Summary
:for="item in store.signalRef" (bare reference) works correctly. :for="item in store.signalRef()" (call syntax) silently iterates zero items, even when the signal IS populated. No console error, no warning — the dropdown just stays empty.
This is inconsistent with every other binding site (:text, :if, x-class, x-style, …) where both bare and call syntax work interchangeably. :for is the only attribute where call syntax silently fails.
Reproducer
```html
<button :for="judge in judgesStore.searchResults" :text="judge.name"></button>
<button :for="judge in judgesStore.searchResults()" :text="judge.name"></button> ```
Tested with judgesStore.searchResults defined as `const searchResults = state<judge>[]>([])` exported from a `defineStore('judges', ...)`.
Root cause
`signals.ts:1980-2227` (`bindFor`). The list expression is evaluated via `evalLazy` (line 2092) which uses an auto-unwrap proxy as the `with()` scope:
```ts const fn = new Function('scope', 'with(scope) { return ' + expression + ' }') return fn(unwrapScope) ```
The proxy's get-trap (line 281-292) unwraps signals on access: if a property is a function with `_isSignal` / `_isDerived`, it calls it and returns the value. For stx stores (`_isStxStore`), it recursively wraps.
So:
- `judgesStore.searchResults` (bare): proxy.get('judgesStore') → sub-proxy → proxy.get('searchResults') → detects signal, calls it, returns the array. ✓
- `judgesStore.searchResults()` (call): proxy auto-unwraps as above, returns the array. Then `()` is applied to the array → `TypeError: array is not a function` → caught by evalLazy's try/catch → returns `''`. Then `bindFor` at line 2227: `if (!Array.isArray(list)) return` → silently bails. ✗
The TypeError is swallowed at `signals.ts:2101-2104` because the catch filter skips `ReferenceError` and `TypeError` for the warn path.
Impact
This is a real footgun. The exact same expression that works fine everywhere else (`{{ store.signal() }}`, `:text="store.signal()"`, `x-class="store.signal().length > 0 ? '' : 'hidden'"`) silently breaks the iteration in `:for`. Caught while migrating a header search dropdown — the same data was reactive via `x-class` (so the section headers showed) but the `:for` button list rendered zero items. No warning. Section headers + empty list looked like the component was "loaded but with no data."
Even the framework's own components have leave-behind comments documenting this gotcha (`resources/components/Bench/Judge/ReviewJudgeSearch.stx`):
"Bind `:for` to the store signal directly (not wrapped in a local derived) so the list re-renders when results land."
…which is the same issue rediscovered manually.
Suggested fix
Pick one:
Accept both forms (preferred). In bindFor's effect, after `evalLazy(listExpr)`, if the result is not an array but IS a function with signal markers, call it. Symmetric with how other bindings already tolerate both.
Strip trailing `()` from the listExpr at parse time if the bare reference also resolves to a signal. Cheap and avoids any runtime cost.
At minimum, log a clear warning. Currently the silent return at `signals.ts:2227` masks the error. `if (!Array.isArray(list)) { console.warn('[STX] :for expected an array, got', typeof list, '— did you call a signal with
()inside :for? Try the bare reference instead.'); return; }` would have saved several iterations of debugging here.
Workaround in the meantime
Drop the parens: `:for="item in signal"` instead of `:for="item in signal()"`.
Environment
- stx vendored via stacks pantry (symlinked to a local checkout)
- Bun 1.3.13
- Caught while implementing a global header search typeahead on a Stacks app — `<button :for="">"judge in judgesStore.searchResults()">` rendered zero rows despite the same store signal driving `x-class="...length > 0 ? '' : 'hidden'"` on the wrapper just fine.