Ask the model for its DOM name instead of approximating it

[Fivell]

Follow-up to #53, from the same review. #53 fixed how a label becomes a CSS class; this is the other half — how a model becomes one.

The bug

Arbre derives a record's DOM name from record.class.model_name.singular, falling back to name.underscore.gsub('/', '_') (arbre html/tag.rb:164-168), and ActiveAdmin builds the index table id from active_admin_config.resource_name.plural (index_as_table.rb:240). Both underscore a namespace.

The gem used gsub(' ', '_').downcase plus an unconditional singularize/pluralize, which leaves :: in place:

have_attributes_table(model: 'Billing::Employee')
# => div.attributes_table.billing::employee

have_table(resource_name: Billing::Employee)
# => table#index_table_billing::employees

Neither is a selector. Under cuprite the first aborts the example with Ferrum::JavaScriptError, under rack_test with Nokogiri::CSS::SyntaxError — in both cases with nothing pointing at the cause, so it reads as a driver problem rather than a bad selector.

table_selector also had no class branch at all: it called .to_s on whatever it was given, so passing a model class was broken for every namespaced model.

Secondary: the unconditional singularize mangles irregulars even on the happy path — parse_model_name('Statistics') returned statistic.

What it renders, measured

Printed from the dummy app with a real browser rather than reasoned about:

<table id="index_table_business_employees">
<div class="attributes_table billing_employee">

Note the two come from different sources: the table id follows the name the resource was registered under (register Billing::Employee, as: 'Business Employee'), while the attributes-table class follows the model class. A renamed resource therefore cannot be addressed by its class — that is ActiveAdmin's design, not something this PR can fix, so it is documented at the call site instead of papered over.

The fix

parse_model_name asks the model when it can, and otherwise mirrors Arbre's fallback:

if model_name.is_a?(Class) && model_name.respond_to?(:model_name)
  name = model_name.model_name
  return singular ? name.singular : name.plural
end

dom_model_name(model_name.is_a?(Class) ? model_name.name : model_name, singular: singular)

with dom_model_name doing underscore.tr('/', '_').tr(' ', '_'). table_selector now routes through it instead of carrying its own copy.

Verification

'Business Employee'  -> table#index_table_business_employees   (unchanged)
'Billing::Employee'  -> table#index_table_billing_employees    (was billing::employees)
Billing::Employee    -> table#index_table_billing_employees    (was billing::employees)
model: Billing::Employee  -> div.attributes_table.billing_employee   (unchanged)
model: 'Billing::Employee' -> div.attributes_table.billing_employee  (was billing::employee)

Six unit examples, plus spec/system/namespaced_model_names_spec.rb asserting both names against a real rendered page — restoring the old derivation fails it:

Failure/Error: expect(page).to have_attributes_table(model: 'Billing::Employee')
Ferrum::JavaScriptError

Suite: 42 examples, 0 failures.

Also

Six dead have_table_cell examples in README.md (lines 58-68). The matcher has always been have_table_cell(options = {}) reading :column, but the README shows have_table_cell('John Doe', row_id: …, col_name: …), which raises TypeError: no implicit conversion of Symbol into String or ArgumentError: wrong number of arguments. docs/guide/README.md already uses the correct form; only the top-level README was stale.

Depends on

Nothing, but it overlaps #55 in spirit — that one adds the missing require 'capybara/active_admin/util' to these same two selector files. Different hunks, so they do not conflict; either order merges.

Left out deliberately

The review also flagged table_scope_selector as a bare li.scope while ActiveAdmin::Scope renders its id as a class. I measured it on a real page:

<li class="scope on_off_payroll selected">   text: "On / Off payroll (1)"

have_table_scope(title) keys on exact_text: title, which cannot match while scopes_show_count is on by default. But the class is not reliably derivable from the title either — Scope#initialize takes the id from the method when one is given (scope 'Pending Review', :awaiting → awaiting), and only otherwise from the name. Fixing it properly means deciding on an API, so it wants a maintainer's call rather than a guess from me.