Навык racket-programmer

Racket Programmer

Expert-level Racket programming skill focused on judgment frameworks, language-oriented programming philosophy, and production practices.

Related skills:

• software-engineer - Core programming philosophy and SICP principles that underpin Racket's design
• functional-programmer - FP patterns and thinking (Racket is fundamentally functional)
• test-driven-development - General testing philosophy; this skill covers RackUnit specifics

Version Targeting

Always target the latest stable Racket version unless explicitly working on a codebase with version constraints.

• Latest stable: Racket 8.18 (as of August 2025)
• Download: https://download.racket-lang.org/
• Version history: https://github.com/racket/racket/releases
• For historical context: https://en.wikipedia.org/wiki/Racket_(programming_language)#History

Racket CS (Chez Scheme-based) is the default implementation since 8.0. Strong backward compatibility commitment means minimal version fragmentation.

For owned codebases: Aggressive upgrade philosophy. Use latest features and libraries.

For third-party codebases: Respect existing version targets and conventions. When contributing to open source, follow the project's existing standards. Don't introduce modern features to projects targeting older versions. Propose improvements through proper channels (issues, governance).

<language_oriented_programming>

Language-Oriented Programming: Racket's Core Philosophy

Central Insight: Racket's defining characteristic is building solutions as languages, not just libraries. Every problem is approached by asking "what's the right notation?" before "what's the implementation?"

This is not a feature of Racket—it IS Racket's design philosophy. Understanding this distinction separates using Racket from merely writing Scheme with extra features.

From "The Racket Manifesto" (Felleisen et al., SNAPL 2015)

"A programmable programming language." Racket is designed to let programmers create languages tailored to specific problem domains. The language grows through the creation of new languages, not just new libraries.

Key principle: "If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization." —Gerald Weinberg

The right notation eliminates entire classes of errors at compile time. Language design is not an esoteric activity—it's the solution to the problem of making correct code easy to write and incorrect code hard to write.

Decision Framework: #lang vs Library vs Macro

Does solution need non-S-expression syntax?
  → Reader extension needed → #lang

Does solution benefit from compile-time validation of entire program?
  → Custom #%module-begin → #lang

Just adding new operations/abstractions?
  → Macros + provide → library

Need IDE integration (syntax coloring, specialized checks)?
  → #lang with language info

Notation precision outweighs learning curve?
  → #lang

When Language-Oriented Design Excels

1. Novice interfaces - Create simpler notations for less-skilled programmers (HtDP teaching languages)
2. Simplified notation - Regular expressions, configuration DSLs where domain notation is clearer
3. Existing notation - BNF grammars (brag package), standard formats that users already know
4. Intermediate compilation targets - JSON with embedded computation, specialized formats
5. Configuration-heavy domains - Test specifications, API definitions where structure matters

Essential Reading:

• Beautiful Racket: https://beautifulracket.com/ (comprehensive LOP tutorial—read this)
• "Why Language-Oriented Programming?": https://beautifulracket.com/appendix/why-lop-why-racket.html
• "Creating Languages in Racket" (Flatt): https://queue.acm.org/detail.cfm?id=2068896

Trade-offs:

• ✅ Perfect notation-to-problem fit, early error detection via compiler, domain expert usability
• ❌ Learning curve, infrastructure maintenance overhead, tooling support complexity
• Decision heuristic: Use when minimum notation with maximum precision is critical

Language Towers: Composable Language Abstractions

Pattern: Build languages on languages, creating towers where each layer provides abstractions for the next. Unlike traditional macros that compete, Racket macros cooperate through shared compile-time information.

Key Paper: "Macros that Work Together" (Flatt et al., 2012)

Racket achieves this through:

• Partial expansion: Macros can expand sub-forms partially, inspect results, continue with additional context
• Definition contexts: Macros cooperate by maintaining lexical scope through hygiene
• Phase separation: Compile-time vs runtime code cleanly separated

Example: Racket's class system is implemented as macros over struct, which is itself a macro. Each layer maintains proper scoping and composition.

This is a fundamental difference from other Lisp macro systems where macros often interfere with each other. </language_oriented_programming>

<contracts_and_blame>

Code as Contracts: Module Boundaries and Blame

Core Pattern: Modules are units of abstraction AND units of blame for contracts. Design module boundaries intentionally with this in mind.

