Derive CSS classes the way ActiveAdmin does by Fivell · Pull Request #53 · activeadmin-plugins/capybara_active_admin

Fivell · 2026-10-01T08:04:38Z

The bug

Three selectors each hand-rolled their own approximation of how ActiveAdmin turns a label into a CSS class:

attributes_row_selector   label.gsub(/[\s\/]/, '_').downcase
table_header_selector     text.tr(' ', '_').downcase
table_cell_selector       text.downcase.gsub(/[^a-z0-9\s]/, '').gsub(/\s+/, '_').squeeze('_')

ActiveAdmin does none of those. It calls parameterize(separator: '_'):

AttributesTable#row — attributes_table.rb:30 — classes << "row-#{title.to_s.parameterize(separator: "_")}"
TableFor::Column#html_class — table_for.rb:152 — html_classes << "col-#{@title.to_s.parameterize(separator: "_")}"

So all three drift from the markup that was actually rendered, each in its own direction:

label	ActiveAdmin renders	`attributes_row_selector`	`table_cell_selector`
`'VAT / TAX Number'`	`vat_tax_number`	`vat___tax_number` ✗	`vat_tax_number`
`'E-mail'`	`e-mail`	`e-mail`	`email` ✗
`"Customer's Name"`	`customer_s_name`	`customer's_name` ✗	`customers_name` ✗
`'# of DIDs'`	`of_dids`	`#_of_dids` ✗	`_of_dids` ✗
`'Setup (one-time)'`	`setup_one-time`	`setup_(one-time)` ✗	`setup_onetime` ✗
`'Price, USD'`	`price_usd`	`price,_usd` ✗	`price_usd`
`'DID Number'`	`did_number`	`did__number` ✗	`did_number`
`'ISO 3166-1'`	`iso_3166-1`	`iso_3166-1`	`iso_31661` ✗

The failure is silent in the worst way: have_table_cell(column: 'E-mail') just doesn't match, and the spec reads as "the cell is missing".

This has already cost downstream users a workaround. #43 is the same bug seen from the table_cell_selector side, and at least one app carries a helper that pre-normalizes labels before handing them to the gem — which then breaks again when the gem's own normalization changes, because the current table_cell_selector strips _ along with everything else.

The fix

One helper that makes the same call ActiveAdmin makes:

def css_class_for_label(label)
  label.to_s.parameterize(separator: '_')
end

All three selectors route through it, so a label can only ever produce the class that was rendered. parameterize is idempotent on its own output, so apps that pre-normalize as a workaround keep working and can drop it at their leisure.

Verification

Unit examples for the labels in the table above, in both spec/lib/selectors/table_spec.rb and spec/lib/selectors/attributes_table_spec.rb.

End to end, the dummy's show page now has a deliberately awkward label:

row('VAT / TAX Number') { |r| r.salary }

and spec/system/business_employee_show_spec.rb asserts the gem finds it. With the previous derivation restored, that spec fails against the real rendered page:

expected to find css "tr.row.row-vat___tax_number > td" but there were no matches

Full suite: 27 examples, 0 failures.

Compatibility

Only labels that the old code mapped to a class ActiveAdmin never rendered change behaviour — those lookups were already finding nothing. Plain labels ('Full name', 'ID', 'Country / Region', 'Price (USD)') are unaffected; the existing examples for them pass untouched.

Three selectors each hand-rolled their own approximation of how ActiveAdmin turns a label into a CSS class: attributes_row_selector label.gsub(/[\s\/]/, '_').downcase table_header_selector text.tr(' ', '_').downcase table_cell_selector text.downcase.gsub(/[^a-z0-9\s]/, '') .gsub(/\s+/, '_').squeeze('_') ActiveAdmin does neither. It calls `parameterize(separator: '_')` -- `AttributesTable#row` (attributes_table.rb:30) and `TableFor::Column#html_class` (table_for.rb:152) -- so the three drift from the rendered markup, each differently: label ActiveAdmin row selector cell selector 'VAT / TAX Number' vat_tax_number vat___tax_number vat_tax_number 'E-mail' e-mail e-mail email "Customer's Name" customer_s_name customer's_name customers_name '# of DIDs' of_dids #_of_dids _of_dids 'Price, USD' price_usd price,_usd price_usd 'DID Number' did_number did__number did_number Route all three through `Util.css_class_for_label`, which makes the same call ActiveAdmin makes, so a label can only ever produce the class that was actually rendered. Covered by unit examples for the labels above and, end to end, by a `row('VAT / TAX Number')` added to the dummy's show page: with the old derivation that spec looks for `tr.row.row-vat___tax_number` and finds nothing.

Copilot

Copilot review overview

🟡 Changes recommended

The changed table-header normalization path lacks direct or system-level test coverage.

Review effort: Balanced
Findings: 1

Open (1)

Add tests for table header normalization and column override · New

What changed in this PR

Aligns label-derived selectors with ActiveAdmin’s parameterize(separator: '_') behavior.

Changes:

Adds shared label-to-CSS-class normalization.
Updates table and attribute-row selectors.
Adds unit and system regression coverage.

