runs the action, similar to urql
's useMutation
hook. useGlobalAction
must be passed one of the global action functions from an instance of your application's generated API client. Options:
globalActionFunction
: The action function from your application's API client. Gadget generates these global action functions for each global action defined in your Gadget backend. Required. Example:api.runSync
, orapi.purgeData
(corresponding to Global Actions namedRun Sync
orPurge Data
).options
: Options for making the call to the backend. Not required and all keys are optional.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docs
useGlobalAction
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a act
function to actually run the backend global action. useGlobalAction
is a rule-following React hook that wraps action execution, which means it doesn't just run the action as soon as the hook is invoked. Instead, useGlobalAction
returns a configured function which you need to call in response to some event. This act
function accepts the action inputs as arguments. useGlobalAction
's result will return the data
, fetching
, and error
details for the most recent execution of the action.
data: Record<string, any> | null
: The data returned by the global action from the backend. Isnull
while before the mutation is run and while it is currently ongoing.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the mutation. Will contain an error if the client passed invalid data, if server failed to complete the mutation, or if a network error was encountered. See the Errors section.
For example, we can create a button which runs a global action called purgeData
when clicked
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const PurgeData = () => {
// invoke the `useGlobalAction` hook, getting back a result object and an action runner function every render
const [{ data, fetching, error }, runPurge] = useGlobalAction(api.purgeData);
// deal with all the states of the result object
if (fetching) return "Loading...";
if (error) return `Error running global action: ${error}`;
if (!data) {
return (
<button
onClick={() => {
// run the global action when the button is clicked
// the action runner function accepts action inputs in the same format as api.purgeData, and the GraphQL API
purgeData({ foo: "bar" });
}}
>
Purge data
</button>
);
} else {
return "Purge completed";
}
};
useGet(singletonModelManager: SingletonModelManager, options: GetOptions = {}): [{data, fetching, error}, refetch]
useGet
fetches a singleton record for an api.currentSomething
style model manager. useGet
fetches one global record, which is most often the current session. useGet
doesn't require knowing the record's ID in order to fetch it. useGet(api.currentSession)
is the React equivalent of await api.currentSession.get()
. useGet
accepts the following options:
singletonModelManager
: The singleton model manager available on the generated API client for your application. The passed model manager must be one of thecurrentSomething
model managers.useGet
can't be used with other model managers that don't have a.get
function. Example:api.currentSession
.options
: Options for making the call to the backend. Not required and all keys are optional.select
: A list of fields and subfields to select. See the Select option docs.requestPolicy
: Theurql
request policy to make the request with. Seeurql
's docspause
: Should the hook make a request right now or not. Seeurql
's docssuspense
: Should this hook suspend when fetching data. See Suspense for more info
useGet
returns two values: a result object with the data
, fetching
, and error
keys for inspecting in your React component's output, and a refetch
function to trigger a refresh of the hook's data.
data: GadgetRecord | null
: The record fetched from the backend. Isnull
while the data is being loaded, or if the record wasn't found.fetching: boolean
: A boolean describing if the hook is currently making a request to the backend.error: Error | null
: An error from the client or server side, if encountered during the request. Will contain an error if the singleton record isn't found. See the Errors section.
useGet(api.currentSession)
retrieves the current global session for the current browser
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const CurrentSessionId = () => {
const [{ data, fetching, error }, _refetch] = useGet(api.currentSession);
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<span>
Current session: {data.id} is {data.state}
</span>
);
};
useGet
can take options which allow the customization of which fields are returned on the selected record:
// fetch only the session id and state fields
const [{ data, fetching, error }, _refetch] = useGet(api.currentSession, { select: { id: true, state: true } });
The refetch
function returned as the second element can be executed in order to trigger a refetch of the most up-to-date data from the backend. See urql
's docs on re-executing queries for more information.
useFetch
is a low-level hook for making an HTTP request to your Gadget backend's HTTP routes. useFetch
preserves client-side authentication information by using api.fetch
under the hood, which means fetches will use the same request identity as other GraphQL API calls using the other hooks.
