H5Web App & Providers

H5Web is a collection of React components to visualize and explore data. It consists of two main packages:

• @h5web/lib: visualization components built with react-three-fiber.
• @h5web/app: a standalone, web-based viewer to explore HDF5 files (this library).

@h5web/app exposes the HDF5 viewer component App, as well as the following built-in data providers:

• H5GroveProvider for use with server implementations based on H5Grove, like jupyterlab-h5web;
• HsdsProvider for use with HSDS;
• MockProvider for testing purposes.

Prerequisites

The react and react-dom dependencies must be installed in your project. Note that as of version 10, @h5web/app requires React 18.

This package supports TypeScript out of the box without the need to install a separate @types/ package.

Getting started 🚀

npm install @h5web/app

import '@h5web/app/styles.css';

import React from 'react';
import { App, MockProvider } from '@h5web/app';

function MyApp() {
  return (
    <div style={{ height: '100vh' }}>
      <MockProvider>
        <App />
      </MockProvider>
    </div>
  );
}

export default MyApp;

Examples

The following code sandboxes demonstrate how to set up and use @h5web/app with various front-end development stacks:

• Vite

Browser support

H5Web works out of the box on Firefox 102 ESR. Support for older versions might be achieved by polyfilling specific web platform features like Object.hasOwn().

API reference

App

Renders the HDF5 viewer.

For App to work, it must be wrapped in a data provider:

<MockProvider>
  <App />
</MockProvider>

sidebarOpen?: boolean (optional)

Whether the viewer should start with the sidebar open. The sidebar contains the explorer and search panels. Defaults to true. Pass false to hide the sidebar on initial render, thus giving more space to the visualization. This is useful when H5Web is embedded inside another app.

<App sidebarOpen={false} />

This replaces prop explorerOpen, which was deprecated in v7.1.0 and removed in v8.0.0.

initialPath?: string (optional)

The path to select within the file when the viewer is first rendered. Defaults to '/'.

<MockProvider>
  <App initialPath="/arrays/threeD" />
</MockProvider>

getFeedbackURL?: (context: FeedbackContext) => string (optional)

If provided, a "Give feedback" button appears in the breadcrumbs bar, which invokes the function when clicked. The function should return a valid URL, for instance a mailto: URL with a pre-filled subject and body: mailto:[email protected]?subject=Feedback&body=<url-encoded-text>. If the app is publicly available, we recommend returning the URL of a secure online contact form instead.

<App getFeedbackURL={() => 'https://my-feedback-form.com'} />

<App
  getFeedbackURL={(context) => {
    const {
      filePath, // path of current file
      entityPath, // path of currently selected entity
    } = context;

    return `mailto:[email protected]?subject=Feedback&body=${encodeURIComponent(...)}`;
  }}
/>

disableDarkMode?: boolean (optional)

By default, the viewer follows your browser's and/or operating system's dark mode setting. This prop disables this behavior by forcing the viewer into light mode.

<App disableDarkMode />

propagateErrors?: boolean (optional)

The viewer has a top-level ErrorBoundary that, by default, handles errors thrown outside of the visualization area. These include errors thrown by the data provider when fetching metadata for the explorer. If you prefer to implement your own error boundary, you may choose to let errors through the viewer's top-level boundary:

import { ErrorBoundary } from 'react-error-boundary';

<ErrorBoundary FallbackComponent={MyErrorFallback}>
  <MockProvider>
    <App propagateErrors />
  </MockProvider>
</ErrorBoundary>;

H5GroveProvider

Data provider for H5Grove.

<H5GroveProvider url="https://h5grove.server.url" filepath="some-file.h5">
  <App />
</H5GroveProvider>

url: string (required)

The base URL of the H5Grove server.

filepath: string (required)

The path and/or name of the file to display in the UI.

fetcher?: Fetcher (optional)

An optional asynchronous function to fetch data and metadata from an h5grove back-end. The function accepts a URL together with some query parameters and options, and is expected to return a promise that resolves to an ArrayBuffer.

If fetcher is not provided, H5GroveProvider creates one based on the native Fetch API using createBasicFetcher but without any form of authentication. In production, we recommend deploying a secure h5grove back-end and initialising a fetcher that sends the expected authentication headers.

