Because I had so much trouble finding a good auto-linking implementation out in
the wild, I decided to roll my own. It seemed that everything I found was either
an implementation that didn't cover every case, had many false positives linked,
or was just limited in one way or another.
So, this utility attempts to handle everything. It:
Autolinks URLs, whether or not they start with the protocol (i.e. 'http://').
In other words, it will automatically link the text "google.com", as well as
"http://google.com". Will also autolink IPv4 addresses.
Will properly handle URLs with query params, anchors,
and special characters, and not include chars like a trailing . at the end
of a sentence, or a ) char if the URL is inside parenenthesis.
Will autolink email addresses
Will autolink phone numbers
Will autolink mentions (Twitter, Instagram, Soundcloud, TikTok, Youtube)
Will autolink hashtags (Twitter, Instagram, Facebook, TikTok, Youtube)
Won't clobber URLs with hash anchors by treating them as hashtags (like some other libraries do). For example: google.com/#anchor is properly linked.
Will properly handle HTML input. The utility will not overwrite an href
attribute inside anchor (<a>) tags (or any other tag/attribute), and will not
accidentally wrap the inner text of <a>/<script>/<style> tags with a new
one (which cause doubly-nested anchor tags, or mess with scripts)
Will do all of this in O(n) (linear) time with low constant factors and without the possibility of RegExp Catastrophic Backtracking, making it extremely fast and unsusceptible to pathological inputs.
Simply clone this repository or download a zip of the project, and link to
either dist/Autolinker.js or dist/Autolinker.min.js with a <script> tag.
Importing Autolinker
ES6+/TypeScript/Webpack/Node.js ESM:
import Autolinker from 'autolinker';
Node.js CommonJS:
const Autolinker = require('autolinker');// note: npm wants an all-lowercase package name, but the utility is a class and// should be aliased with a capital letter
Browser
<!-- 'Autolinker.js' or 'Autolinker.min.js' - non-minified is better for debugging, minified is better for users' download time --><script src="path/to/autolinker/dist/Autolinker.min.js"></script>
const autolinker = new Autolinker([ options ]);const linkedText = autolinker.link(textToAutoLink);
Note: if using the same options to autolink multiple pieces of html/text, it's
slightly more efficient to create a single Autolinker instance, and run the
link()
method repeatedly (i.e. use the "class" form above).
Examples:
const linkedText = Autolinker.link("Check out google.com");// Produces: "Check out <a href="http://google.com" target="_blank" rel="noopener noreferrer">google.com</a>"const linkedText = Autolinker.link("Check out google.com", { newWindow: false});// Produces: "Check out <a href="http://google.com">google.com</a>"
Options
The following are the options which may be specified for linking. These are
specified by providing an Object as the second parameter to Autolinker.link().
These include:
newWindow : boolean true to have the links should open in a new window when clicked, false
otherwise. Defaults to true.
urls : boolean/Object true to have URLs auto-linked, false to skip auto-linking of URLs. Defaults
to true.
This option also accepts an Object form with 3 properties to allow for
more customization of what exactly gets linked. All default to true:
schemeMatches (boolean): true to match URLs found prefixed with a scheme,
i.e. http://google.com, or other+scheme://google.com, false to
prevent these types of matches.
tldMatches: true to match URLs with known top level domains (.com, .net,
etc.) that are not prefixed with a scheme (i.e. 'http://'). Ex: google.com,
asdf.org/?page=1, etc. Set to false to prevent these types of matches.
ipV4Matches (boolean): true to match IPv4 addresses. Ex: 192.168.0.1.
false to prevent these types of matches. Note that if the IP address had
a prefixed scheme (such as 'http://'), and schemeMatches is true, it
will still be linked.
email : boolean true to have email addresses auto-linked, false to skip auto-linking of
email addresses. Defaults to true.
phone : boolean true to have phone numbers auto-linked, false to skip auto-linking of
phone numbers. Defaults to true.
mention : string
A string for the service name to have mentions (@username) auto-linked to. Supported values at this time are 'twitter', 'soundcloud', 'instagram', 'tiktok', and 'youtube'. Pass false to skip auto-linking of mentions. Defaults to false.
hashtag : boolean/string
A string for the service name to have hashtags auto-linked to. Supported values at this time are 'twitter', 'facebook', 'instagram', 'tiktok', and 'youtube'. Pass false to skip auto-linking of hashtags. Defaults to false.
stripPrefix : boolean true to have the 'http://' (or 'https://') and/or the 'www.'
stripped from the beginning of displayed links, false otherwise.
Defaults to true.
This option also accepts an Object form with 2 properties to allow for
more customization of what exactly is prevented from being displayed.