useFetch
accepts the following arguments:
path
: the server-side URL to fetch from. Corresponds to an HTTP route defined on in your backend Gadget app'sroutes
folderoptions
: options configuring thefetch
call, corresponding exactly to those you might send with a normalfetch
.method
: the request method, like"GET"
,"POST"
, etc. Defaults to"GET"
headers
: the request headers, like{ "content-type": "application/json" }
body
: the request body to send to the server, like"hello"
orJSON.stringify({foo: "bar"})
json
: If true, expects the response to be returned as JSON, and parses it for conveniencestream
: If true, responds with aReadableStream
allowing you to work with the response as it arrives. If "string" or any valid encoding label it will update the responsedata
as a string as it arrives. This last option is ideal when working with response streams from LLMs.sendImmediately
: If true, sends the first fetch on component mount. If false, waits for thesend
function to be called to send a request. Defaults totrue
for GET requests andfalse
for any other HTTP verbs.- See all the
fetch
options on MDN
useFetch
returns a tuple with the current state of the request and a function to send or re-send the request. The state is an object with the following fields:
data
: the response data, if the request was successfulfetching
: a boolean describing if the fetch request is currently in progresserror
: an error object if the request failed in any way
The second return value is a function for sending or resending the fetch request.
Here's an example user component which uses useFetch
to make a request to the routes/users/GET-me.js
backend Gadget route:
export function UserByEmail(props) {
const [{data, fetching, error}, refresh] = useFetch("/users/me", {
method: "GET",
body: JSON.stringify({ email: props.email }})
headers: {
"content-type": "application/json",
}
json: true,
});
if (result.error) return <>Error: {result.error.toString()}</>;
if (result.fetching && !result.data) return <>Fetching...</>;
if (!result.data) return <>No user found with id={props.id}</>;
return <div>{result.data.name}</div>;
}
By default, GET
requests are sent as soon as the hook executes. GET
requests can also be refreshed by calling the second return value to re-send the fetch request and fetch fresh data.
// GET request will be sent immediately, can be refreshed by calling `send()` again
const [{ data, fetching, error }, send] = useFetch("/some/route", { method: "GET" });
// ... sometime later
data; // => will be populated
Other request methods like POST
, DELETE
, etc will not be sent automatically. The request will only be sent when the send
functions is called explicitly, often in a click handler or similar.
// POST requests will not be sent until `send` is called explicitly
const [{ data, fetching, error }, send] = useFetch("/some/route", { method: "POST" });
// ... sometime later
data; // => will still be undefined
await send();
data; // => will be populated
Unlike the useFind*
hooks, useFetch
doesn't automatically parse and return rich data from your HTTP route. By default, useFetch
returns a string of the response for the data
. But, there's a couple convenience options for quickly parsing the response into the shape you need.
Pass the { json: true }
option to expect a JSON response from the server, and to automatically parse the response as JSON.
Pass the { stream: true }
to get a ReadableStream
object as a response from the server, allowing you to work with the response as it arrives. Otherwise, the response will be returned as a string
object.
When possible, the hooks which make requests to your structured GraphQL API should be preferred. Your app's GraphQL API is autogenerated, which means you don't need to wire up custom routes on your backend to serve data. These hooks are also type safe and benefit from automatic client side caching, and useFetch
does not.
@gadgetinc/react
's useFetch
hook calls fetch
under the hood both client side and server side, which means you can use it to make HTTP requests to services other than your Gadget backend. You don't have to use useFetch
to make calls elsewhere, but it is handy for avoiding adding other dependencies to your frontend code.
For example, we can call a third-party JSON API at dummyjson.com
:
export function DummyProducts(props) {
const [{ data, fetching, error }, resend] = useFetch("https://dummyjson.com/products", {
method: "GET",
json: true,
});
if (result.error) return <>Error: {result.error.toString()}</>;
if (result.fetching && !result.data) return <>Fetching...</>;
return <div>{JSON.stringify(result.data.products)}</div>;
}
useFetch
will not send your Gadget API client's authentication headers to third party APIs. It will behave like a normal browser fetch
call, just with the added React wrapper and json: true
option for easy JSON parsing.