If you have to initialise the fetcher during render, make sure to memoise it so the fetching cache isn't cleared every time your app re-renders.

getExportURL?: (...args) => URL | (() => Promise<URL | Blob>) | undefined (optional)

Some visualizations allow exporting the current dataset/slice to various formats. For instance, the Line visualization allows exporting to CSV and NPY; the Heatmap visualization to NPY and TIFF, etc.

For each format, the viewer invokes the provider's getExportURL method. If this method returns a URL or an async function, then the export menu in the toolbar shows an entry for the corresponding export format.

In the case of JSON and CSV, the viewer itself takes care of the export by providing its own "exporter" function to the getExportURL method. When this happens, the getExportURL method just returns a function that calls the exporter.

In the case of NPY and TIFF, H5GroveApi#getExportURL returns a URL so the export can be generated server-side by h5grove.

The optional getExportURL prop is called internally by the getExportURL method and allows taking over the export process. It enables advanced use cases like generating exports from an authenticated endpoint.

// Fetch export data from authenticated endpoint
getExportURL={(format, dataset, selection) => async () => {
  const query = new URLSearchParams({ format, path: dataset.path, selection });
  const response = await fetch(`${AUTH_EXPORT_ENDPOINT}?${query.toString()}`, {
    headers: { /* authentication header */ }
  })

  return response.blob();
}}

// Fetch a one-time export link
getExportURL={(format, dataset, selection) => async () => {
  const query = new URLSearchParams({ format, path: dataset.path, selection });
  const response = await fetch(`${AUTH_TOKEN_ENDPOINT}?${query.toString()}`, {
    headers: { /* authentication header */ }
  })

  // Response body contains temporary, pre-authenticated export URL
  return new URL(await response.body());
}}

// Tweak a built-in export payload in some way (round or format numbers, truncate lines, etc.)
getExportURL={(format, dataset, selection, builtInExporter) => async () => {
  if (!builtInExporter || format !== 'csv') {
    return undefined;
  }

  const csvPayload = builtInExporter();
  return csvPayload.split('\n').slice(0, 100).join('\n'); // truncate to first 100 lines
}}

resetKeys?: unknown[] (optional)

You can pass variables in resetKeys that, when changed, will reset the provider's internal fetch cache. You may want to do this, for instance, when the content of the current file changes and you want the viewer to refetch the latest metadata and dataset values.

It is up to you to decide what sort of keys to use and when to update them. For instance:

• Your server could send over a hash of the file via WebSocket.
• You could show a toast notification with a Refresh button when the file changes and simply increment a number when the button is clicked (cf. contrived example below).

function MyApp() {
  const [key, setKey] = useState(0);
  const incrementKey = useCallback(() => setKey((val) => val + 1), []);

  return (
    <>
      <button type="button" onClick={incrementKey}>
        Refresh
      </button>
      <H5GroveProvider resetKeys={[key]} /* ... */>
        <App />
      </H5GroveProvider>
    </>
  );
}

HsdsProvider

Data provider for HSDS.

<HsdsProvider
  url="https://hsds.server.url"
  username="foo"
  password="abc123"
  filepath="/home/reader/some-file.h5"
>
  <App />
</HsdsProvider>

url: string (required)

The base URL of the HSDS server.

username: string; password: string (required)

The credentials to use to authenticate to the HSDS server. Note that this authentication mechanism is not secure; please do not use it to grant access to private data.

filepath: string (required)

The path of the file to request.

fetcher?: Fetcher (optional)

An asynchronous function to fetch data and metadata from an HSDS back-end. The function accepts a URL together with some query parameters and options, and is expected to return a promise that resolves to an ArrayBuffer. The fetcher must also send the required HSDS authentication headers.

To get you started, if your HSDS back-end is configured with basic HTTP authentication, you can use createBasicFetcher together with buildBasicAuthHeader:

const fetcher = createBasicFetcher({
  headers: buildBasicAuthHeader(USERNAME, PASSWORD),
});

However, beware that this authentication mechanism is not secure — do not use it to grant access to private data.

If you have to initialise the fetcher during render, make sure to memoise it so the fetching cache isn't cleared every time your app re-renders.

getExportURL?: (...args) => URL | (() => Promise<URL | Blob>) | undefined (optional)

See H5GroveProvider#getExportURL.