Both default to true:
scheme (boolean): true to prevent the scheme part of a URL match
from being displayed to the user. Example: 'http://google.com'
will be displayed as 'google.com'. false to not strip the
scheme. NOTE: Only an 'http://' or 'https://' scheme will be
removed, so as not to remove a potentially dangerous scheme (such
as 'file://' or 'javascript:').
www (boolean): true to prevent the 'www.' part of a URL match
from being displayed to the user. Ex: 'www.google.com' will be
displayed as 'google.com'. false to not strip the 'www'.
stripTrailingSlash : boolean true to remove the trailing slash from URL matches, false to keep
the trailing slash. Example when true: http://google.com/ will be
displayed as http://google.com. Defaults to true.
truncate : number/Object
A number for how many characters long URLs/emails/Twitter handles/Twitter
hashtags should be truncated to inside the text of a link. If the match is
over the number of characters, it will be truncated to this length by
replacing the end of the string with a two period ellipsis ('..').
In the object form, both length and location may be specified to perform
truncation. Available options for location are: 'end' (default), 'middle',
or 'smart'. Example usage:
truncate: { length: 32, location: 'middle' }
The 'smart' truncation option is for URLs where the algorithm attempts to
strip out unnecessary parts of the URL (such as the 'www.', then URL scheme,
hash, etc.) before trying to find a good point to insert the ellipsis if it is
still too long. For details, see source code of:
className : string
A CSS class name to add to the generated anchor tags. This class will be added
to all links, as well as this class plus "url"/"email"/"phone"/"hashtag"/"mention"
suffixes for styling url/email/phone/hashtag/mention links differently.
The name of the hashtag/mention service is also added as a CSS class for those
types of matches.
For example, if this config is provided as "my-link", then:
URL links will have the CSS classes: "my-link my-link-url"
Email links will have the CSS classes: "my-link my-link-email"
Phone links will have the CSS classes: "my-link my-link-phone"
Twitter mention links will have the CSS classes: "my-link my-link-mention my-link-twitter"
Instagram mention links will have the CSS classes: "my-link my-link-mention my-link-instagram"
Hashtag links will have the CSS classes: "my-link my-link-hashtag my-link-twitter"
decodePercentEncoding: boolean true to decode percent-encoded characters in URL matches, false to keep
the percent-encoded characters.
Example when true: https://en.wikipedia.org/wiki/San_Jos%C3%A9 will
be displayed as https://en.wikipedia.org/wiki/San_José.
Defaults to true.
replaceFn : Function
A function to use to programmatically make replacements of matches in the
input string, one at a time. See the section
Custom Replacement Function for
more details.
true to HTML-encode the start and end brackets of existing HTML tags found
in the input string. This will escape < and > characters to < and
>, respectively.
Setting this to true will prevent XSS (Cross-site Scripting) attacks,
but will remove the significance of existing HTML tags in the input string. If
you would like to maintain the significance of existing HTML tags while also
making the output HTML string safe, leave this option as false and use a
tool like https://github.com/cure53/DOMPurify (or others) on the input string
before running Autolinker.
Defaults to false.
For example, if you wanted to disable links from opening in new windows, you could do:
const linkedText = Autolinker.link("Check out google.com", { newWindow: false});// Produces: "Check out <a href="http://google.com">google.com</a>"
And if you wanted to truncate the length of URLs (while also not opening in a new window), you could do:
A custom replacement function (replaceFn)
may be provided to replace url/email/phone/mention/hashtag matches on an
individual basis, based on the return from this function.
Full example, for purposes of documenting the API:
const input = "..."; // string with URLs, Email Addresses, Mentions (Twitter, Instagram), and Hashtagsconst linkedText = Autolinker.link(input, { replaceFn : function(match) { console.log("href = ", match.getAnchorHref()); console.log("text = ", match.getAnchorText()); switch(match.type) { case 'url': console.log("url: ", match.getUrl()); return true; // let Autolinker perform its normal anchor tag replacement case 'email': const email = match.getEmail(); console.log("email: ", email); if(email === "my@own.address") { return false; // don't auto-link this particular email address; leave as-is } else { return; // no return value will have Autolinker perform its normal anchor tag replacement (same as returning `true`) } case 'phone': console.log("Phone Number: ", match.getPhoneNumber()); return '<a href="http://newplace.to.link.phone.numbers.to/">' + match.getPhoneNumber() + '</a>'; case 'mention': console.log("Mention: ", match.getMention()); console.log("Mention Service Name: ", match.getServiceName()); return '<a href="http://newplace.to.link.mention.handles.to/">' + match.getMention() + '</a>'; case 'hashtag': console.log("Hashtag: ", match.getHashtag()); return '<a href="http://newplace.to.link.hashtag.handles.to/">' + match.getHashtag() + '</a>'; } }} );
Modifying the default generated anchor tag
const input = "..."; // string with URLs, Email Addresses, Mentions (Twitter, Instagram), and Hashtagsconst linkedText = Autolinker.link( input, { replaceFn : function( match ) { console.log("href = ", match.getAnchorHref()); console.log("text = ", match.getAnchorText()); const tag = match.buildTag(); // returns an `Autolinker.HtmlTag` instance for an <a> tag tag.setAttr('rel', 'nofollow'); // adds a 'rel' attribute tag.addClass('external-link'); // adds a CSS class tag.setInnerHtml('Click here!'); // sets the inner html for the anchor tag return tag; }} );
Internet Explorer support has been removed since its official demise in June
2022.
