Extract rule: template-table-groups

[NullVoxPopuli]

Split from #2371.

[NullVoxPopuli-ai-agent]

Review: template-table-groups

Compared against ember-template-lint table-groups.js.

General Correctness

1. Faithful port: This is a complex rule and the port does a thorough job. The core logic -- checking that <table> elements only contain allowed table grouping children (caption, colgroup, thead, tbody, tfoot) in the correct order, with control flow transparency -- is well implemented.

2. Empty table handling: The port adds special handling for truly empty tables (lines 279-288) by parsing the source text to detect self-closing-like patterns. The original ETL rule does NOT have this explicit check -- it would just iterate over an empty children array and not report. This means the ESLint port is stricter than the original for empty <table></table>. In the original, <table></table> would not trigger an error because the children array is empty and the loop simply doesn't execute. Verify whether this added strictness is intentional.

3. dasherize function: The port implements its own dasherize() using regex. The original imports dasherizeComponentName from a helper. The port's implementation converts :: to / and camelCase to kebab-case, which matches the expected behavior for Ember component name conversion. Looks correct.

4. isControlFlowHelper list: The port includes ['if', 'unless', 'each', 'each-in', 'let', 'with'], which matches the original's AstNodeInfo.isControlFlowHelper(). Good.

5. Control flow scoping for if/unless branches: The original uses scopedIndices as a stack and, on CONTROL_FLOW_START_MARK, creates a new set by reducing all scoped indices. The ESLint port instead creates a copy of just the current set: new Set(currentAllowedMinimumIndices). This is a behavioral difference from the original. The original merges ALL parent scopes into the new scope, while the port only copies the current scope. For nested control flow with if/else branches, this could produce different results. Consider aligning with the original's reduce approach.

6. Tests: Excellent coverage -- 893 lines of tests covering valid/invalid cases, control flow, ordering, custom component mappings via options, and both gjs and hbs modes. The test suite is comprehensive and well-structured.

7. Schema alignment: The schema correctly uses the same allowed-*-components keys as the original's configuration. Good.

Scope Analysis (gjs/gts)

This rule checks structural properties: HTML tag names (table, thead, etc.), attribute names (@tagName), and control flow helper names (if, unless, each, etc.). It also checks node.path?.original === 'yield' for the {{yield}} keyword. All of these are Ember built-in keywords or HTML structural elements. No scope analysis needed -- shadowing yield, if, unless, etc. in JS would be extremely unusual and not a realistic concern.

---

🤖 Automated review comparing with ember-template-lint source