rust-architecture-guide

name	rust-architecture-guide
description	Comprehensive Rust engineering guide covering priority pyramid, architecture decisions, idiomatic coding style, memory layout transparency, breakwater pattern, and physical feasibility audit. Invoke when starting Rust projects, making trade-offs, writing idiomatic code, or designing system architecture. Based on Jeet Kune Do philosophy — intercept boilerplate, flow with hardware, fold logic to maximum energy density, unity of false and real.
license	MIT
metadata	{"version":"9.1.0","philosophy":"Dialectical Materialism & Jeet Kune Do — Unity of False and Real","domain":"general Rust engineering","author":"rust-architect","rust_edition":"2024","aligned_with":["Rust API Guidelines","Rust 2024 Edition","Tokio Best Practices","Clippy pedantic/nursery","Safety-Critical Rust Coding Guidelines"]}

Rust Architecture & Engineering Decision Guide V9.1.0 — The Rust 2024 Edition

This document serves as the constitutional foundation for Rust engineering decisions. It is structured for deterministic, reproducible decision-making suitable for AI-assisted development.

V9.0.0 introduced three paradigm shifts:

1. Memory Layout Transparency (ref/30): From "ideological safety" to "materialist physics" — struct padding audit, #[repr(C)] mandate, cache-friendly design
2. Breakwater Architecture (ref/31): Facade/Core layered pattern — ergonomic facade absorbs chaos, zero-overhead core maintains determinism
3. Physical Feasibility Audit (ref/32): Mandatory audit between initiation and design — I/O budget, memory ceiling, concurrency true cost

V9.1.0 — Rust 2024 Edition Alignment:

• Adopts Rust 2024 Edition idioms: unsafe extern blocks, RPIT lifetime capture, impl Trait in return position sugar
• Aligns with Rust API Guidelines checklist (C-CASE, C-CONV, C-MACRO-VIS, C-EXAMPLE)
• Integrates Safety-Critical Rust Coding Guidelines for unsafe code governance
• Adds cargo-mutants mutation testing, kani formal verification, turmoil network chaos to advanced testing arsenal

Execution Mode

Before applying any rule, determine the execution mode:

• rapid (prototype): Enforce P0 only; anyhow in libraries, unlimited .clone(), no doc-tests
• standard (default): Enforce P0+P1; warn on P2 violations
• strict (production): Enforce P0-P3; all deviations require formal annotation

See references/00-mode-guide.md for full mode definitions.

Mandatory Output Contract

Every code generation or review output MUST end with a Decision Summary block:

## Decision Summary
- **Mode**: [rapid|standard|strict]
- **Edition**: [2021|2024] — Rust edition in use
- **Rules Applied**: [list specific rules from P0-P3]
- **Conflicts Resolved**: [PX > PY with justification, or "None"]
- **Deviations**: [list with `// DEVIATION: reason` references, or "None"]
- **Trade-offs**: [key trade-off decisions made]
- **External Alignment**: [relevant Rust API Guidelines / Safety-Critical guidelines matched]

---

1. Priority Decision Matrix

When conflicts arise, apply decisions top-down. Upper levels have absolute veto power.

Level	Core Principle	Accept	Reject
P0	Safety & Correctness	Compile-time borrow checking, Miri-verified unsafe	Any UB, race conditions, unproven unsafe
P1	Maintainability	Semantic naming, ownership transfer, trait decoupling	Deeply nested lifetimes, overly complex generics
P2	Engineering Efficiency	Box<dyn Trait>, workspace division, feature isolation	Binary bloat from monomorphization
P3	Runtime Performance	SoA layout, lock-free, SIMD, profiling-proven	Intuition-based premature optimization

---

2. Ownership & Memory Architecture Strategy

Business Logic Layer

• Prefer owned types (String, Vec<T>)
• Allow .clone() to eliminate lifetime interference on non-hotpaths
• Prohibit storing non-'static references in structs, unless struct is transient view

Performance Hotpaths Layer

• Zero-Copy Design: Cow<'a, T> or bytes::Bytes for cross-layer sharing
• Memory Layout: Enforce AoS → SoA conversion for bulk data
• Alignment: #[repr(align(64))] to eliminate false sharing