By default, useFetch
will immediately issue HTTP requests for GET
s when run. This makes it easy to use useFetch
to retrieve data for use rendering your component right away.
export function GetRequest(props) {
// will automatically send the request when the component renders the first time, as it is a GET
const [{ data, fetching, error }, resend] = useFetch("/products");
// fetching will be `true`
}
useFetch
will not immediately issue HTTP requests for HTTP verbs other than GET
, like POST
, PUT
, etc. The HTTP request will only be sent when you call the returned send
function.
export function PostRequest(props) {
// will not automatically send the request when the component renders, call `send` to issue the request
const [{ data, fetching, error }, send] = useFetch("/products", { method: "POST" });
// fetching will be `false`
}
This behavior can be overridden with the sendImmediately
option. You can avoid sending GET requests on render by passing sendImmediately: false
:
export function DelayedGetRequest(props) {
// will not automatically send the request when the component renders, call `send` to issue the request
const [{ data, fetching, error }, send] = useFetch("/products", { sendImmediately: false });
// fetching will be `false`
}
You can also have POST
or PUT
requests immediately issued by passing sendImmediately: true
:
export function ImmediatePutRequest(props) {
// will automatically send the request when the component renders the first time
const [{ data, fetching, error }, send] = useFetch("/products", { method: "POST", sendImmediately: true });
// fetching will be `true`
}
The select
option allows you to choose which fields and subfields are returned by your Gadget app's GraphQL API. Your app's API supports returning only some fields of each model you request, as well as fields of related models through the Gadget relationship field types. The select
option is an object with keys representing the apiIdentifier
of fields in your Gadget models, and values holding a boolean describing if that field should be selected or not, or a subselection for object-typed fields.
For example, you can limit the fields selected by a finder to only return some fields, lowering the amount of bandwidth used and making your requests faster:
export const OnlySomeWidgetFields = (props: { id: string }) => {
// fetch only the widget id and name fields
const [{ data, fetching, error }, _refetch] = useFindOne(api.widget, props.id, { select: { id: true, name: true } });
return (
<span className="widget-name">
{data?.id}: {data?.name}
</span>
);
};
You can also use the select
option for selecting fields of related models. For example, if we have a backend Blog Post model which has a HasMany
field to a Comment model, we can fetch a blog post and it's related comments:
export const BlogWithComments = (props: { id: string }) => {
// fetch only the widget id and name fields
const [{ data, fetching, error }, _refetch] = useFindOne(api.blogPost, props.id, {
select: {
id: true,
title: true,
body: true,
// fetch the post's `comments` HasMany relationship field on the post
comments: {
edges: {
node: {
id: true,
body: true,
// fetch the author's BelongsTo User relationship field also
author: {
email: true,
},
},
},
},
},
});
if (!data) return null;
return (
<>
<h2>{data.title}</h2>
<ul>
{data.comments.edges.map((edge) => (
<li>
{edge.node.author.email} says {edge.node.body}
</li>
))}
</ul>
</>
);
};
Note: The shape of the options you pass in the select
option matches exactly the shape of the GraphQL API for your application. Gadget applications use Relay-style GraphQL pagination, which means lists of records are accessed using the relatedField: { edges: { node: true } }
style. BelongsTo and HasOne field types are accessed without any intermediate fields.
For TypeScript users, the select
option is fully typesafe, allowing you to typecheck which fields you're fetching from the backend as well as ensure that the fields you render in your components are actually selected.
Running queries or mutations can produce a few different kinds of errors your client side should handle:
- network errors where the browser is unable to connect to the server at all
- validation errors where the client sent information to the server successfully, but the server deemed it invalid and rejected it
- server side errors where the client sent information to the server but the server failed to process it due to a bug or transient issue.
Each of these error cases is broken out on the error
object returned by useAction
(and any of the other hooks). The error
object is an ErrorWrapper
object, which has a number of properties for figuring out exactly what went wrong:
error.message: string
- A top level error message which is always presenterror.networkError: Error | undefined
- An error thrown by the browser when trying to communicate with the servererror.executionErrors: (GraphQLError | GadgetError)[] | undefined
- Any errors thrown by the GraphQL API, like missing parameters or invalid selections, and any errors thrown by the server concerning invalid data or backend processing errors.error.validationErrors: { apiIdentifier: string, message: string }[] | undefined
- Any validation errors returned by the server. A shortcut to accessing the.validationErrors
property of the firstInvalidRecordError
in the.executionErrors
of the outerErrorWrapper
object. Useful for building form validations.
