m04-zero-cost

Original：🇺🇸 English

Translated

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

2installs

Sourcezhanghandong/rust-skills

Added on2026-02-04

NPX Install

npx skill4agent add zhanghandong/rust-skills m04-zero-cost

SKILL.md Content

View Translation Comparison →

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:

What abstraction is needed?
- Same behavior, different types → trait
- Different behavior, same type → enum
- No abstraction needed → concrete type
When is type known?
- Compile time → generics (static dispatch)
- Runtime → trait objects (dynamic dispatch)
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

rust

// 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