Contracts in Racket are not optional documentation—they are executable specifications that assign blame when violated. This is fundamentally different from assertions (which blame the entire program) or types (which reject programs statically).

Contract-Out: The Standard Pattern

(provide
  (contract-out
    [my-function (→ number? (>/c 0) boolean?)]
    [my-struct (struct/c point [real? real?])]
    [complex-fn (→i ([x number?]
                      [y (x) (>/c x)])  ; dependent contract
                     [result number?])]))

Decision Framework for Contracts

Public API boundary?
  → REQUIRED: contract-out with precise contracts

Internal module function?
  → Optional: Consider for complex invariants

Performance-critical inner loop?
  → Profile first (contracts typically cheap, but measure)

Complex invariant (e.g., balanced tree)?
  → Use dependent contracts (→i)

Simple type-like guarantee?
  → Flat contracts (number?, string?)

Contract Hierarchy

1. Always: Module boundaries (public APIs) - MANDATORY for production
2. Often: Complex data structures with invariants
3. Sometimes: Internal boundaries in large modules
4. Rarely: Simple, obvious functions

Chaperone vs Impersonator Contracts:

• Chaperone: Monitoring only, guarantees no behavior change (preferred)
• Impersonator: Can change behavior, heavier weight
• Use chaperones unless you need interposition

Key Resources:

• Contracts Guide: https://docs.racket-lang.org/guide/contracts.html
• Contract Reference: https://docs.racket-lang.org/reference/contracts.html
• "Contracts for Higher-Order Functions" (Findler & Felleisen): Foundational paper </contracts_and_blame>

<macro_system>

Macro System: Production Standards

Decision Matrix

Need	Tool	When to Use
Simple substitution	define-syntax-rule	One-off macros, prototyping
Multiple patterns	syntax-rules	Standard pattern matching, no validation
Validation + errors	syntax/parse	Production macros (recommended)
Complex compile logic	define-syntax + procedure	Programmatic code generation
Syntax templates	with-syntax, quasisyntax	Building syntax from data

Syntax/Parse: The Production Standard

(define-syntax (my-let stx)
  (syntax-parse stx
    [(_ ([var:id rhs:expr] ...) body:expr ...+)
     #:fail-when (check-duplicate-identifier (syntax->list #'(var ...)))
                 "duplicate variable name"
     #'(let ([var rhs] ...) body ...)]))

Why syntax/parse for production:

• ✅ Declarative pattern matching with side conditions
• ✅ Domain-specific error messages via #:fail-when
• ✅ Syntax classes enable reusable patterns
• ✅ Better than syntax-rules for anything non-trivial

Essential Resources:

• "Fear of Macros" (Greg Hendershott): https://www.greghendershott.com/fear-of-macros/all.html (best macro tutorial—read this)
• syntax/parse docs: https://docs.racket-lang.org/syntax/stxparse.html
• Macro debugger: https://docs.racket-lang.org/macro-debugger/

Hygiene: Racket macros are hygienic by default. This prevents variable capture but means you must understand syntax objects. Use datum→syntax only when intentionally breaking hygiene (rare cases like anaphoric macros).

Reader Extensions: High Power, High Cost

Decision Tree:

Modify single character interpretation?
  → Readtable (make-readtable)

Embedded DSL with different syntax?
  → Readtable + #reader

Complete language with custom notation?
  → #lang + reader module

Compatible with S-expressions?
  → DON'T use reader - use macros instead!

When to Extend Reader:

• Need non-S-expression syntax (BNF, JSON-like, etc.)
• Domain notation is well-established and incompatible with parens
• Target audience uncomfortable with S-expressions
• Notation is fundamentally incompatible with Lisp syntax

When NOT to:

• Macros suffice (most cases!)
• Syntax is still S-expressions with new forms
• Team not ready for the complexity
• Tool support matters more than notation

Trade-offs:

• ✅ Perfect notation for domain, can parse existing formats
• ❌ Complexity, harder debugging, less tool support, team must learn custom syntax </macro_system>

<structs_vs_classes>

Structs vs Classes: The 90/10 Rule

Decision Framework:

Representing data with operations?
  → Struct (90% of use cases)

Need inheritance-based polymorphism with method overriding?
  → Class

Building GUI application?
  → Class (Racket GUI framework uses classes)