HsdsProvider doesn't support the NPY and TIFF export formats out of the box.

resetKeys?: unknown[] (optional)

See H5GroveProvider#resetKeys.

MockProvider

Data provider for demonstration and testing purposes.

<MockProvider>
  <App />
</MockProvider>

getExportURL?: (...args) => URL | (() => Promise<URL | Blob>) | undefined (optional)

See H5GroveProvider#getExportURL.

MockProvider doesn't support the NPY and TIFF export formats out of the box.

Utilities

createBasicFetcher: (fetchOpts?: Omit<RequestInit, 'signal'>) => Fetcher

Create a fetcher function based on the native Fetch API. Accepts an optional RequestInit object to configure requests, for instance with an Authentication header:

const fetcher = createBasicFetcher({
  headers: { Authorization: `Bearer ${token}` },
});

To add custom query parameters to the requests:

const basicFetcher = createBasicFetcher();

async function fetcher(...args: Parameters<Fetcher>) {
  const [url, params, opts] = args;
  return basicFetcher(url, { ...params, myOwnParam }, opts);
}

createAxiosFetcher: (axiosInstance: AxiosInstance) => Fetcher

Create a fetcher function from an axios instance. Note that you will need to install axios in your application.

const fetcher = createAxiosFetcher(axios.create({ adapter: 'fetch' }));

function MyApp() {
  //...
  return (
    <H5GroveProvider url={URL} filepath={FILE} fetcher={fetcher}>
      {/*...*/}
    </H5GroveProvider>
  );
}

You can configure the axios instance as you see fit. For instance, you can specify authentication headers, or set up request interceptors to refresh tokens and retry requests automatically. However, do note that some options have no effect, notably url/baseUrl, responseType, signal and onDownloadProgress.

buildBasicAuthHeader: (username: string, password: string) => Record<string, string>

Build an Authorization header for basic HTTP authentication from a username and password.

getFeedbackMailto: (context: FeedbackContext, email: string, subject?) => string

Generate a feedback mailto: URL using H5Web's built-in feedback email template.

(context: FeedbackContext, email: string, subject = 'Feedback') => string;

import { getFeedbackMailto } from '@h5web/app';
...
<App getFeedbackURL={(context) => {
  return getFeedbackMailto(context, '[email protected]');
}} />

enableBigIntSerialization: () => void

Invoke this function before rendering your application to allow the Scalar visualization and metadata inspector to serialize and display big integers:

enableBigIntSerialization();
createRoot(document.querySelector('#root')).render(<MyApp />);

This is recommended if you work with a provider that supports 64-bit integers — i.e. one that may provide dataset and attribute values that include primitive bigint numbers — currently only MockProvider.

The Scalar visualization and metadata inspector rely on JSON.stringify() to render dataset and attribute values. By default, JSON.stringify() does not know how to serialize bigint numbers and throws an error if it encounters one. enableBigIntSerialization() teaches JSON.stringify() to convert big integers to strings:

> JSON.stringify(123n);
TypeError: Do not know how to serialize a BigInt

> enableBigIntSerialization();
> JSON.stringify(123n);
"123n"

The n suffix (i.e. the same suffix used for bigint literals as demonstrated above) is added to help distinguish big integer strings from other strings.

If you're application already implements bigint serialization, you don't need to call enableBigIntSerialization(). Doing so would override the existing implementation, which might have unintended effects.

Context

The viewer component App communicates with its wrapping data provider through a React context called DataContext. This context is available via a custom hook called useDataContext. This means you can use the built-in data providers in your own applications:

<MockProvider>
  <MyApp />
</MockProvider>;

function MyApp() {
  const { filename } = useDataContext();
  return <p>{filename}</p>;
}

useDataContext returns the following object:

interface DataContextValue {
  filepath: string;
  filename: string;

  entitiesStore: EntitiesStore;
  valuesStore: ValuesStore;
  attrValuesStore: AttrValuesStore;
}

The three stores are created with the react-suspense-fetch library, which relies on React Suspense. A component that uses one of these stores (e.g. entitiesStore.get('/path/to/entity')) must have a Suspense ancestor to manage the loading state.

