Today, I was adding a React Select element to a page, and I really wanted to test it.

This is, surprisingly, not the first time I’ve attempted this.

Since 2016, React Select has undergone some changes, released a brand-new (and much better!) version. One of these big features is that it uses Emotion for styling. A consequence of this is that React Select no longer has .Select, or .Select-input classes that you could use as selectors in the tests. Instead, it will generate hashed CSS selectors, like css-2b097c-container and css-1rhbuit-multiValue. These have the potential to change: if the underlying CSS changes, then the hash will change. So they are not reliable in tests.

To fix this, I used another prop of ReactSelect called classNamePrefix:

import CreatableSelect from "react-select/creatable";

// ...

const options = [
  { label: "Admin", value: "Admin" },
  { label: "HIU", value: "HIU" },
  { label: "Organisational", value: "Organisational" },
  { label: "Paid", value: "Paid" },
];

// "tags" here comes from a prop on this component
const selectedOptions = options.filter(({ value }) => tags.includes(value));

return (
  <CreatableSelect
    options={options}
    defaultValue={selectedOptions}
    isMulti
    classNamePrefix="tags"
  />
);

This prop adds additional classes to the React Select component, such as .tags__control and .tags__multi-value – values that are reliable and predictable!

This means that in the Capybara test, I can now use these to interact with this React Select component, like this:

# spec/support/filter_helpers.rb
module FilterHelpers
  class SelectControl
    include Capybara::DSL

    attr_reader :prefix

    def initialize(prefix:)
      @prefix = prefix
    end

    def select(option)
      within_control do
        find("input").fill_in(with: option)
      end

      find(".#{prefix}__option", text: option).click
    end

    def value
      find(single_value_selector).text
    end

    def blank?
      page.has_no_selector?(single_value_selector)
    end

    def values
      all(multi_value_selector).map(&:text)
    end

    def remove(label)
      value = find(multi_value_selector, text: label)
      within(value) do
        find("#{multi_value_selector}__remove").click
      end
    end

    def visible?
      page.has_selector?(control_selector)
    end

    def hidden?
      !visible?
    end

    private

    def single_value_selector
      ".#{prefix}__single-value"
    end

    def multi_value_selector
      ".#{prefix}__multi-value"
    end

    def control_selector
      ".#{prefix}__control"
    end

    def within_control(&block)
      within(control_selector, &block)
    end
  end

  module TagFilterControl
    def tag_filter_control
      SelectControl.new(prefix: "tags")
    end

    def add_new_tag(tag)
      tags_control.select(tags)
    end

    def expect_tags_selected(tags)
      expect(tags_control.value).to eq(tags)
    end

    def expect_no_tags_selected
      expect(tags_control).to be_blank
    end

    def remove_tags(*tags)
      tags.each do |tag|
        tags_control.remove(tag)
      end
    end
  end
end

# spec/features/updating_tags_spec.rb

RSpec.describe "Updating tags" do
  include TagFilterControl

  scenario "Can update a User's tags", js: true do
    visit edit_admin_user_path(user)

    expect_tags_selected("Existing tag")
    remove_tag("HIU")
    add_new_tag("Paid")
    add_new_tag("Custom tag")
  end
end

I have the following helpers:

  1. expect_selected_tag: When the page loads, the Select element should contain a tag that is already assigned to the user.
  2. remove_tag: Used to remove a tag that appears in the Select element.
  3. add_new_tag: Used to add either a tag from the list of options, or to create a brand new tag.

With these helpers, I can effectively test React Select within Capybara and Selenium.