Download websites that use Choices

About Choices

Not every entry in a technology dataset is something a site owner chose. Choices is a small JavaScript component that replaces native select elements with searchable, taggable dropdowns, published under the MIT license and around since 2016. Nobody sets out to build a website with a dropdown library. The library arrives inside something else: an admin theme, a booking form, a checkout template, a plugin that needed a multi select and pulled one in.

The consequence shows up in the version numbers. Across the installs we can fingerprint, we record 31 distinct releases in circulation, with 10.2.0 the one that recurs most often. A component people upgrade deliberately clusters on a couple of recent releases. A component that ships inside other people's bundles spreads out instead, because each install is frozen at whatever the surrounding theme or plugin depended on the last time it was built.

The spread is the useful part. If you maintain a front end component, the distribution of pinned versions across live sites tells you how long a breaking change actually takes to drain out of the ecosystem, which is a different question from the one your download statistics answer. Package registry downloads count builds, including continuous integration runs that install a package and throw it away. A crawl of live pages counts sites that real people load.

There is a second reading of the same fingerprint. A dropdown replacement is a small usability and accessibility decision, and sites carrying one have usually been through a round of form work rather than shipping raw browser controls. Presence is weak evidence of a maintained front end rather than proof of it, since a template can carry the dependency for years after anyone last looked at the form.

As of October 2026 we see Choices on 4,456 sites, filed under JavaScript libraries. The count reflects only installs where the library is served as an identifiable file rather than minified into one application bundle with everything else, so it should be read as a floor. Bundling hides a great deal of front end code, and the smaller and more embeddable the component, the more of it is hidden.

Two audiences get value from a list like that. Maintainers and security researchers want the population still running an old release when an advisory lands, and a domain list is what turns an advisory into an actual notification effort. The other audience is commercial: companies selling form, checkout or accessibility products look for sites that already cared enough about their forms to replace the browser default. Neither question can be answered one domain at a time.