xml-zero-lexer
Friendly and forgiving HTML5/XML5/React-JSX lexer/parser with lots of tests. Memory-efficient and Web Worker compatible.
Features
- HTML, XML, and React JSX parsing/lexing
- Zero-copy (returns only indexes into the original string, not copies of values)
- TypeScript
- Lots of tests
- Limited in scope. Easy to audit
- Zero dependencies
"Zero Copy"?
Generally speaking (very generally) the term 'zero copy' refers to a way of parsing input and returning a 'table of contents' data structure providing to you —the programmer— a list of indexes into the input.
The term "Zero Copy" is an attempt to distinguish between 'table of contents' parsing Vs a 'fork the input data, and copy every distinct part into a new data structure without any reference to the original' approach to parsing.
This is all quite philosophical so unless you really care about this then choose a more popular library, please.
Usage
import Lexx, { NodeTypes } from "xml-zero-lexer";
const xml = "<p>his divine shadow</p>";
const tokens = Lexx(xml);
const firstToken = tokens[0];
const isElement = firstToken[0] === NodeTypes.ELEMENT_NODE;
console.log(xml.substring(firstToken[1], firstToken[2]));
const secondToken = tokens[1];
const isTextNode = secondToken[0] === NodeTypes.TEXT_NODE;
console.log(xml.substring(secondToken[1], secondToken[2]));
API
xml-zero-lexer
's default export is a function that can be imported as import Lexx from 'xml-zero-lexer';
.
You can provide this function one or two arguments:
-
(Required) The input string
of HTML, XML, or React JSX;
-
(Optional) { jsx?: boolean; html?: boolean; blackholes?: string[] }
where:
jsx (boolean, optional, default=false): Whether to parse {expression}
in the middle of text / child nodes as a distinct JSX token. React-JSX attributes are always parsed by xml-zero-lexer
so this option is only whether JSX in text / child nodes should be parsed (true
) or treated as text (false
).
html (boolean, optional, default=false): Currently, only affects whether to treat <br>
and <link>
and <img>
(and other HTML self-closing tags) as self-closing tags, affecting whether to return another token of [NodeTypes.CLOSE_ELEMENT]
when parsing these elements.
blackholes (string[], optional, default=["script", "style"]): an array of elements names (typically ["script", "style"]
etc.) that have special parsing rules meaning input of <script> <p> </script>
should be parsed as a script element with a text node of " <p> "
. The special rule is that such an element encompasses everything until it closes with the same element name.
This function returns Token[]
. Each Token
has the shape [NodeType: NODE_TYPE, ...restIndexes: number[] ]
where:
NodeType: NODE_TYPE = number (integer)
. A number representing the type of Node such as 'open element', 'attribute', 'jsx attribute', 'text node', 'close element' and so on. For convenience the NodeTypes
export has named keys for each type NodeTypes.OPEN_ELEMENT
, NodeTypes.CLOSE_ELEMENT
, NodeTypes.TEXT_NODE
, NodeTypes.ATTRIBUTE_NODE
(etc...). This constant can be accessed through the NodeTypes
export eg import { NodeTypes } from 'xml-zero-lexer';
....restIndexes: number[]
an array of integer indexes into the input string that can be used as eg. inputString.substring(restIndexes[0], restIndexes[1])
. Each type of NODE_TYPE
can have a different restIndexes.length
, from 0 to 4. Some NodeType
s also have a variable length, such as NodeTypes.ATTRIBUTE
, because attribute values are optional in HTML5 (ie <input hidden>
lacks a value, and in HTML5 should be treated as a boolean true
, but is essentially valueless for the purposes of parsing). So an input string of <a href="//zombo.com" hidden>
would return:
[
[NodeTypes.ELEMENT_NODE, 1, 2],
[NodeTypes.ATTRIBUTE_NODE, 3, 7, 9, 20],
[NodeTypes.ATTRIBUTE_NODE, 22, 28],
];
Note the variation in length of the ATTRIBUTE_NODE
tokens.
To learn more about the expected outputs for various inputs, please read the tests in index.test.ts
... they're designed to be readable, just look for the const cases = [
variable around line 41.
4KB gzipped.
Want to turn it back into formatted HTML or XML? Try xml-zero-beautify
.
CHANGELOG
If you're upgrading to 3.1.x or greater please note that parsing </>
no longer results in an array with two items of an opening element and a closing element, it now just results in an array with one item of a closing element. This better adheres to expected semantics for React-JSX parsing. Tests have been updated accordingly.
Part of XML-Zero.js