Need polymorphism but not inheritance?
  → Struct with generic interfaces (#:methods gen:custom)

Porting OO codebase?
  → Class (but consider refactoring to structs)

Struct Advantages

• ✅ Lighter weight, faster
• ✅ Work with pattern matching
• ✅ Simpler mental model
• ✅ Can add methods via #:methods with generics
• ✅ Transparent/opaque control

Class Use Cases

• Full OO with inheritance
• Method overriding, super calls
• Mixins and traits
• GUI programming (required by framework)

Hybrid Approach: Structs + Generic Interfaces

(define-generics printable
  (print-it printable port))

(struct point (x y)
  #:methods gen:printable
  [(define (print-it p port)
     (fprintf port "(~a,~a)" (point-x p) (point-y p)))])

Staff Guidance: Default to structs. Only use classes when working with GUI framework, need true inheritance, or building OO framework for others. When unsure, start with structs + generics. </structs_vs_classes>

<typed_racket>

Typed Racket: When and Why

Decision Framework

Performance-critical numeric code?
  → YES (25-50× speedups possible)

Large-scale refactoring/maintenance?
  → YES (types document invariants)

Complex data structure invariants?
  → YES (occurrence typing + refinements)

Heavy typed/untyped interaction?
  → CAUTION (contract overhead can be severe)

Exploratory/prototype code?
  → NO (friction not worth it)

Macro-heavy DSLs?
  → NO (limitations on macro usage)

Language Declaration

#lang typed/racket              ; Deep types (default, full checking)
#lang typed/racket/shallow      ; Shallow types (lower overhead)
#lang typed/racket/optional     ; Optional types (no runtime cost)

Occurrence Typing (Unique Feature)

Racket's type system has occurrence typing—type refinement based on predicates:

(: flexible-length (→ (U String (Listof Any)) Integer))
(define (flexible-length x)
  (if (string? x)
      (string-length x)  ; x refined to String here
      (length x)))       ; x refined to (Listof Any) here

This is a distinctive feature compared to most gradual type systems.

Performance Best Practices

• Use Float not Real for numeric optimization
• Prefer structs to vectors (always optimized)
• Use typed versions of standard libraries
• Minimize boundary crossings

Essential Resources:

• Typed Racket Guide: https://docs.racket-lang.org/ts-guide/
• Contract Profiler: https://docs.racket-lang.org/contract-profile/
• Optimization Coach: https://docs.racket-lang.org/optimization-coach/ </typed_racket>

<functional_idioms>

Racket-Specific Functional Idioms

What Makes Racket FP Unique (not generic FP advice that you already know):

1. Contracts as First-Class Behavioral Specifications

Not just type checking; express behavioral constraints. Contracts are values that can be computed and composed.

2. Macros Enable Compile-Time FP

Compile-time computation is functional programming. Phase separation keeps concerns clean.

3. Parameters for Dynamic Binding

Better than global variables for cross-cutting concerns:

;; Standard pattern for context
(parameterize ([current-output-port (open-output-file "log.txt")])
  (displayln "logged"))

Example: current-output-port, current-directory

4. Sequence Abstractions

;; Prefer for/* comprehensions over explicit recursion
(for/list ([x (in-range 10)]
           #:when (even? x))
  (* x x))

5. Match-Based Dispatch

(define (eval expr)
  (match expr
    [(? number? n) n]
    [`(+ ,a ,b) (+ (eval a) (eval b))]
    [`(* ,a ,b) (* (eval a) (eval b))]))

When to Mutate in Racket

• Algorithm is inherently stateful (union-find, hash table implementation)
• Performance-critical after profiling shows bottleneck
• Implementing imperative interface for interop
• Default: Prefer functional style at module boundaries, mutation only internally </functional_idioms>

<pattern_matching>

Pattern Matching: When and How

Use match when:

• Deconstructing structured data (lists, structs, syntax)
• Dispatch on data shape
• Working with recursive data structures
• Replacing verbose accessor chains

Don't use match for:

• Simple conditionals (use cond or if)
• Single binding extraction (use let)
• Just checking predicates (use predicate directly)

Key Patterns:

;; Struct patterns
(match point
  [(point x y) ...])

;; Quasiquote for S-expressions
(match expr
  [`(if ,cond ,then ,else) ...])

;; Guards
(match x
  [(? number? n) ...]
  [(list a b) #:when (> a b) ...])

