This is a fork of the main library. Please consider using that instead.
A library which creates efficient and robust CSS selectors for HTML elements.
The closest you can get if you're looking for a fingerprint of an HTML element
- shortest path and fastest selection in comparison
- configurations allow to define custom options for skip, priority and ignore patterns
- allows single and multiple element as inputs
- provide UMD integration (usage via script, AMD or CommonJS)
- in addition to regular browsers it support the htmlparser2 DOM for virtual environments
- micro library (~ 11kb, no external dependency)
Aside of the prebundled versions the library is also available via npm:
npm install --save optimal-select
import { select } from 'optimal-select' // global: 'OptimalSelect'
document.addEventListener('click', (e) => {
var selector = select(e.target)
console.log(selector)
})
By default following attributes are ignored for robustness towards changes:
- style (inline styles often temporary and used for dynamic visualizations)
- data-reactid (reacts element identifier which depends on the current DOM structure)
- data-react-checksum (react string rendered markup which depends on the current DOM structure)
To define custom filters you can pass the 'ignore' property as a secondary optional parameter.
You can then specify a validation function for the different types (id
, class
, attribute
, tag
).
var selector = select(element, {
// default reference
root: document,
skip (traverseNode) {
// ignore select information of the direct parent
return traverseNode === element.parentNode
},
// define order of attribute processing
priority: ['id', 'class', 'href', 'src'],
// define patterns which should't be included
ignore: {
class (className) {
// disregard short classnames
return className.length < 3
},
attribute (name, value, defaultPredicate) {
// exclude HTML5 data attributes
return (/data-*/).test(name) || defaultPredicate(name, value)
},
// define simplified ignore patterns as a boolean/string/number/regex
tag: 'div'
}
})
As shown the root
property allows to define the container element (default: document
).
The skip
value allows to define a function
, a single node
or an array
of nodes which should be ignored as the selector is created (default: null
). With the priority
value can the order of processed attributes be customized. Finally individual filter functions can be defined through ignore
.
getQuerySelector (input, [options]) // alias: 'select'
Convenience function which automatically uses either getSingleSelector
or getMultiSelector
getSingleSelector(element, [options])
Retrieve a unique CSS selector of the element Element is a DOM ode
getMultiSelector(elements, [options])
Retrieve a unique CSS selector of the elements Elements is an array with a list of DOM nodes
optimize(selector, elements, [options])
Improve the CSS selector
getCommonAncestor(elements, [options])
Retrieve the closest ancestor of the elements
getCommonProperties(elements, [options])
Retrieve a set of common properties of the elements
The latest version of optimal-select
allows the generation and optimization of selectors on virtual environments. It uses the basic structure the htmlparser2 DOM provides and adds some utilities to create the same results as the browser (note: the withDOMLv1
option has to be enabled). Other libraries like cheerio are built on top of these and therefore compatible.
In contrast to the browser does server environments not have a global context which defines their scope. Therefore one can either be specified explicit through a node using the context
options field or automatically extracted from the provided input element. Checkout the example for more details.
- extend documentation
- add automatic tests (e.g. using jsdom)
- improve child-relation and grouping of
getMultiSelector
- define
strict
option for optimizations of multiple elements - check attributes for complex classnames
- fix #8 - Full coverage for "nth-of-type" optimization
- consider
:not
- selector to exclude other elements matching (for multiple element matching consider the :not selector to exclude exceptions)
To build your own version run npm run dev
for development (incl. watch) or npm run build
for production (minified).