67 lines
4.0 KiB
Markdown
67 lines
4.0 KiB
Markdown
# Parse `data:` URLs
|
||
|
||
This package helps you parse `data:` URLs [according to the WHATWG Fetch Standard](https://fetch.spec.whatwg.org/#data-urls):
|
||
|
||
```js
|
||
const parseDataURL = require("data-urls");
|
||
|
||
const textExample = parseDataURL("data:,Hello%2C%20World!");
|
||
console.log(textExample.mimeType.toString()); // "text/plain;charset=US-ASCII"
|
||
console.log(textExample.body); // Uint8Array(13) [ 72, 101, 108, 108, 111, 44, … ]
|
||
|
||
const htmlExample = parseDataURL("data:text/html,%3Ch1%3EHello%2C%20World!%3C%2Fh1%3E");
|
||
console.log(htmlExample.mimeType.toString()); // "text/html"
|
||
console.log(htmlExample.body); // Uint8Array(22) [ 60, 104, 49, 62, 72, 101, … ]
|
||
|
||
const pngExample = parseDataURL("data:image/png;base64,iVBORw0KGgoAAA" +
|
||
"ANSUhEUgAAAAUAAAAFCAYAAACNbyblAAAAHElEQVQI12P4" +
|
||
"//8/w38GIAXDIBKE0DHxgljNBAAO9TXL0Y4OHwAAAABJRU" +
|
||
"5ErkJggg==");
|
||
console.log(pngExample.mimeType.toString()); // "image/png"
|
||
console.log(pngExample.body); // Uint8Array(85) [ 137, 80, 78, 71, 13, 10, … ]
|
||
```
|
||
|
||
## API
|
||
|
||
This package's main module's default export is a function that accepts a string and returns a `{ mimeType, body }` object, or `null` if the result cannot be parsed as a `data:` URL.
|
||
|
||
- The `mimeType` property is an instance of [whatwg-mimetype](https://www.npmjs.com/package/whatwg-mimetype)'s `MIMEType` class.
|
||
- The `body` property is a `Uint8Array` instance.
|
||
|
||
As shown in the examples above, you can easily get a stringified version of the MIME type using its `toString()` method. Read on for more on getting the stringified version of the body.
|
||
|
||
### Decoding the body
|
||
|
||
To decode the body bytes of a parsed data URL, you'll need to use the `charset` parameter of the MIME type, if any. This contains an encoding [label](https://encoding.spec.whatwg.org/#label); there are [various possible labels](https://encoding.spec.whatwg.org/#names-and-labels) for a given encoding. You can use the [`TextDecoder`](https://developer.mozilla.org/en-US/docs/Web/API/TextDecoder) API for this:
|
||
|
||
```js
|
||
const parseDataURL = require("data-urls");
|
||
|
||
const dataURL = parseDataURL(arbitraryString);
|
||
|
||
// If there's no charset parameter, e.g. if `arbitraryString` is `"data:text/plain,H%C3%A9llo!"`,
|
||
// then let's guess UTF-8.
|
||
const encodingLabel = dataURL.mimeType.parameters.get("charset") ?? "utf-8";
|
||
const decoder = new TextDecoder(encodingLabel);
|
||
|
||
const bodyDecoded = decoder.decode(dataURL.body);
|
||
```
|
||
|
||
(Note that as of the time of this writing in 2026-01, Node.js's built-in `TextDecoder` has many correctness bugs, so we suggest using the polyfill from the [`@exodus/bytes`](https://www.npmjs.com/package/@exodus/bytes) package until they are fixed.)
|
||
|
||
Using the parsed charset is quite important, since [the spec requires](https://fetch.spec.whatwg.org/#data-url-processor) that if no parseable MIME type is given, the default is `"US-ASCII"`, [aka windows-1252](https://encoding.spec.whatwg.org/#note-latin1-ascii)—not UTF-8, like you might asume. So for example, given an `arbitraryString` of `"data:,H%E9llo!"`, the above code snippet will correctly produce a `bodyDecoded` of `"Héllo!"` by using the windows-1252 decoder, whereas if you used a UTF-8 decoder you'd get back `"H<>llo!"`.
|
||
|
||
### Advanced functionality: parsing from a URL record
|
||
|
||
If you are using the [`whatwg-url`](https://www.npmjs.com/package/whatwg-url) package, you may already have a "URL record" object on hand, as produced by that package's `parseURL` export. In that case, you can use this package's `fromURLRecord` export to save a bit of work:
|
||
|
||
```js
|
||
const { parseURL } = require("whatwg-url");
|
||
const dataURLFromURLRecord = require("data-urls").fromURLRecord;
|
||
|
||
const urlRecord = parseURL("data:,Hello%2C%20World!");
|
||
const dataURL = dataURLFromURLRecord(urlRecord);
|
||
```
|
||
|
||
In practice, we expect this functionality only to be used by consumers like [jsdom](https://www.npmjs.com/package/jsdom), which are using these packages at a very low level.
|