---

3. Error Handling Layering Specification

Library Crate

• Required: thiserror or manual Display + Error trait
• Prohibited: Never return anyhow::Error from public API
• Error variants must contain sufficient context

Application Binary

• Use anyhow with ? operator for error propagation
• Attach .with_context(|| ...) at error propagation points
• Register global Panic Hook at main entry

---

4. Idiomatic Style: Jeet Kune Do Coding Philosophy

Core Philosophy

Principle	Description
Intercepting Boilerplate	If logic can be expressed in 1 line of pattern matching, never use 5 lines of nesting
Economy of Motion	Every line of code should point directly to intent. Eliminate redundant intermediate variables and implicit copies
Hardware Sympathy	Leverage iterators and zero-copy types, align with the compiler's inline optimization

Agent Self-Check List

1. Are code paths flat? Can let else or ? eliminate if let nesting?
2. Are manual loops eliminated? Can collection processing be converted to iterator adapter chains?
3. Are variable scopes minimized? Can shadowing remove no-longer-needed mut?
4. Are there implicit copies? Is .to_string() or .clone() misused on hot paths?
5. Is naming stuttering? e.g., user::UserConfig should be user::Config.
6. Is Rust 2024 Edition leveraged? Are unsafe extern blocks used? Are RPIT captures explicit with use<...>?
7. Are Clippy pedantic/nursery lints addressed? For strict mode, no unjustified #[allow] on pedantic lints.

Reduction Directive

When receiving a Rust coding task, fold the logic with Jeet Kune Do intuition before output — ensuring every line of code carries maximum energy density. The goal is high signal-to-noise ratio.

---

5. Agent Execution & Self-Review Directives

MUST execute the following workflow before outputting any Rust solution:

Trade-off Analysis

// Trade-off Analysis:
// - P1 (simplicity): Using String.clone() adds ~2KB runtime allocation
// - P3 (performance): Acceptable for this cold path (config loading)
// - Decision: Clone for clarity; revisit if profiling shows bottleneck

Review Checklists

• Ownership: Unnecessary lifetime annotations removed? Large objects boxed before .await?
• Concurrency: std::sync::Mutex held during await? Blocking ops in async context?
• Error Robustness: All unwrap() have expect("REASON")? External IO handles timeout?
• Safety: Unsafe blocks have SAFETY comments? FFI uses opaque handles? catch_unwind on extern "C"?

---

6. Typical Refactoring Paths

From	To	Signal
Option/bool flags	Enum state machine	Business model stabilizing
String storage	Cow<'a, str> or Arc<str>	Hot path optimization needed
Manual for loops	Iterator chaining	Collection processing code
MVP Option fields	Type-driven state machine	Core domain logic stable

---

7. Production Observability Specification

• Tracing: All pub async fn annotated with #[tracing::instrument]; skip large objects
• Metrics: Static labels; atomic counters on hot paths; Prometheus pull model
• Panic Hook: Global hook at main; synchronous write; std::process::abort() over unwinding

---

8. Reference Files

Document Relationship Map

flowchart TD
    SKILL[SKILL.md<br/>Constitutional Foundation] --> Mode[00-mode-guide.md<br/>Execution Modes]
    SKILL --> Pyramid[01-priority-pyramid.md<br/>P0-P3 Hierarchy]
    
    Pyramid --> Conflict[02-conflict-resolution.md<br/>Conflict Scenarios]
    Pyramid --> Concurrency[11-concurrency.md<br/>Lock Spectrum]
    
    Mode --> Progressive[03-progressive-architecture.md<br/>MVP → Production]
    Mode --> Deviation[06-deviation-process.md<br/>Rule Exceptions]
    
    Concurrency --> StateMachine[07-state-machine.md<br/>Type-Driven States]
    Concurrency --> Async[12-async-internals.md<br/>Pin/Unpin/Waker]
    
    Async --> CloudInfra[Cloud Infra Guide<br/>Vertical Deepening]
    
    Pyramid --> DataArch[09-data-architecture.md<br/>Zero-Copy Principles]
    DataArch --> Performance[25-performance-tuning.md<br/>Mechanical Sympathy]
    
    StateMachine --> Newtype[08-newtype.md<br/>Semantic IDs]
    
