Awesome Reviewers

When adding telemetry, previews, or fallback paths, ensure you don’t (a) repeat expensive work, (b) fetch/scan unbounded amounts of data, or (c) hard-fail when size/time limits are exceeded.

Apply these rules: 1) Reuse captured results

• If you already called a function that performs a costly capture/subprocess/clipboard round-trip, reuse its output for both the user-facing emission and any “success/failure” telemetry.

Example pattern (fetch once, reuse):

// Fetch once
const terminalSnapshot = await getLatestTerminalOutput();

// Emit based on the single snapshot
await returnCurrentTerminalContentsFromSnapshot(terminalSnapshot);

// Telemetry based on the same snapshot
telemetryService.captureTerminalOutputFailure(
  terminalSnapshot.state === "ok" ? /*...*/ : /*...*/,
);

(If the existing API can’t take a snapshot, refactor so the snapshot is computed once and shared.)

2) Bound “wide” Git/remote operations

• Avoid listing all remote heads or downloading large histories when you only need one branch or a small subset.
• Prefer upstream/default branch targets or local remote-tracking refs over enumerating every head.

3) Degrade gracefully under limits

• If subprocess output could exceed buffer/timeout (e.g., large repo status), don’t drop everything.
• Return { state: "partial" } (preserving critical fields like HEAD/branch) when limits are hit, so downstream can still function.

4) Gate unnecessary work during transitional UI states

• During handoffs/optimistic transitions, ensure your combined “should render loader” condition correctly suppresses avoidable loader insertion/removal.

These practices reduce latency, avoid side-effect duplication, and prevent expensive tasks from causing { state: "unavailable" } or UI churn.

This site is open source. Improve this page.