Extract rule: template-require-presentational-children

[NullVoxPopuli]

Split from #2371.

[NullVoxPopuli-ai-agent]

Review: template-require-presentational-children (PR #2611)

Comparison with ember-template-lint source

General correctness:

1. Significant behavioral difference — Case 1 (presentational role on parent elements): The ESLint version adds a whole new "Case 1" that the original ETL rule does NOT have. The original ETL only checks ROLES_THAT_CANNOT_SUPPORT_SEMANTIC_CHILDREN (button, checkbox, img, etc.). It does NOT check for role="presentation" or role="none" on <ul>, <table>, etc. with their semantic children (li, tr, td, etc.). The PRESENTATIONAL_CHILDREN map and associated logic is entirely new behavior not present in the original rule. While this may be useful, it should be documented as an intentional enhancement, not a migration.

2. role="none" check difference: In the original ETL, getAllElementNodeDescendants checks if attributeTextValue(findAttribute(childNode, 'role')) !== 'presentation'. This means it only skips nodes with role="presentation" — NOT role="none". The ESLint version's hasPresentationalRole() checks for both "none" and "presentation", which is technically more correct per the ARIA spec but diverges from the original.

3. Recursion into presentational children: The original ETL's getAllElementNodeDescendants does NOT recurse into children of presentational nodes (it returns [...descendants] where descendants is only populated via hasChildren check, but the node itself is excluded and its children are still walked). Actually, looking more carefully, the original DOES recurse — it calls getAllElementNodeDescendants(childNode) regardless and just excludes the presentational node itself from the results. The ESLint version also always recurses. This matches.

4. Error message difference: Original: "<ul> has a role of presentation, it cannot have semantic descendants like <li>". ESLint version: "Element <ul> has role=\"presentation\" but contains semantic child <li>. Presentational elements should only contain presentational children." — different wording.

5. Only checks direct children for Case 1: In the new Case 1 (presentational parent), only direct children are checked, not deeper descendants. If <ul role="presentation"><div><li>...</li></div></ul> is used, the nested <li> wouldn't be caught. However, since Case 1 is new behavior not in the original, this is just a note for completeness.

Scope analysis (gjs/gts):

Not applicable. This rule checks HTML element tags and role attribute values only — no helper/component name matching is involved.

Tests: Good coverage including the new Case 1, ETL-matching tests, named blocks, SVG skipping, and additionalNonSemanticTags option.

Summary: The main concern is the undocumented addition of "Case 1" behavior that wasn't in the original ETL rule. Consider either documenting this as an intentional enhancement or splitting it into a separate concern.

🤖 Automated review comparing with ember-template-lint source