@thi.ng/sax

@thi.ng/sax

This project is part of the @thi.ng/umbrella monorepo.

About

@thi.ng/transducers-based, SAX-like, non-validating, configurable, speedy & tiny XML parser (~1.8KB gzipped).

Unlike the classic event-driven approach of SAX, this parser is implemented as a transducer function, transforming an XML input into a stream of SAX-event-like objects. Being a transducer, the parser can be used in novel ways as part of a larger processing pipeline and can be composed with other pre or post-processing steps, e.g. to filter or transform element / attribute values or only do partial parsing with early termination based on some condition.

Additionally, since by default the parser emits any children as part of "element end" events, it can be used like a tree-walking DOM parser as well (see SVG parsing example further below). The choice is yours!

Installation

yarn add @thi.ng/sax

Dependencies

• @thi.ng/transducers
• @thi.ng/transducers-fsm

Usage examples

import * as sax from "@thi.ng/sax";
import * as tx from "@thi.ng/transducers";

src=`<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE foo bar>
<!-- comment -->
<a>
    <b1>
        <c x="23" y="42">ccc
            <d>dd</d>
        </c>
    </b1>
    <b2 foo="bar" />
</a>`

doc = [...tx.iterator(sax.parse(), src)]

// (see description of `type` values further below)

// [ { type: 0,
//     tag: 'xml',
//     attribs: { version: '1.0', encoding: 'utf-8' } },
//   { type: 1, body: 'foo bar' },
//   { type: 2, body: ' comment ' },
//   { type: 4, tag: 'a', attribs: {} },
//   { type: 6, tag: 'a', body: '\n    ' },
//   { type: 4, tag: 'b1', attribs: {} },
//   { type: 6, tag: 'b1', body: '\n        ' },
//   { type: 4, tag: 'c', attribs: { x: '23', y: '42' } },
//   { type: 6, tag: 'c', body: 'ccc\n            ' },
//   { type: 4, tag: 'd', attribs: {} },
//   { type: 6, tag: 'd', body: 'dd' },
//   { type: 5, tag: 'd', attribs: {}, children: [], body: 'dd' },
//   { type: 5,
//     tag: 'c',
//     attribs: { x: '23', y: '42' },
//     children: [ [Object] ],
//     body: 'ccc\n            ' },
//   { type: 5,
//     tag: 'b1',
//     attribs: {},
//     children: [ [Object] ],
//     body: '\n        ' },
//   { type: 4, tag: 'b2', attribs: { foo: 'bar' } },
//   { type: 5, tag: 'b2', attribs: { foo: 'bar' } },
//   { type: 5,
//     tag: 'a',
//     attribs: {},
//     children: [ [Object], [Object] ],
//     body: '\n    ' } ]

Partial parsing & result post-processing

As mentioned earlier, the transducer nature of this parser allows for its easy integration into larger transformation pipelines. The next example parses an SVG file, then extracts and selectively applies transformations to only the <circle> elements in the first group (<g>) element. Btw. The transformed elements can be serialized back into SVG syntax using @thi.ng/hiccup...

Given the composed transducer below, parsing stops immediately after the first <g> element is complete. This is because the matchFirst() transducer will cause early termination once that element has been processed.

svg=`
<?xml version="1.0"?>
<svg version="1.1" height="300" width="300" xmlns="http://www.w3.org/2000/svg">
    <g fill="yellow">
        <circle cx="50.00" cy="150.00" r="50.00" />
        <circle cx="250.00" cy="150.00" r="50.00" />
        <circle cx="150.00" cy="150.00" fill="rgba(0,255,255,0.25)" r="100.00" stroke="#ff0000" />
        <rect x="80" y="80" width="140" height="140" fill="none" stroke="black" />
    </g>
    <g fill="none" stroke="black">
        <circle cx="150.00" cy="150.00" r="50.00" />
        <circle cx="150.00" cy="150.00" r="25.00" />
    </g>
</svg>`;

[...tx.iterator(
    tx.comp(
        // transform into parse events (see parser options below)
        sax.parse({ children: true }),
        // match 1st group end
        tx.matchFirst((e) => e.type == sax.Type.ELEM_END && e.tag == "g"),
        // extract group's children
        tx.mapcat((e) => e.children),
        // select circles only
        tx.filter((e) => e.tag == "circle"),
        // transform attributes
        tx.map((e)=> [e.tag, {
            ...e.attribs,
            cx: parseFloat(e.attribs.cx),
            cy: parseFloat(e.attribs.cy),
            r:  parseFloat(e.attribs.r),
        }])
    ),
    svg
)]
// [ [ 'circle', { cx: 50, cy: 150, r: 50 } ],
//   [ 'circle', { cx: 250, cy: 150, r: 50 } ],
//   [ 'circle', { cx: 150, cy: 150, fill: 'rgba(0,255,255,0.25)', r: 100, stroke: '#ff0000' } ] ]

DOM-style tree parsing using defmulti

import { defmulti, DEFAULT } from "@thi.ng/defmulti";

// coerces given attribute IDs into numeric values and
// keeps all other attribs
const numericAttribs = (e, ...ids: string[]) =>
    ids.reduce(
        (acc, id) => (acc[id] = parseFloat(e.attribs[id]), acc),
        { ...e.attribs }
    );

// returns iterator of parsed & filtered children of given element
// (iterator is used to avoid extraneous copying at call sites)
const parsedChildren = (e) =>
    tx.iterator(
        tx.comp(
            tx.map(parseElement),
            tx.filter((e)=> !!e),
        ),
        e.children
    );

// define multiple dispatch function, based on element tag name
const parseElement = defmulti((e) => e.tag);

// tag specific implementations
parseElement.add("circle", (e) =>
    [e.tag, numericAttribs(e, "cx", "cy", "r")]);

parseElement.add("rect", (e) =>
    [e.tag, numericAttribs(e, "x", "y", "width", "height")]);

parseElement.add("g", (e) =>
    [e.tag, e.attribs, ...parsedChildren(e)]);

parseElement.add("svg", (e) =>
    [e.tag, numericAttribs(e, "width", "height"), ...parsedChildren(e)]);

// implementation for unhandled elements
parseElement.add(DEFAULT, () => null);

// using the same SVG source as in previous example:
// the `last()` reducer just returns the ultimate value
// which in this case is the SVG root element's ELEM_END parse event
// this also contains all children (by default)
parseElement(tx.transduce(sax.parse(), tx.last(), svg));

// ["svg",
//     {
//         version: "1.1",
//         height: 300,
//         width: 300,
//         xmlns: "http://www.w3.org/2000/svg"
//     },
//     ["g",
//         { fill: "yellow" },
//         ["circle", { cx: 50, cy: 150, r: 50 }],
//         ["circle", { cx: 250, cy: 150, r: 50 }],
//         ["circle",
//             {
//                 cx: 150,
//                 cy: 150,
//                 fill: "rgba(0,255,255,0.25)",
//                 r: 100,
//                 stroke: "#ff0000"
//             }],
//         ["rect",
//             {
//                 x: 80,
//                 y: 80,
//                 width: 140,
//                 height: 140,
//                 fill: "none",
//                 stroke: "black"
//             }]],
//     ["g",
//         { fill: "none", stroke: "black" },
//         ["circle", { cx: 150, cy: 150, r: 50 }],
//         ["circle", { cx: 150, cy: 150, r: 25 }]]]

Error handling

If the parser encounters a syntax error, an error event value incl. a description and input position will be produced (but no JS error will be thrown) and the entire transducer pipeline stopped.

[...tx.iterator(sax.parse(), `a`)]
// [ { type: 7, body: 'unexpected char: \'a\' @ pos 1' } ]

[...tx.iterator(sax.parse(), `<a><b></c></a>`)]
// [ { type: 4, tag: 'a', attribs: {} },
//   { type: 4, tag: 'b', attribs: {} },
//   { type: 7, body: 'unmatched tag: c @ pos 7' } ]

Emitted result type IDs

The type key in each emitted result object is a TypeScript enum with the following values:

ID	Enum	Description
0	Type.PROC	Processing instruction incl. attribs
1	Type.DOCTYPE	Doctype declaration body
2	Type.COMMENT	Comment body
3	Type.CDATA	CDATA content
4	Type.ELEM_START	Element start incl. attributes
5	Type.ELEM_END	Element end incl. attributes, body & children
6	Type.ELEM_BODY	Element text body
7	Type.ERROR	Parse error description

Parser options

Option	Type	Default	Description
children	boolean	true	If true, recursively includes children elements in ELEM_END events. For very large documents, this should be disabled to save (or even fit into) memory.
entities	boolean	false	If true, unescape standard XML entities in body text and attrib values.
trim	boolean	false	If true, trims element body, comments and CDATA content. If the remaining string is empty, no event will be generated for this value.

Authors

• Karsten Schmidt

License

© 2018 Karsten Schmidt // Apache Software License 2.0