<MockProvider>
  <Suspense fallback={<span>Loading...</span>}>
    <MyApp />
  </Suspense>
</MockProvider>;

function MyApp() {
  const { entitiesStore } = useDataContext();
  const group = entitiesStore.get('/resilience/slow_metadata');
  return <pre>{JSON.stringify(group, null, 2)}</pre>;
}

A common need is to find specific datasets in a file and retrieve their values. You can do so with hooks useDatasets and useValues as follows:

const DATASETS_DEFS = {
  twoD: { path: '/twoD', shape: ShapeClass.Array, type: DTypeClass.Float }
  title: { path: '/title', shape: ShapeClass.Scalar, type: DTypeClass.String }
};

function MyApp() {
  const datasets = useDatasets(DATASETS_DEFS);
  const { twoD, title } = useValues(datasets); // `number[] | TypedArray` and `string` respectively

  // Or if you just need a specific slice from the `twoD` dataset:
  const { slice, title } = useValues({
    slice: { dataset: datasets.twoD, selection: '2,:' },
    title: dataset.title,
  })
}

We also provide two simpler hooks, useEntity and useValue, as well as a large number of type guards and assertion functions to narrow down the kind/shape/type of HDF5 entities returned by useEntity.

const entity = useEntity('/arrays/twoD'); // ProvidedEntity
assertDataset(entity); // Dataset
assertArrayShape(entity); // Dataset<ArrayShape>
assertFloatType(entity); // Dataset<ArrayShape, FloatType>

const value = useValue(entity); // number[] | TypedArray

// Or a specific slice:
const slice = useValue(entity, '2,:');

Once you have a raw value array, you can use the memoised hook useNdArray to wrap it in an ndarray, and then pass it down to a visualization component from @h5web/lib:

const value = useValue(entity); // number[] | TypedArray
const dataArray = useNdArray(value, entity.shape); // NdArray<number[] | TypedArray>
const domain = useDomain(dataArray); // [number, number]

return (
  <HeatmapVis
    style={{ width: '100vw', height: '100vh' }}
    dataArray={dataArray}
    domain={domain}
  />
);

Every store comes with a prefetch method that works like get but doesn't trigger the Suspense boundary and doesn't return a value. If you work with a remote provider like H5Grove and need to access multiple entities/values at once, it's important to prefetch every entity/value first so the requests are done in parallel. useDatasets and useValues do this automatically, but not useEntity and useValue:

const { valuesStore } = useDataContext();
valuesStore.prefetch(abscissasDataset);
valuesStore.prefetch(ordinatesDataset);

const abscissas = useValue(abscissasDataset);
const ordinates = useValue(ordinatesDataset);

To work with HDF5 attributes, retrieve an entity object with useEntity or useDatasets and pass it to findAttribute. Then, you can check or assert its type and shape and retrieve its value with getAttributeValue:

const { attrValuesStore } = useDataContext();
const entity = useEntity('/arrays/twoD'); // ProvidedEntity

// If you just want to know whether the attribute is present
const hasAttr = hasAttribute(entity, 'my_attr'); // boolean

// Otherwise, find it
const attribute = findAttribute(entity, 'my_attr'); // Attribute | undefined

// If the attribute must be present and have the expected shape and type, use type assertions
assertDefined(attribute);
assertArrayShape(attribute);
assertStringType(attribute); // now `Attribute & HasShape<ArrayShape> & HasType<StringType>`

// Otherwise, use type guards and an `if` block
if (
  isDefined(attribute) &&
  hasArrayShape(attribute) &&
  hasStringType(attribute)
) {
  const someStr = getAttributeValue(entity, attribute, attrValuesStore); // string
  someStr.startWith('foo'); // `someStr` is fully type-checked; no need to use `typeof`
}

With scalar string and numeric attributes, use findScalarStrAttr and findScalarNumAttr for convenience:

const strAttr = findScalarStrAttr(entity, 'my_str_attr');
const numAttr = findScalarNumAttr(entity, 'my_num_attr');

assertDefined(strAttr); // or `isDefined` + `if` block
assertDefined(numAttr);

const str = getAttributeValue(entity, strAttr, attrValuesStore); // string
const num = getAttributeValue(entity, numAttr, attrValuesStore); // number | bigint