State and Pipelines

State is where bugs live: once a function's result depends on things other than its inputs, the reader must reconstruct history to reason about it. A pure function — output fully determined by inputs, no side effects — can be verified from its own lines alone, which is exactly what a diff reviewer can actually check.

Flag:

• Manual bookkeeping loops doing filter/transform/accumulate:

  const names: string[] = [];
  for (let i = 0; i < users.length; i++) {
    if (users[i].isActive) names.push(users[i].email);
  }
  // ↓ states what, hides how — nothing to get wrong
  const names = users.filter(u => u.isActive).map(u => u.email);
  

  Chains of filter → map → sort → slice form a declarative pipeline; each stage is independently readable and there's no index or accumulator to fumble.
• Computation interleaved with effects: a function that calculates a total and writes the database and sends an email makes domain reasoning depend on infrastructure history. Separate a coherent pure calculation (computeInvoice(items): Invoice) when that creates a real domain operation, then orchestrate I/O through the project's native effect mechanism. Do not fragment straightforward workflows or introduce interfaces merely to unit-test each call; test the shell at the smallest boundary that faithfully exercises its behavior.
• Mutable flags steering distant control flow: a module-level connectionDisabled = true set in one function and consulted in another is action at a distance — no local reading of either function reveals the link. Return values or pass state explicitly so the data flow is visible in signatures.