Pressing Enter to choose an input-method candidate should not accidentally spend a timed quiz answer. That is an input-handling case worth separating from a countdown test.
I build Streakdle, an independent rare-answer browser game with seven prompts per round and 25 seconds per prompt. The example gives this checklist a concrete setting; this is an original, AI-assisted test note, not a claim that Streakdle has passed these checks or uses the sample code below.
Separate text composition from answer submission
MDN describes KeyboardEvent.isComposing as a flag for a keyboard event inside a composition session. Its compatibility notice is a reason to test the browsers you support. The compositionend event can mean composition finished or was cancelled; it should not itself be treated as an instruction to submit a quiz answer.
This small predicate expresses only one proposed rule:
export function shouldConsiderEnter(event) {
return event.key === "Enter" && !event.isComposing;
}
The name deliberately says “consider”. Passing it is not permission to score an answer. A separate layer still checks whether the prompt is open, an answer has already been accepted, and the text is usable under the game's published rules. It also does not establish a server-side deadline.
Start with a tiny deterministic check
These are fictional event objects. They test the predicate, not the browser's composition behavior:
const cases = [
[{ key: "Enter", isComposing: true }, false],
[{ key: "Enter", isComposing: false }, true],
[{ key: "a", isComposing: false }, false],
[{ key: "Escape", isComposing: true }, false],
];
for (const [event, expected] of cases) {
if (shouldConsiderEnter(event) !== expected) {
throw new Error("Enter predicate mismatch");
}
}
Keep that check small. A passing synthetic test does not prove that a particular operating system and keyboard emit the event sequence you expected.
Then run the human input matrix
| Interaction | Observe before deciding pass or fail |
|---|---|
| Type ordinary Latin text, then press Enter | Does one deliberate action reach the normal answer path? |
| Start a Chinese or Japanese input-method composition and select a candidate with Enter | Does candidate selection leave the prompt available for the intended answer? |
| Cancel composition | Does cancellation leave the current text and submission state understandable? |
| Choose a candidate, pause, then deliberately submit | Can the participant distinguish text confirmation from answer confirmation? |
| Let the timer finish while a candidate panel is open | Does the UI clearly explain whether an answer was accepted, without accepting an unintended fragment? |
| Submit using the visible button instead of Enter | Do keyboard and pointer paths agree about prompt state and duplicate acceptance? |
Write down the operating system, browser version, keyboard/input method and the exact action sequence. Use fictional text or participant-approved examples; a useful incident report does not need a full collection of real answers.
Preserve the failure category
Record three separate observations: what text was visible, whether the UI acknowledged an answer, and which prompt state followed. “The candidate changed” and “the answer was accepted” describe different events. A network retry is also different from a second accepted answer.
If a failure appears only on one input method, retain that scope in the report. If you have not checked a browser, label it untested. This checklist is meant to produce a reproducible input report, not a broad accessibility or compatibility certification.
The linked game is my own product. This is maker-authored educational material, with no independent endorsement implied.
Top comments (0)