Awesome

PostgREST Data Provider For React-Admin

PostgREST Data Provider for react-admin, the frontend framework for building admin applications on top of REST/GraphQL services.

Installation

npm install --save @raphiniert/ra-data-postgrest

Citation

If you use the PostgREST Data Provider in your research or project, please consider citing the corresponding paper. Feel free to include the following BibTeX entry in your publications:

@article{SCHEIBLE2024100699,
    title = {PostgREST Data Provider for React-Admin: Bootstrap the creation of user interfaces on top of PostgreSQL databases},
    journal = {Software Impacts},
    volume = {21},
    pages = {100699},
    year = {2024},
    issn = {2665-9638},
    doi = {https://doi.org/10.1016/j.simpa.2024.100699},
    author = {Raphael Scheible}
}

REST Dialect

This Data Provider fits REST APIs using simple GET parameters for filters and sorting. This is the dialect used for instance in PostgREST.

Note: The PostgREST data provider expects the API to include a Content-Range header in the response to getList calls. The value must be the total number of resources in the collection. This allows react-admin to know how many pages of resources there are in total, and build the pagination controls.

Content-Range: posts 0-24/319

If your API is on another domain as the JS code, you'll need to whitelist this header with an Access-Control-Expose-Headers CORS header.

Access-Control-Expose-Headers: Content-Range

Usage

// in src/App.js
import * as React from 'react';
import { Admin, Resource, fetchUtils } from 'react-admin';
import postgrestRestProvider,
     { IDataProviderConfig,
       defaultPrimaryKeys,
       defaultSchema } from '@raphiniert/ra-data-postgrest';

import { PostList } from './posts';

const config: IDataProviderConfig = {
    apiUrl: 'http://path.to.my.api/',
    httpClient: fetchUtils.fetchJson,
    defaultListOp: 'eq',
    primaryKeys: defaultPrimaryKeys,
    schema: defaultSchema
}

const App = () => (
    <Admin dataProvider={postgrestRestProvider(config)}>
        <Resource name="posts" list={PostList} />
    </Admin>
);

export default App;

Adding Custom Headers

The provider function accepts an HTTP client function as second argument. By default, they use react-admin's fetchUtils.fetchJson() as HTTP client. It's similar to HTML5 fetch(), except it handles JSON decoding and HTTP error codes automatically.

That means that if you need to add custom headers to your requests, you just need to wrap the fetchJson() call inside your own function:

import { fetchUtils, Admin, Resource } from 'react-admin';
import postgrestRestProvider from '@raphiniert/ra-data-postgrest';

const httpClient = (url, options = {}) => {
    if (!options.headers) {
        options.headers = new Headers({ Accept: 'application/json' });
    }
    // add your own headers here
    options.headers.set('X-Custom-Header', 'foobar');
    return fetchUtils.fetchJson(url, options);
};

const config: IDataProviderConfig = {
    ...
    httpClient: httpClient,
    ...
}

const dataProvider = postgrestRestProvider(config);

render(
    <Admin dataProvider={dataProvider} title="Example Admin">
        ...
    </Admin>,
    document.getElementById('root')
);

Now all the requests to the REST API will contain the X-Custom-Header: foobar header.

Tip: The most common usage of custom headers is for authentication. fetchJson has built-on support for the Authorization token header:

const httpClient = (url, options = {}) => {
    options.user = {
        authenticated: true,
        token: 'SRTRDFVESGNJYTUKTYTHRG',
    };
    return fetchUtils.fetchJson(url, options);
};

Now all the requests to the REST API will contain the Authorization: SRTRDFVESGNJYTUKTYTHRG header.

Special Filter Feature

As postgRest allows several comparators, e.g. ilike, like, eq... The dataProvider is designed to enable you to specify the comparator in your react filter component:

<Filter {...props}>
    <TextInput label="Search" source="post_title@ilike" alwaysOn />
    <TextInput label="Search" source="post_author" alwaysOn />
    // some more filters
</Filter>

One can simply append the comparator with an @ to the source. In this example the field post_title would be filtered with ilike whereas post_author would be filtered using eq which is the default if no special comparator is specified.

