m04-zero-cost

m04-zero-cost

热门

CRITICAL: Use for generics, traits, zero-cost abstraction. Triggers: E0277, E0308, E0599, generic, trait, impl, dyn, where, monomorphization, static dispatch, dynamic dispatch, impl Trait, trait bound not satisfied, 泛型, 特征, 零成本抽象, 单态化

534Star
53Fork
更新于 1/24/2026
SKILL.md
readonly只读
name
m04-zero-cost
description

"CRITICAL: Use for generics, traits, zero-cost abstraction. Triggers: E0277, E0308, E0599, generic, trait, impl, dyn, where, monomorphization, static dispatch, dynamic dispatch, impl Trait, trait bound not satisfied, 泛型, 特征, 零成本抽象, 单态化"

Zero-Cost Abstraction

Layer 1: Language Mechanics

Core Question

Do we need compile-time or runtime polymorphism?

Before choosing between generics and trait objects:

  • Is the type known at compile time?
  • Is a heterogeneous collection needed?
  • What's the performance priority?

Error → Design Question

Error Don't Just Say Ask Instead
E0277 "Add trait bound" Is this abstraction at the right level?
E0308 "Fix the type" Should types be unified or distinct?
E0599 "Import the trait" Is the trait the right abstraction?
E0038 "Make object-safe" Do we really need dynamic dispatch?

Thinking Prompt

Before adding trait bounds:

  1. What abstraction is needed?

    • Same behavior, different types → trait
    • Different behavior, same type → enum
    • No abstraction needed → concrete type
  2. When is type known?

    • Compile time → generics (static dispatch)
    • Runtime → trait objects (dynamic dispatch)
  3. What's the trade-off priority?

    • Performance → generics
    • Compile time → trait objects
    • Flexibility → depends

Trace Up ↑

When type system fights back:

E0277 (trait bound not satisfied)
    ↑ Ask: Is the abstraction level correct?
    ↑ Check: m09-domain (what behavior is being abstracted?)
    ↑ Check: m05-type-driven (should use newtype?)
Persistent Error Trace To Question
Complex trait bounds m09-domain Is the abstraction right?
Object safety issues m05-type-driven Can typestate help?
Type explosion m10-performance Accept dyn overhead?

Trace Down ↓

From design to implementation:

"Need to abstract over types with same behavior"
    ↓ Types known at compile time → impl Trait or generics
    ↓ Types determined at runtime → dyn Trait

"Need collection of different types"
    ↓ Closed set → enum
    ↓ Open set → Vec<Box<dyn Trait>>

"Need to return different types"
    ↓ Same type → impl Trait
    ↓ Different types → Box<dyn Trait>

Quick Reference

Pattern Dispatch Code Size Runtime Cost
fn foo<T: Trait>() Static +bloat Zero
fn foo(x: &dyn Trait) Dynamic Minimal vtable lookup
impl Trait return Static +bloat Zero
Box<dyn Trait> Dynamic Minimal Allocation + vtable

Syntax Comparison

// Static dispatch - type known at compile time
fn process(x: impl Display) { }      // argument position
fn process<T: Display>(x: T) { }     // explicit generic
fn get() -> impl Display { }         // return position

// Dynamic dispatch - type determined at runtime
fn process(x: &dyn Display) { }      // reference
fn process(x: Box<dyn Display>) { }  // owned

Error Code Reference

Error Cause Quick Fix
E0277 Type doesn't impl trait Add impl or change bound
E0308 Type mismatch Check generic params
E0599 No method found Import trait with use
E0038 Trait not object-safe Use generics or redesign

Decision Guide

Scenario Choose Why
Performance critical Generics Zero runtime cost
Heterogeneous collection dyn Trait Different types at runtime
Plugin architecture dyn Trait Unknown types at compile
Reduce compile time dyn Trait Less monomorphization
Small, known type set enum No indirection

Object Safety

A trait is object-safe if it:

  • Doesn't have Self: Sized bound
  • Doesn't return Self
  • Doesn't have generic methods
  • Uses where Self: Sized for non-object-safe methods

Anti-Patterns

Anti-Pattern Why Bad Better
Over-generic everything Compile time, complexity Concrete types when possible
dyn for known types Unnecessary indirection Generics
Complex trait hierarchies Hard to understand Simpler design
Ignore object safety Limits flexibility Plan for dyn if needed

Related Skills

When See
Type-driven design m05-type-driven
Domain abstraction m09-domain
Performance concerns m10-performance
Send/Sync bounds m07-concurrency

You Might Also Like

Related Skills

coding-agent

coding-agent

179Kdev-codegen

Run Codex CLI, Claude Code, OpenCode, or Pi Coding Agent via background process for programmatic control.

openclaw avataropenclaw
获取
add-uint-support

add-uint-support

97Kdev-codegen

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint support.

pytorch avatarpytorch
获取
at-dispatch-v2

at-dispatch-v2

97Kdev-codegen

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

pytorch avatarpytorch
获取
skill-writer

skill-writer

97Kdev-codegen

Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md files, frontmatter, or skill structure.

pytorch avatarpytorch
获取

Implements JavaScript classes in C++ using JavaScriptCore. Use when creating new JS classes with C++ bindings, prototypes, or constructors.

oven-sh avataroven-sh
获取

Creates JavaScript classes using Bun's Zig bindings generator (.classes.ts). Use when implementing new JS APIs in Zig with JSC integration.

oven-sh avataroven-sh
获取