;; Or patterns
(match expr
  [(or (list 'add x y) (list '+ x y)) ...])

</pattern_matching>

<tooling_and_standards>

Mandatory Tooling and Standards

#lang racket/base vs #lang racket (CRITICAL)

Decision Rule:

Production library code → #lang racket/base (REQUIRED)
Scripts/demos/exploration → #lang racket (acceptable)

Why racket/base for libraries:

• Faster startup (loads only core)
• Explicit dependencies (clearer what's needed)
• Smaller dependency footprint
• Better practice for maintainability

Quote from Beautiful Racket: "As a rule of thumb, it's better practice to start with racket/base and use require to import what you need."

raco fmt: The Automatic Formatter (Mandatory)

Installation: raco pkg install fmt

Usage:

raco fmt -i file.rkt                  # Format in-place
raco fmt --width 102 file.rkt         # Custom width (style guide default)
raco fmt --check file.rkt             # Check without modifying (CI usage)

CI Integration: Add raco fmt --check to fail builds on formatting violations

What it Enforces:

• Default 80-character width (use --width 102 for style guide compliance)
• Max 1 consecutive blank line
• Converts to brackets [...] in specific contexts: let, for, cond, match, case
• Handles 183+ standard forms

raco review: Linting (Mandatory for CI)

Installation: raco pkg install review

Usage:

raco review file.rkt
raco review .                # Review entire directory

What It Detects:

• Syntax errors
• Unused identifiers
• Variable shadowing
• Missing else clauses
• Incorrect require ordering
• Style issues

Ignoring Warnings:

#|review: ignore|#           ; Ignore entire file
(define x 5)  ;; review: ignore  ; Ignore line

RackUnit: Testing (Mandatory)

For general testing philosophy, see the test-driven-development skill. Core principle: Tests are contracts that document expected behavior at module boundaries. This skill covers RackUnit-specific patterns.

Basic Usage:

(require rackunit)

;; Simple checks
(check-equal? actual expected "message")
(check-= actual expected epsilon)
(check-true value)
(check-pred predicate value)
(check-exn exn-predicate (thunk ...))

Test Suites:

(define my-tests
  (test-suite
   "Suite name"
   (test-case "Test 1"
     (check-equal? (foo 5) 10))
   (test-case "Test 2"
     (check-true (bar "test")))))

(run-tests my-tests)

module+ test Pattern:

#lang racket/base

(define (my-function x)
  (+ x 1))

(provide my-function)

(module+ test
  (require rackunit)
  (check-equal? (my-function 5) 6)
  (check-equal? (my-function 0) 1))

Running Tests:

raco test file.rkt              # Test single file
raco test -p package-name       # Test entire package

Scribble: Documentation (Mandatory for Production)

Basic Document Structure:

#lang scribble/manual
@(require (for-label racket my-lib))

@title{My Library}

@defmodule[my-lib]

@defproc[(my-function [arg1 string?] [arg2 integer?])
         boolean?]{
  Description of what the function does.

  @examples[
    #:eval my-eval
    (my-function "hello" 42)
  ]
}

MANDATORY: All exported bindings in production code must have Scribble documentation.

Building:

scribble --html manual.scrbl
raco setup --doc-index --pkgs my-package

Code Style: Mandatory for Production

Official Style Guide

Primary Resource: https://docs.racket-lang.org/style/index.html

Critical Rules:

1. Parentheses: ALL closing parens on LAST line (never on separate lines)
2. Indentation: Use DrRacket's "indent all" as STANDARD (no tabs, spaces only)
3. Line width: 102 characters maximum
4. Naming: kebab-case with full English words (NOT camelCase, NOT snake_case)
5. End files with newline, no trailing whitespace

Naming Conventions (Mandatory)

Character	Meaning	Example
?	Predicates	boolean?, even?
!	Mutators	set!, vector-set!
%	Classes	button%, frame%
<>	Interfaces	dc<%>, window<%>
^	Unit signatures	game-context^
@	Units	testing-context@
#%	Kernel forms	#%app, #%module-begin
/	"with" preposition	call/cc, for/list

Standard Pattern: type-operation (e.g., string-append, hash-ref, list->vector)

File Organization

#lang racket/base          ; Language line (first line)

;; Exports (provide forms first, ideally)
(provide (contract-out
           [function (→ Type Type)]
           [struct-name ...]))

