Naming Forge

"There are only two hard things in Computer Science: cache invalidation and naming things." — Phil Karlton

What It Does

You need to name a function. It takes a list of orders, filters out cancelled ones, groups them by customer, and returns the totals. You stare at the cursor. processOrders? Too vague. filterAndGroupAndSummarizeOrdersByCustomerExcludingCancelled? Too long. getOrderSummary? Misleading — it does more than "get."

Naming Forge generates names that are precise (says what it does), consistent (matches your codebase's conventions), discoverable (other developers can guess what to search for), and proportional (name length matches scope importance).

The Five Laws of Good Names

Law 1: A Name Should Be a Contract

The name is a promise. calculateTotal promises it calculates and returns a total. If it also sends an email, the name is a lie.

BAD:  updateUser()          — does it update the DB? The UI? Both? What fields?
GOOD: saveUserProfile()     — saves the user's profile (to persistent storage)
GOOD: refreshUserDisplay()  — updates what the user sees on screen

Law 2: Scope Determines Length

Small scope → short name. Large scope → descriptive name. A loop variable can be i. A public API method should not be.

Scope	Name Length	Example
Loop variable (3 lines)	1-2 chars	i, ch, tx
Local variable (10-20 lines)	1 word	total, users, query
Private method (module scope)	1-2 words	parseInput, buildQuery
Public method (cross-module)	2-3 words	calculateOrderTotal, validateAddress
Exported constant (global)	2-4 words	MAX_RETRY_ATTEMPTS, DEFAULT_TIMEOUT_MS

Law 3: Verbs for Actions, Nouns for Things

Functions do things → verbs. Variables hold things → nouns. Types describe things → nouns/adjectives.

FUNCTIONS (verb-first):
├── get/fetch/load    → retrieves data (getSessions, fetchUser, loadConfig)
├── set/update/save   → modifies data (setTheme, updateProfile, saveOrder)
├── create/build/make → constructs new things (createUser, buildQuery, makeHandler)
├── delete/remove     → eliminates things (deleteAccount, removeItem)
├── is/has/can/should → returns boolean (isValid, hasPermission, canEdit)
├── parse/format/transform → converts between formats (parseJSON, formatDate)
├── validate/check/verify → confirms correctness (validateEmail, checkStatus)
└── handle/process/on → responds to events (handleClick, processPayment, onSubmit)

VARIABLES (noun/adjective):
├── Collections: plural nouns (users, orders, activeConnections)
├── Singles: singular nouns (user, order, currentConnection)
├── Booleans: is/has/can prefix (isLoading, hasErrors, canSubmit)
├── Counts: noun + Count (retryCount, errorCount, userCount)
└── Maps/Indices: noun + By + Key (userById, ordersByDate)

Law 4: Consistency Beats Cleverness

If your codebase says fetchUser, don't introduce retrieveUser. If it says isValid, don't introduce checkValidity. Match what exists.

CODEBASE AUDIT:
├── Existing pattern: fetch* for API calls → USE fetchOrders, NOT getOrders
├── Existing pattern: *Service for modules → USE PaymentService, NOT PaymentManager
├── Existing pattern: on* for handlers → USE onSubmit, NOT handleSubmit
└── Existing pattern: is* for booleans → USE isActive, NOT active or checkActive

Law 5: Avoid Semantic Noise

Words that add length without adding meaning: data, info, item, thing, object, value, manager, handler, processor, helper, utils.

BAD:  userData, userInfo, userObject  → just "user"
BAD:  orderItem  → just "order" (unless distinguishing from order summary)
BAD:  StringHelper, DateUtils  → what do they actually do? Be specific.
GOOD: formatDate, parseCSV, slugify  → specific actions

The Forge Process

INPUT: What does this thing do? (natural language description)
CONTEXT: What part of the codebase is this in?

Phase 1: SEMANTIC EXTRACTION
├── Extract the core action or concept from the description
├── Identify: Is this a function, variable, type, file, or route?
├── Identify: What's the scope? (local, module, public, global)
└── Identify: What domain vocabulary applies? (business terms, tech terms)

Phase 2: CONVENTION SCAN
├── Scan existing codebase for naming patterns:
│   ├── Verb preferences (get vs. fetch vs. load vs. retrieve)
│   ├── Noun preferences (User vs. Account vs. Profile)
│   ├── Casing convention (camelCase, snake_case, PascalCase, kebab-case)
│   ├── Prefix/suffix patterns (is*, *Service, *Controller, I* for interfaces)
│   └── Domain vocabulary already in use
├── Identify the dominant convention for this type of name
└── Flag any existing inconsistencies

Phase 3: CANDIDATE GENERATION
├── Generate 3-5 candidates following conventions
├── For each candidate, evaluate:
│   ├── Precision: Does the name accurately describe the thing?
│   ├── Consistency: Does it match existing patterns?
│   ├── Discoverability: Could a teammate guess this name?
│   ├── Proportionality: Is the length right for the scope?
│   └── Uniqueness: Does it conflict with any existing name?
├── Rank candidates by composite score
└── Flag tradeoffs between candidates

