Generate URL-safe representations of some arbtirary JSON data in as small a space as possible that can be shared in a bookmark / link.
Although designed to work in Node, a standalone client-side library is provided that can be used directly on the browser.
var codec = require('json-url')('lzw');
var obj = { one: 1, two: 2, three: [1,2,3], four: 'red pineapples' };
codec.compress(obj).then(result => console.log(result));
/* Result: woTCo29uZQHCo3R3bwLCpXRocmVlwpMBAgPCpGZvdXLCrsSOZCBwacSDYXBwbGVz */
var codec = require('json-url')('lzma');
codec.decompress(someCompressedString).then(json => { /* operate on json */ })
var codec = require('json-url')('lzstring');
codec.stats(obj).then(
({ rawencoded, compressedencoded, compression }) => {
console.log(`Raw URI-encoded JSON string length: ${rawencoded}`);
console.log(`Compressed URI-encoded JSON string length: ${compressedencoded}`);
console.log(`Compression ratio (raw / compressed): ${compression}`);
}
);
<script type="text/javascript" src="/dist/browser/json-url.js"></script>
<script>
const lib = JsonUrl('lzma'); // JsonUrl is added to the window object
lib.compress(parsed).then(output => { result.value = output; });
</script>
To see it in action, download the source code and run npm run example, or simply visit this link.
- The browser bundle is generated using Webpack and consists of multiple chunks, with the main chunk entry point located at
dist/browser/json-url.js. Chunks must be located in the same folder as the main module itself. - I've tried my best to reduce the bundle sizes, but the module (at least the entry) is still surprisingly large on the browser (56kb minified, 20kb gzipped). Most of it is due to the buffer.js shim, and partly due to the regenerator-runtime - I may revisit this later to try and improve efficiency.
- Although not all algorithms are asynchronous, all functions return Promises to ensure compatibility.
- Instantiate an instance with appropriate compression codec before using.
- Valid codecs:
- lzw
- lzma
- lzstring - runs lzstring against a stringified JSON instead of using MessagePack on JSON
- pack - this just uses MessagePack and converts the binary buffer into a Base64 URL-safe representation, without any other compression
Typically when you want to shorten a long URL with large amounts of data parameters, the approach is to generate a "short URL" where compression is achieved by using a third-party service which stores the true URL and redirects the user (e.g. bit.ly or goo.gl).
However, if you want to:
- share bookmarks with virtually unlimited combinations of state and/or
- want to avoid the third-party dependency
you would encode the data structure (typically JSON) in your URL, but this often results in very large URLs.
This approach differs by removing that third-party dependency and encodes it using common compression algorithms such as LZW or LZMA.
Note: It is arguable that a custom dictionary / domain specific encoding would ultimately provide better compression, but here we want to
- avoid maintaining such a dictionary and/or
- retain cross-application compatibility (otherwise you need a shared dictionary)
I explored several options, the most popular one being MessagePack. However, I noticed that it did not give the best possible compression as compared to LZMA and LZW.
At first I tried to apply the binary compression directly on a stringified JSON, then I realised that packing it first resulted in better compression.
For small JS objects, LZW largely outperformed LZMA, but for the most part you'd probably be looking to compress large JSON data rather than small amounts (otherwise a simple stringify + base64 is sufficient). You can choose to use whatever codec suits you best.
In addition, there is now support for LZSTRING, although the URI encoding still uses urlsafe-base64 because LZSTRING still uses unsafe characters via their compressToURIEncodedString method - notably the + character
Finally, I went with urlsafe-base64 to encode it in a URL-friendly format.
Find a way to improve bundle sizes for browser usage.