;; Imports (require forms grouped)
(require racket/list
         racket/match
         (for-syntax racket/base))

;; Implementation
(define ...)
(struct ...)

Project Layout: The lib/test/doc Pattern

Multi-Collection Package Structure (Production Standard)

For non-trivial libraries:

my-library/
  my-library-lib/         # Runtime code only
    info.rkt
    main.rkt
    core/
      module1.rkt

  my-library-test/        # Tests only
    info.rkt
    tests/
      test-module1.rkt

  my-library-doc/         # Documentation only
    info.rkt
    scribblings/
      my-library.scrbl

  my-library/             # Composite meta-package
    info.rkt              # Aggregates all components

Why This Pattern:

1. ✅ Focused dependencies - Apps depend only on -lib, not test/doc deps
2. ✅ Minimizes failure points - Doc build failures don't break library installation
3. ✅ Efficiency - Users download only what they need
4. ✅ Flexibility - Can install just tests or just docs

Authoritative Resource: https://countvajhula.com/2022/02/22/how-to-organize-your-racket-library/

Single-Collection Package (Simple Projects)

For small libraries:

my-simple-lib/
  main.rkt
  helper.rkt
  tests/
    main-test.rkt
  scribblings/
    my-simple-lib.scrbl
  info.rkt

Package Management: raco pkg

Core Commands:

raco pkg install package-name     # From catalog
raco pkg install github://user/repo/branch
raco pkg update package-name
raco pkg update --all
raco pkg remove package-name --auto  # Remove + unused deps
raco pkg show --all                  # List all packages

Declaring Dependencies in info.rkt (MANDATORY):

#lang info

(define deps
  '("base"                           # Always required
    "rackunit-lib"
    "web-server-lib"
    ["other-package" "2.0"]))        # Version constraint

(define build-deps
  '("scribble-lib"                   # For docs
    "racket-doc"))                   # For doc cross-refs

Key Fields:

• deps: Runtime dependencies (REQUIRED)
• build-deps: Build-time only (tests, docs)
• implies: Re-exports for meta-packages </tooling_and_standards>

<ci_cd_integration>

CI/CD Integration (MANDATORY for Production)

GitHub Actions Standard Setup:

name: CI

on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    strategy:
      matrix:
        racket-version: ['8.16', '8.17', 'stable']
        variant: ['CS']

    steps:
      - uses: actions/checkout@v3

      - name: Setup Racket
        uses: Bogdanp/setup-racket@v1.14
        with:
          architecture: 'x64'
          distribution: 'full'
          variant: ${{ matrix.variant }}
          version: ${{ matrix.racket-version }}

      - name: Install Dependencies
        run: raco pkg install --auto --batch

      - name: Check Dependencies
        run: raco setup --check-pkg-deps --pkgs my-package

      - name: Run Tests
        run: raco test --package my-package

      - name: Build Documentation
        run: raco setup --doc-index --pkgs my-package

      - name: Lint Code
        run: |
          raco pkg install --auto review
          raco review .

      - name: Check Formatting
        run: |
          raco pkg install --auto fmt
          raco fmt --check .

What MUST Fail Builds:

1. ✅ raco test failures (REQUIRED)
2. ✅ raco setup --check-pkg-deps failures (REQUIRED)
3. ✅ raco review violations (RECOMMENDED)
4. ✅ raco fmt --check failures (if using formatter)
5. ✅ Documentation build failures (RECOMMENDED) </ci_cd_integration>

<common_mistakes>

Common Mistakes from Other Languages

<from_scheme>

From Scheme (R5RS/R6RS/R7RS)

Key Differences:

1. Immutable Pairs by Default: Racket pairs are immutable. No set-car!/set-cdr!. Use mcons/mcar/mcdr for mutable pairs.

2. Module System: Racket's modules are NOT R6RS-compatible. Must use #lang racket not R6RS module syntax.

3. Letrec Semantics: Racket's letrec is sequential (like R6RS letrec*), not R5RS letrec.

4. Case Sensitivity: Racket is case-sensitive by default (R5RS was case-insensitive).

5. Pattern Matching: Not in standard Scheme; heavily used in Racket. Learn match.

Resources:

• Standards Compatibility: https://docs.racket-lang.org/guide/standards.html </from_scheme>

<from_common_lisp>