The urls.wwwMatches config has been removed. A www. prefix is now treated
like any other subdomain of a top level domain (TLD) match (such as
'subdomain.google.com').
Match.getType() should be replaced with Match.type. This allows for
TypeScript type narrowing of Match objects returned by the parse()
method or inside the replaceFn.
The Matcher classes have been removed in favor of a single finite state
machine parser, greatly improving the performance of Autolinker (3x
performance improvement over the 3.x branch), but removing some of the
customizability of the old regular expressions. Will address this
customizability in a future release.
Autolinker.AnchorTagBuilder, Autolinker.HtmlTag, and Autolinker.match.*
references have been removed. These shouldn't be needed as public APIs, but
please raise a GitHub issue if these are for some reason needed.
The HtmlParser class has been removed in favor of an internal parseHtml()
function which replaces the old regexp-based implementation with a state
machine parser that is guaranteed to run in linear time. If you were using
the HtmlParser class directly, I recommend switching to htmlparser2, which implements the HTML semantics
better. The internal parseHtml() function that Autolinker now uses is
fairly geared towards Autolinker's purposes, and may not be useful in a
general HTML parsing sense.
The codebase has been converted to TypeScript, and uses ES6 exports. You can
now use the import statement to pull in the Autolinker class and related
entities such as Match:
// ES6/TypeScript/Webpackimport Autolinker, { Match } from 'autolinker';
The require() interface is still supported as well for Node.js:
You will no longer need the @types/autolinker package as this package now
exports its own types
You will no longer be able to override the regular expressions in the
Matcher classes by assigning to the prototype (for instance, something like
PhoneMatcher.prototype.regex = ...). This is due to how TypeScript creates
properties for class instances in the constructor rather than on prototypes.
The idea of providing your own regular expression for these classes is a
brittle notion anyway, as the Matcher classes rely on capturing groups in
the RegExp being in the right place, or even multiple capturing groups for
the same piece of information to support a different format. These capturing
groups and associated code are subject to change as the regular expression
needs to be updated, and will not involve a major version release of
Autolinker.
In the future you will be able to override the default Matcher classes
entirely to provide your own implementation, but please raise an issue (or
+1 an issue) if you think the library should support a currently-unsupported
format.
Upgrading from v0.x -> v1.x (Breaking Changes)
twitter option removed, replaced with mention (which accepts 'twitter',
'instagram' and 'soundcloud' values)
Matching mentions (previously the twitter option) now defaults to
being turned off. Previously, Twitter handle matching was on by
default.
replaceFn option now called with just one argument: the Match
object (previously was called with two arguments: autolinker and
match)
(Used inside the replaceFn) TwitterMatch replaced with
MentionMatch, and MentionMatch.getType() now returns 'mention'
instead of 'twitter'
(Used inside the replaceFn) TwitterMatch.getTwitterHandle() ->
MentionMatch.getMention()
Developing / Contributing
Pull requests definitely welcome. To setup the project, make sure you have
Node.js installed. Then open up a command prompt and type
the following:
npm install -g pnpm@latest # this project uses pnpm workspaces, and pnpm is a faster npm anyway :)cd Autolinker.js # where you cloned the projectpnpm install
To run the tests:
pnpm run test
Make sure to add tests to check your new functionality/bugfix
These benchmarks attempt to set up all libraries by configuring comparable features to Autolinker (e.g.: linking emails, hashtags, mentions, etc.) to try to get an apples-to-apples comparison.
While developing, recommend running the benchmarks a few times both before and after making any changes if developing.
Unfortunately, JSDuck is a very old project that is no longer maintained. As
such, it doesn't support TypeScript or anything from ES6 (the class keyword,
arrow functions, etc). However, I have yet to find a better documentation
generator that creates such a useful API site. (Suggestions for a new one are
welcome though - please raise an issue.)
Since ES6 is not supported, we must generate the documentation from the ES5
output. As such, a few precautions must be taken care of to make sure the
documentation comes out right:
@cfg documentation tags must exist above a class property that has a
default value, or else it won't end up in the ES5 output. For example:
// Will correctly end up in the ES5 output/** * @cfg {String} title */readonly title: string = '';// Will *not* end up in ES5 output, and thus, won't end up in the generated// documentation/** * @cfg {String} title */readonly title: string;
The @constructor tag must be replaced with @method constructor
To build the documentation, you will need Ruby
installed (note: Ruby comes pre-installed on MacOS), with the
JSDuck gem.