html-parse-stringify

This is an _experimental lightweight approach_ to enable quickly parsing HTML into an AST and stringify'ing it back to the original string.

As it turns out, if you can make a the simplifying assumptions about HTML that all tags must be closed or self-closing. Which is OK for _this_ particular application. You can write a super light/fast parser in JS with regex.

"Why on earth would you do this?! Haven't you read: http://stackoverflow.com/questions/1732348/regex-match-open-tags-except-xhtml-self-contained-tags ?!?!"

Why yes, yes I have :)

But the truth is. If you _could_ do this in a whopping grand total of ~600 bytes (min+gzip) as this repo shows. It potentially enables DOM diffing based on a HTML strings to be super light and fast in a browser. What is that you say? DOM-diffing?

Yes.

React.js essentially pioneered the approach. With React you render to a "virtual DOM" whenever you want to, and the virtual DOM can then diff against the real DOM (or the last virtual DOM) and then turn that diff into whatever transformations are necessary to get the _real_ DOM to match what you rendered as efficiently as possible.

As a result, when you're building a single page app, you don't have to worry so much about bindings. Instead, you simple re-render to the virtual DOM whenever you know something's changed. All of a sudden being able to have change events for individual properties becomes less important, instead you can just reference those values in your template whenever you think something changed.

Cool idea, right?!

So why this?

Well, there are other things React expects me to do if I use it that I don't like. Such as the custom templating and syntax you have to use.

If, hypothetically, you could instead diff an HTML string (generated by _whatever_ templating language of your choice) against the DOM, then you'd get the same benefit, sans React's impositions.

This may all turn out to be a bad idea altogether, but initial results seem promising when paired with virtual-dom.

But you can't just diff HTML strings, as simple strings, very easily, in order to diff two HTML node trees you have to first turn that string into a tree structure of some sort. Typically, the thing you generate from parsing something like this is called an AST (abstract syntax tree).

This lib does exactly that.

It has two methods:

1. parse
2. stringify

.parse(htmlString, options)

Takes a string of HTML and turns it into an AST, the only option you can currently pass is an object of registered components whose children will be ignored when generating the AST.

.stringify(AST)

Takes an AST and turns it back into a string of HTML.

What does the AST look like?

See comments in the following example:

``js
var HTML = require('html-parse-stringify')

// this html:
var html = '

console.log(ast)
/*
{
// can be tag, text or component
type: 'tag',

// name of tag if relevant
name: 'div',

// parsed attribute object
attrs: {
class: 'oh'
},

// whether this is a self-closing tag
// such as
voidElement: false,

// an array of child nodes
// we see the same structure
// repeated in each of these
children: [
{
type: 'tag',
name: 'p',
attrs: {},
voidElement: false,
children: [
// this is a text node
// it also has a type
// but nothing other than
// a content containing
// its text.
{
type: 'text',
content: 'hi'
}
]
}
]
}
*/
`

the AST node types

$3

properties:

- type - will always be tag for this type of node
- name - tag name, such as 'div'
- attrs - an object of key/value pairs. If an attribute has multiple space-separated items such as classes, they'll still be in a single string, for example: class: "class1 class2"
- voidElement - true or false. Whether this tag is a known void element as defined by spec.
- children - array of child nodes. Note that any continuous string of text is a text node child, see below.

$3

properties:

- type - will always be text for this type of node
- content - text content of the node

$3

If you pass an object of components as part of the options object passed as the second argument to .parse() then the AST won't keep parsing that branch of the DOM tree when it one of those registered components.

This is so that it's possible to ignore sections of the tree that you may want to handle by another "subview" in your application that handles it's own DOM diffing.

properties:

- type - will always be component for this type of node
- name - tag name, such as 'div'
- attrs - an object of key/value pairs. If an attribute has multiple space-separated items such as classes, they'll still be in a single string, for example: class: "class1 class2"
- voidElement - true or false. Whether this tag is a known void element as defined by spec.
- children - it will still have a children array, but it will always be empty.

changelog

- 3.0.1 Merged #47 which makes void elements check case insensitive. Thanks again, @adrai for this contribution!
- 3.0.0 Merged #46 which fixed an issue with handling of whitespace. Doing major version bump since this changes behavior if you have whitespace only nodes (see merged PR and #45 for more details). Thanks @adrai for this contribution!
- 2.1.1 Merged #41 which fixed an issue with tag nesting. Thanks @ericponto.
- 2.1.0 Merged support for numeric tags. This allows a use case described in this PR. Thanks @kachkaev.
- 2.0.3 Fixed failed publish. Accidentally published an empty package :sweat_smile:
- 2.0.2 Fixed incorrect attribution for vulnerability disclosure. The vulnerability was discovered by Yeting Li. Sam Sanoop was the one who reached out to me about it.
- 2.0.1 Addressing a reported regular expression denial of service issue found by Yeting Li and reported to me by Sam Sanoop of Snyk THANK YOU!. The issue was that sending certain input would cause one of the regular expressions we used to lock up and not finish, freezing the process. See the test that was added for details. To be clear, this lib wasn't meant for parsing non-well formed HTML. But, better safe than sorry! So we're fixing it.
- 2.0.0 updated to more modern dependencies/build system. Switched to prettier, etc. No big feature differences, just new build system/project structure. Added support for top level text nodes thanks to @jperl. Added support for comments thanks to @pconerly.
- 1.0.0 - 1.0.3` no big changes, bug fixes and speed improvements.

credits

If this sounds interesting you should probably follow @HenrikJoreteg and @Philip_Roberts on twitter to see how this all turns out.

license

MIT