RPC Functions

Given a RPC call as GET /rpc/add_them?post_author=Herbert HTTP/1.1, the dataProvider allows you to filter such endpoints. As they are no view, but a SQL procedure, several postgREST features do not apply. I.e. no comparators such as ilike, like, eq are applicable. Only the raw value without comparator needs to be send to the API. In order to realize this behavior, just add an "empty" comparator to the field, i.e. end source with an @ as in the example:

<Filter {...props}>
    <TextInput label="Search" source="post_author@" alwaysOn />
    // some more filters
</Filter>

Compound primary keys

If one has data resources without primary keys named id, one will have to define this specifically. Also, if there is a primary key, which is defined over multiple columns:

const config: IDataProviderConfig = {
    ...
    primaryKeys: new Map([
        ['some_table', ['custom_id']],
        ['another_table', ['first_column', 'second_column']],
    ]),
    ...
}

const dataProvider = postgrestRestProvider(config);

Custom schema

PostgREST allows to select and switch the database schema by setting a custom header. Thus, one way to use this function would be adding the custom header as a string while using react-admin hooks within meta.schema (compare to next section) or to set it up as function of () => (string) while using the data provider just component based. The latter can be done as follows and gives the opportunity to use some central storage (e.g. localStorage) which can be changed at multiple points of the application:

const config: IDataProviderConfig = {
    ...
    schema: () => localStorage.getItem("schema") || "api",
    ...
}

const dataProvider = postgrestRestProvider(config);

Passing extra headers via meta

Postgrest supports calling functions with a single JSON parameter by sending the header Prefer: params=single-object with your request according to its docs.

Within the data provider one can add any kind of header to the request while calling react-admin hooks, e.g.:

const [create, { isLoading, error }] = useCreate(
    'rpc/my-function',
    {
        data: { ... },
        meta: { headers: { Prefer: 'params=single-object' } },
    }
);

Null sort order

Postgrest supports specifying the position of nulls in sort ordering. This can be configured via an optional data provider parameter:

import { PostgRestSortOrder, IDataProviderConfig } from '@raphiniert/ra-data-postgrest';

const config: IDataProviderConfig = {
    ...
    sortOrder: PostgRestSortOrder.AscendingNullsLastDescendingNullsLast
    ...
}

const dataProvider = postgrestRestProvider(config);

This parameter impacts the getList and getManyReference calls.

It is important to note that null positioning in sort will impact index utilization so in some cases you'll want to add corresponding index on the database side.

You can also override this parameter on a per-query basis by passing nullsfirst: true or nullslast: true in the meta object of the query:

const { data, total, isLoading, error } = useGetList(
    'posts',
    {
        pagination: { page: 1, perPage: 10 },
        sort: { field: 'published_at', order: 'DESC' },
        meta: { nullslast: true }
    }
);

Vertical filtering

Postgrest supports a feature of Vertical Filtering (Columns). Within the react-admin hooks this feature can be used as in the following example:

const { data, total, isLoading, error } = useGetList(
    'posts',
    {
        pagination: { page: 1, perPage: 10 },
        sort: { field: 'published_at', order: 'DESC' }
        meta: { columns: ['id', 'title'] }
    }
);

Further, one should be able to leverage this feature to rename columns:

columns: ['id', 'somealias:title']

, to cast columns:

columns: ['id::text', 'title']

and even get bits from a json or jsonb column"

columns: ['id', 'json_data->>blood_type', 'json_data->phones']

Note: not working for create and updateMany.

Developers notes

The current development of this library was done with node v19.10 and npm 8.19.3. In this version the unit tests and the development environment should work.

License

This data provider is licensed under the MIT License and sponsored by raphiniert.com.