Gadget makes a default selection when you don't pass the select
option to a finder, which will include all the model's scalar fields and a small representation of its related records. This default is also type safe, so you can rely on the returned objects from default finder methods returning type safe results conforming to the default shape. To figure out exactly what your client will select by default for a model, see the documentation for that model in your generated API documentation.
The refetch
function returned as the second return value for some hooks can be executed in order to trigger a refetch of the most up to date data from the backend. This is useful for powering refresh buttons in user-facing UI, or for periodically updating the client side data. See urql
's docs on re-executing queries for more information.
As an example, we could use the refetch
function to power a refresh button in a table:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
export const ShowWidgetNames = () => {
// get the second return value of `useFindMany`, which is the refetch function
const [{ data, fetching, error }, refetch] = useFindMany(api.widget);
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<table>
<tr>
<th>ID</th>
<th>Name</th>
</tr>
{data.map((widget) => (
<tr key={widget.id}>
<td>{widget.id}</td>
<td>{widget.name}</td>
</tr>
))}
<tr colspan="2">
<button onClick={() => void refetch()}>Refresh</button>
</tr>
</table>
);
};
@gadgetinc/react
supports two modes for managing loading states: the fetching
return value, which will be true when making requests under the hood, as well as using <Suspense/>
, React's next generation tool for managing asynchrony. Read more about <Suspense/>
in the React docs.
To suspend rendering when fetching data, pass the suspense: true
option to the useFind*
hooks.
const Posts = () => {
// pass suspense: true, and the component will only render once data has been returned
const [{ data, error }, refresh] = useFindMany(api.post, { suspense: true });
// note: no need to inspect the fetching prop
return (
<>
{data
.map
//...
()}
</>
);
};
All the read hooks support suspense: useFindOne
, useMaybeFindOne
, useFindMany
, useFindFirst
, useMaybeFindFirst
, and useGet
.
suspense: true
is most useful when a parent component wraps a suspending-child with the <Suspense/>
component for rendering a fallback UI while the child component is suspended:
<Suspense fallback={"loading..."}>
<Posts />
</Suspense>
With this wrapper in place, the fallback prop will be rendered while the data is being fetched, and once it's available, the <Posts/>
component will render with data.
Read more about <Suspense/>
in the React docs. suspense: true
uses urql
's suspense support under the hood.
Since this library uses urql
behind the scenes, it provides a few useful exports directly from urql
so that it does not need to be installed as a peer dependency should you need to write custom queries or mutations.
The following are exported from urql
:
Provider
Consumer
Context
useQuery
useMutation
Example usage:
import React from "react";
import { api } from "../api"; // some file that instantiates @gadget-client/your-app-slug
import { Provider, useQuery } from "@gadgetinc/react";
export const ShowWidgetNames = () => {
// find all widgets and the most recently created gizmo related to the widget
const [{ data, fetching, error }, refetch] = useQuery({
query: `
query GetWidgets {
widgets {
edges {
node {
id
name
gizmos(first: 1, sort:{ createdAt: Descending }) {
edges {
node {
createdAt
}
}
}
}
}
}
}
`,
});
if (fetching) return "Loading...";
if (error) return `Error loading data: ${error}`;
return (
<table>
<tr>
<th>ID</th>
<th>Name</th>
<th>Last Gizmo Created</th>
</tr>
{data.widgets.edges.map(({ node: widget }) => (
<tr key={widget.id}>
<td>{widget.id}</td>
<td>{widget.name}</td>
<td>{widget.gizmos.edges.length > 0 ? widget.gizmos.edges[0].node.createdAt : "Does not have Gizmos"}</td>
</tr>
))}
<tr colspan="2">
<button onClick={() => void refetch()}>Refresh</button>
</tr>
</table>
);
};
export const App = () => (
<Provider api={api}>
<ShowWidgetNames />
</Provider>
);
When working with Gadget auth, there are several hooks and components that can help you manage the authentication state of your application.