File	Description
`lib/capybara/active_admin/util.rb`	Adds shared normalization helper.
`lib/capybara/active_admin/selectors/table.rb`	Normalizes table header and cell classes.
`lib/capybara/active_admin/selectors/attributes_table.rb`	Normalizes attribute-row classes.
`spec/lib/selectors/table_spec.rb`	Covers punctuation in cell labels.
`spec/lib/selectors/attributes_table_spec.rb`	Covers separator normalization.
`spec/dummy/test_application.rb`	Adds an awkward rendered label.
`spec/system/business_employee_show_spec.rb`	Verifies matching against rendered markup.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

The third selector this PR rewrote had no examples at all — not one table_header_selector or have_table_header anywhere in the suite — so its normalization could have regressed unnoticed while the cell and attribute-row cases stayed green. Seven examples: generic form, a plain label, the three labels where the previous `tr(' ', '_').downcase` drifted from parameterize, the `column:` override taking the same path as the visible text, and the sortable/sorted suffixes still being appended. Restoring the old implementation fails four of them: expected: col-vat_tax_number got: col-vat_/_tax_number expected: col-customer_s_name got: col-customer's_name expected: col-of_dids got: col-#_of_dids

Copilot

Copilot review overview

🔵 Needs a closer look

The changed table-header normalization path lacks regression coverage.

Review effort: Balanced
Findings: 1

Open (1)

Add tests for table header normalization and column override

…#55) Follow-up to #53, from a review of it. `parameterize` transliterates to Latin and drops the rest, so 'Имя', '名前', '№' and a blank string all come back as ''. ActiveAdmin makes the same call, so it renders `class="row row-"` for every one of them — and #53 made the gem build `tr.row.row-` to match, which matches them all at once: attributes_row_selector('Имя') => tr.row.row- > td attributes_row_selector('Телефон') => tr.row.row- > td On a page with both rows, `have_attribute_row('Имя', exact_text: ...)` passes against the phone row, and `within_attribute_row('Имя')` raises Capybara::Ambiguous. The derivation #53 replaced produced `row-имя`, matched nothing, and failed honestly. Turning a loud failure into a silent wrong match is the worse half of the trade, so the helper now raises with the label named and a suggestion to match on text. Also here, all from the same review: * `selectors/table.rb` and `selectors/attributes_table.rb` reference `Util` without requiring it. They load today only because `capybara/active_admin.rb` happens to require util one line earlier; requiring the selectors file directly raised NameError. * `table_header_selector(nil, column: 'x')` returned the generic header selector, discarding the override — so `find_table_header` matched the first header on the page whatever column was asked for. * `table_cell_selector(:full_name)` had no example, though that is the commonest input of all and the old derivation broke it (`col-fullname`). Added, plus examples for the new guard. * Two example names still said "strips slashes" / "strips other special characters", the rule #53 deleted — they passed only because those two labels collapse the same way under both. A maintainer reading them as the contract would reinstate the stripping. * CHANGELOG had nothing for #53 at all, in a released gem whose README commits to Keep a Changelog, including the selector changes a consumer needs to know about on `bundle update`.

Follow-up to #53, from the same review. #53 fixed how a *label* becomes a CSS class; these are the two remaining derivations of the same kind, where a *model* becomes one. Arbre builds a record's DOM name from `record.class.model_name.singular` (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 of which underscore a namespace. The gem used `gsub(' ', '_').downcase` plus an unconditional singularize/pluralize, which leaves the `::` alone: 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 naming the cause. `table_selector` also had no class branch at all, so passing a model class was broken for every namespaced model. Verified against real rendered pages: the dummy renders `<div class="attributes_table billing_employee">` and `<table id="index_table_business_employees">`, and the new system spec asserts both. Restoring the old derivation fails it. Note the two names have different sources: the table id follows the name a resource was *registered* under (`as: 'Business Employee'`), so a renamed resource cannot be addressed by its class. Documented at the call site rather than papered over. Also fixes six dead `have_table_cell` examples in README.md, which still showed a positional-text signature the matcher has never had — they raise TypeError or ArgumentError as written.

Fivell requested review from senid231 and a balanced review from Copilot October 1, 2026 08:56

Copilot started reviewing on behalf of Fivell October 1, 2026 08:57 View session

Copilot AI reviewed Oct 1, 2026

View reviewed changes

Comment thread lib/capybara/active_admin/selectors/table.rb

Fivell requested a balanced review from Copilot October 1, 2026 12:13

Copilot started reviewing on behalf of Fivell October 1, 2026 12:14 View session

Copilot AI reviewed Oct 1, 2026

View reviewed changes

senid231 approved these changes Oct 1, 2026

View reviewed changes

Fivell merged commit 4b9fecc into master Oct 1, 2026
37 checks passed

Fivell deleted the fix/match-activeadmin-css-class-derivation branch October 1, 2026 12:36

Fivell mentioned this pull request Oct 1, 2026

Refuse to build a selector from a label that parameterizes to nothing #55

Merged

Fivell mentioned this pull request Oct 1, 2026

Ask the model for its DOM name instead of approximating it #56

Merged

Navigation Menu

Repository navigation

Derive CSS classes the way ActiveAdmin does - #53

Derive CSS classes the way ActiveAdmin does#53
Fivell merged 2 commits into
masterfrom
fix/match-activeadmin-css-class-derivation

Fivell commented Oct 1, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Fivell commented Oct 1, 2026

The bug

The fix

Verification

Compatibility

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🔵 Needs a closer look

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants