You Dont Know JS Async Performance (Kyle Simpson)（Z-Library）

【One-Line Pitch】 A deep, opinionated tour of how JavaScript actually handles time, concurrency, and speed—written for developers who can already ship code but want to stop guessing why async behavior surprises them. If you've ever blamed the language for a callback tangle or a sluggish page, this book is aimed squarely at you. 【Book Arc】 - **Opening (~0%–10%)**: Frames the series' mission—reject "just enough" JavaScript and study the tough parts deliberately—then sets up asynchrony as the central problem of "now vs. later." - **Early (~10%–35%)**: Establishes the mental model: programs run in chunks, the event loop schedules them, and async is not the same as parallel. Introduces callbacks as the default (and flawed) answer. - **Middle (~35%–60%)**: Deepens the event-loop and concurrency discussion, including why `setTimeout` timing is approximate, how single-threaded JS avoids shared-memory races, and where I/O observability gets tricky. - **Late (~60%–85%)**: Moves through the modern toolkit—Promises, then Generators, then Generators combined with Promises—as progressively better ways to express async flow and error handling. - **Ending (~85%–100%)**: Shifts to performance: Web Workers, parallel JS, SIMD, asm.js, then benchmarking methodology and tuning, plus appendices on the `asynquence` library and advanced async patterns. 【Key Takeaways】 - **Async is about the gap between now and later** (Early): Nontrivial programs must manage state across time—user input, network, timers—so understanding that gap is the foundation, not an edge case. - **The event loop, not the JS engine, owns scheduling** (Middle): The engine executes one chunk at a time on demand; the hosting environment queues callbacks. This explains why timers are approximate and why callbacks wait in line. - **Async ≠ parallel** (Middle): Parallelism means simultaneous execution via threads/processes; JS's event loop serializes tasks. Conflating the two leads to wrong assumptions about shared state and ordering. - **Callbacks are sufficient until they aren't** (Early–Middle): They work, but as programs grow, inversion of control and "trust issues" make them harder to reason about—motivating Promises. - **Promises restore trust and composability** (Late): They provide chainable flow, centralized error handling, and predictable scheduling, though the book also examines their limitations rather than treating them as a cure-all. - **Generators + Promises change readability** (Late): Generators alone are more complex than interesting; combined with Promises they become a practical tool for writing async code that reads sequentially. - **Performance work requires measurement discipline** (Ending): Benchmarking context matters, micro-optimizations are often misleading, and tools like jsPerf demand well-designed tests before conclusions. - **Parallelism in JS has real tools but real trade-offs** (Ending): Web Workers, SIMD, and asm.js expand what's possible, but they don't erase the single-threaded model's constraints. 【Reading Tips】 - Read Chapters 1–2 carefully even if you think you know callbacks; the event-loop and "now vs. later" framing underpins everything later. - Treat Promises and Generators as a paired unit—skim neither. The payoff is in how they combine, not in either alone. - Skim the performance chapters if you're not currently tuning code, but return to the benchmarking methodology before trusting any micro-benchmark you write. - Use the appendices as reference, not required reading; `asynquence` and advanced patterns are optional depth. - Keep a scratchpad for the pseudocode examples—tracing the event loop and parallel-thread scenarios by hand is where the concepts click. 【Coverage Limits】 This guide is synthesized from stratified excerpts covering the table of contents, preface, and portions of the early-to-middle chapters; the excerpts do not cover the full text of the Promises, Generators, performance, or appendix chapters in detail, so chapter-level specifics beyond the listed topics are not represented here.