The Provider
component exported from this library accepts an auth
prop which can be used to configure the relative paths to your app's sign in and sign out endpoints. If you do not provide these paths, the default values of /auth/signin
and /auth/signout
will be used.
The hooks use the Gadget client's suspense: true
option, making it easier to manage the async nature of the hooks without having to deal with loading state.
import { Client } from "@gadget-client/my-gadget-app";
import { Provider } from "@gadgetinc/react";
import { Suspense } from "react";
import App from "./App";
// instantiate the API client for our app
const api = new Client({ authenticationMode: { browserSession: true } });
export function main() {
// ensure any components which use the @gadgetinc/react hooks are wrapped with the Provider and a Suspense component
return (
<Provider api={api} auth={{ signInPath: "/auth/signin", signOutActionApiIdentifier: "signOut" }}>
<Suspense fallback={<>Loading...</>}>
<App />
</Suspense>
</Provider>
);
}
React hooks are available to help you manage the authentication state of your application.
Takes your apps API client as well as an optional GetOptions
object to specify things like custom select
or filter
criteria.
Returns the current session, equivalent to await api.currentSession.get()
or useGet(api.currentSession)
, but uses Promises for an easier interface. Throws a Suspense promise while the session is being loaded.
Takes your apps API client as well as an optional GetOptions
object to specify things like custom select
or filter
criteria.
Returns the current user of the session, if present. For unauthenticated sessions, returns null
. Throws a Suspense promise while the session/user are loading.
Returns an object representing the current authentication state of the session. Throws a Suspense promise while the session is being loaded.
user
- the current User
, if signed in. Similar to useUser
.
session
- the current Session
. Similar to useSession
.
isSignedIn
- set to true
if the session has a user associated with it (signed in), false
otherwise.
import { api } from "./api";
export default function App() {
const user = useUser(api);
const { isSignedIn } = useAuth(api);
const gadgetContext = useGadgetContext();
return (
<>
{ isSignedIn ? <p>Hello, {user.firstName} {user.lastName}}</p> : <a href={gadgetContext.auth.signInPath}>Please sign in</a> }
</>
);
}
Returns a callback that you can call to sign out your current Gadget User
from the current Session
. This calls the configured signOutActionApiIdentifier
action, which is the User
signOut
action by default.
const signOut = useSignOut();
<button onClick={signOut}>Sign Out</button>;
It has two options:
redirectOnSuccess
- configuredtrue
by default, will redirect the current viawindow.location.assign
to thesignInPath
.redirectToPath
- WhenredirectOnSuccess
istrue
, will redirect to this path instead of thesignInPath
.
If you are trying to control the layout of your application based on authentication state, it may be helpful to use the Gadget auth React components instead of, or in addition to, the hooks.
Conditionally renders its children if the current session has a user associated with it, similar to the isSignedIn
property of the useAuth()
hook.
<h1>
Hello<SignedIn>, human</SignedIn>!
</h1>
Conditionally renders its children if the current session has a user associated with it.
<SignedOut>
<a href="/auth/signin">Sign In!</a>
</SignedOut>
Conditionally renders its children if the current session has a user associated with it, or redirects the browser via window.location.assign
if the user is not currently signed in. This component is helpful for protecting front-end routes.
<BrowserRouter>
<Routes>
<Route path="/" element={<Layout />}>
<Route index element={<Home />} />
<Route
path="my-profile"
element={
<SignedInOrRedirect>
<MyProfile />
</SignedInOrRedirect>
}
/>
</Route>
</Routes>
</BrowserRouter>
Conditionally renders its children if the current session has no user associated with it, or redirects the browser via window.location.assign
if the user is not currently signed out. This component is helpful for redirecting front-end routes.
<BrowserRouter>
<Routes>
<Route path="/" element={<Layout />}>
<Route index element={
<SignedOutOrRedirect>
<Home />
</SignedOutOrRedirect>
}>
<Route path="my-profile" element={<MyProfile />}
/>
</Route>
</Routes>
</BrowserRouter>