    Conflict --> Tradeoffs[04-trade-offs.md<br/>Decision Framework]
    
    style SKILL fill:#4a90d9,stroke:#2c5aa0,color:#fff
    style Pyramid fill:#5cb85c,stroke:#4cae4c,color:#fff
    style Concurrency fill:#f0ad4e,stroke:#ec971f,color:#000
    style CloudInfra fill:#d9534f,stroke:#c9302c,color:#fff

Execution & Strategy (7)

• 00-mode-guide.md — Execution modes and transition checklists
• 01-priority-pyramid.md — The four-level hierarchy
• 02-conflict-resolution.md — Typical conflicts and resolutions
• 03-progressive-architecture.md — MVP to production migration
• 04-trade-offs.md — Decision analysis framework
• 05-glossary.md — Centralized terminology definitions
• 06-deviation-process.md — Formal rule exception handling

Architecture Patterns (14)

• State & Types

  • 07-state-machine.md — Type-driven state machines
  • 08-newtype.md — Type-safe IDs and credentials

• Data & Memory (P0-P1)

  • 09-data-architecture.md — Ownership, cloning, zero-copy, Arena basics

• Error Handling (P0-P1)

  • 10-error-handling.md — Library (thiserror) vs application (anyhow) strategies

• Concurrency & Async (P0-P3)

  • 11-concurrency.md — Lock spectrum, decision tree, async isolation (P0-P1)
  • 12-async-internals.md — Pin/Unpin, Waker, custom executors (P2-P3)

• API & Safety

  • 13-api-design.md — Public API boundaries, #[non_exhaustive], sealed traits
  • 15-ffi-interop.md — The Defense Wall: FFI safety boundaries (P0)

• Metaprogramming (P1-P2)

  • 14-metaprogramming.md — Declarative/procedural macros, const generics

• Observability

  • 16-observability.md — Tracing, metrics, panic hooks

• Tooling

  • 17-toolchain.md — CI, Clippy, workspace, feature flags, rust-toolchain.toml pinning

Modern CI & Formal Methods (V9.1.0 additions)

• 33-ci-modern.md — Modern CI/CD pipeline: cargo-mutants, kani, turmoil, deterministic RNG, rust-toolchain.toml

• Memory Layout Transparency (P0)

  • 30-memory-layout.md — Struct padding audit, #[repr(C)] mandate, cache-friendly design, false sharing prevention

• Breakwater Architecture (P0-P1)

  • 31-breakwater-pattern.md — Facade/Core layered pattern, boundary interception protocol, type contraction

• Physical Feasibility Audit (P0)

  • 32-physical-audit.md — Pre-design mandatory audit: I/O budget, memory ceiling, concurrency true cost

Idiomatic Style (7)

• 18-control-flow.md — let else, matches!, intercepting deep nesting
• 19-iterators.md — Iterator chains, filter_map, flowing force
• 20-traits.md — From vs Into, Default, Hardware Sympathy
• 21-errors.md — unwrap_or_else, map_err, and_then
• 22-data-struct.md — Field shorthand, type stuttering
• 23-borrowing.md — AsRef, Cow, memory economy
• 24-refactor.md — Agent Self-Check List, Reduction Directive

Deep Dive (P3 Only)

⚠️ Requirement: All P3 optimizations MUST include profiler data proving bottleneck

• 25-performance-tuning.md — Mechanical Sympathy: SoA, SIMD, Arena, PGO (P3 only)
• 26-advanced-testing.md — Formal verification: proptest, fuzz, Loom, Miri, kani, cargo-mutants, turmoil (strict mode)
• 27-review.md — Production review checklist
• 28-usage-examples.md — Real-world scenarios with Decision Summaries

---

9. Constitutional Summary

No rule is absolute. The ultimate measure of code quality is whether it matches:

• Its lifecycle stage (MVP vs production) — governed by execution mode
• Business constraints (deadline vs long-term maintenance)
• Team context (solo vs large organization)

The Pragmatic Principle:

Pursue excellence at system boundaries and hot paths; release mental load for internal flows and cold paths.

Deviation Protocol: When breaking a rule is the correct decision, annotate with // DEVIATION: reason and record in the Decision Summary.

Terminology: All specialized terms are defined in glossary.md.

Remember: This guide is the Agent's constitutional foundation. Before each response, generate a Decision Summary recording rationale for choosing specific implementation paths, ensuring consistency and professional depth.