x-govuk / html-attributes-utils Goto Github PK

We'd like to use this library for managing attributes in ViewComponent models that use Turbo and Stimulus. We'd be happy to contribute PRs if this is in line with your project goals, but want to get your input first.

We have three blockers currently:

1. Support for stimulus controller and action attributes that require merges:

Input:

def default_attributes
  { data: { controller: "default", action: "click->extra#event" }}
end
...
Component.new(data: { controller: "extra", action: "click->extra#event" })

Desired output:

<div data-controller="default extra" data-action="click->default#event click->extra#event">

2. Automatic flattening of merged arrays

... Rails accepts them as arrays so there's no need to convert back to strings.

Adding custom configuration for the above attributes reveals that the documentation is not correct in general.

3. deep_tidy_html_attributes removes empty strings

This is in line with the documented behaviour, but it's surprising given that the usual behaviour of Rails tag rendering is that an attribute with an empty string becomes a bare attribute, e.g.

Input:

tag.div(hidden: "", data: { turbo_stream: "" })

Output:

<div hidden data-turbo-stream>

Rails will remove attributes with nil values, but empty string attributes are usually rendered.

Could you perhaps explain the reasoning for removing empty attributes via deep_tidy_html_attributes? It's not clear from the documentation why this feature is helpful.

If you'd be happy to accept PRs for any or all of these suggested enhancements please let us know.

In Rails there are two ways of setting data and aria attributes when using the tag and content_tag helpers. These are to set the attribute by its full name data-controller: "abc", data-response: "xyz" or to use a nested hash: data: { controller: "abc", response: "xyz" }.This library, and both the components and form builder assume people are using the latter. Using the former might lead to unintended consequences.

It might make sense to detect the full name variant and either apply a conversion or provide some kind of warning to the developer. As this only applies to data- and aria- prefixed attributes the conversion should be straightforward.