Phase 4: RECOMMENDATION
├── Top recommendation with rationale
├── Runner-up alternatives
├── Names to AVOID (and why)
└── If renaming: migration impact (how many references to update)

Domain-Specific Naming

API Routes / Endpoints

PATTERN: /resource/action or RESTful convention

GOOD:
├── GET    /users              → list users
├── GET    /users/:id          → get specific user
├── POST   /users              → create user
├── PUT    /users/:id          → replace user
├── PATCH  /users/:id          → partial update
├── DELETE /users/:id          → delete user
├── POST   /users/:id/verify   → action on user (verb as sub-resource)

BAD:
├── GET /getUsers              → verb in path (GET already implies "get")
├── POST /createNewUser        → redundant (POST already implies "create")
├── GET /user_list             → inconsistent casing, use /users
└── POST /doUserVerification   → too verbose, use /users/:id/verify

Database Tables / Columns

TABLES:
├── Plural nouns: users, orders, payments (not user, order, payment)
├── Join tables: user_roles, order_items (alphabetical or parent_child)
├── Consistent casing: snake_case for SQL, camelCase for NoSQL (match your ORM)

COLUMNS:
├── Foreign keys: user_id, order_id (table_singular + _id)
├── Booleans: is_active, has_verified, can_edit (prefix with state verb)
├── Timestamps: created_at, updated_at, deleted_at (past_participle + _at)
├── Counts: login_count, retry_count (noun + _count)
└── Status: order_status, payment_state (entity + _status/_state)

CSS Classes

PATTERN: BEM (block__element--modifier) or utility classes

GOOD:
├── .card__header--highlighted
├── .nav__link--active
├── .form__input--error
└── .btn--primary, .btn--disabled

BAD:
├── .redButton        → visual, not semantic
├── .leftSidebar      → positional, not semantic
├── .big              → relative to what?
└── .myComponent      → "my" adds nothing

Environment Variables

PATTERN: SCREAMING_SNAKE_CASE, grouped by service

GOOD:
├── DATABASE_URL, DATABASE_POOL_SIZE
├── REDIS_HOST, REDIS_PORT, REDIS_PASSWORD
├── SMTP_HOST, SMTP_PORT, SMTP_FROM_ADDRESS
├── APP_SECRET_KEY, APP_DEBUG, APP_LOG_LEVEL

BAD:
├── db, DB             → too short, too ambiguous
├── databaseUrl        → wrong casing for env vars
├── MY_APP_SETTING     → "MY" adds nothing
└── ENABLE_FEATURE_X   → what feature? be specific

Output Format

╔══════════════════════════════════════════════════════════════╗
║                      NAMING FORGE                           ║
║  Input: "function that takes a list of orders, filters      ║
║  out cancelled ones, groups by customer, returns totals"    ║
║  Scope: Public method in OrderService                       ║
╠══════════════════════════════════════════════════════════════╣
║                                                              ║
║  RECOMMENDATION: summarizeActiveOrdersByCustomer             ║
║  ├── Precise: "summarize" = aggregate, "active" = not        ║
║  │   cancelled, "by customer" = grouping key                 ║
║  ├── Convention match: codebase uses *ByField pattern ✓      ║
║  ├── Discoverable: searching "order" + "customer" finds it ✓ ║
║  └── Proportional: 5 words for a public cross-module method ✓║
║                                                              ║
║  ALTERNATIVES:                                               ║
║  ├── getCustomerOrderTotals — shorter but loses "active"     ║
║  ├── aggregateOrdersByCustomer — "aggregate" is less common  ║
║  │   in this codebase (0 uses vs 12 uses of "summarize")     ║
║  └── calculateCustomerOrderSummary — "calculate" implies     ║
║      math; this is more filter + group                       ║
║                                                              ║
║  AVOID:                                                      ║
║  ├── processOrders — "process" is meaningless                ║
║  ├── getOrderData — "data" is noise                          ║
║  └── doOrderStuff — please                                   ║
╚══════════════════════════════════════════════════════════════╝

When to Invoke

• When you're staring at a cursor trying to name something
• When you're about to name something temp, data, result, or thing
• When refactoring and you realize names no longer match behavior
• When reviewing code with misleading names
• When starting a new module and setting naming conventions
• When naming database tables, API endpoints, or environment variables

Why It Matters

Names are the primary documentation of a codebase. Developers read names thousands of times a day. A precise name eliminates the need to read the implementation. A misleading name causes more damage than no name at all.

You'll spend 10 minutes agonizing over a name, or you'll spend 10 hours explaining what processData actually does. The Forge is faster than either.

Zero external dependencies. Zero API calls. Pure linguistic and codebase analysis.