Use Template Literal Types to Model DSLs and String Relationships

Overview

Template literal types bring the power of JavaScript template literals to TypeScript's type system. They allow you to model structured subsets of strings, parse domain-specific languages (DSLs), and capture relationships between string types. Combined with conditional types and the infer keyword, they enable sophisticated string manipulation at the type level.

This skill is essential for bringing type safety to string-heavy APIs and for building powerful type transformations.

When to Use This Skill

• Modeling structured string patterns (IDs, paths, URLs)
• Parsing domain-specific languages (CSS selectors, query languages)
• Transforming string types (camelCase, snake_case conversion)
• Validating string formats at compile time
• Combining with mapped types for key transformations

The Iron Rule

Use template literal types to model structured string subsets and DSLs. Combine with infer for parsing and mapped types for transformations.

Detection

Watch for these opportunities:

// RED FLAGS - Untyped strings that could be precise
type EventType = string;  // Could be 'click' | 'hover' | etc.
function query(selector: string): Element;  // Could parse CSS selectors
type CSSProperty = string;  // Could validate property names

Basic Template Literal Types

// Match strings starting with a prefix
type PseudoString = `pseudo${string}`;
const science: PseudoString = 'pseudoscience';  // OK
const alias: PseudoString = 'pseudonym';        // OK
const physics: PseudoString = 'physics';        // Error!

// Match specific patterns
type DataAttribute = `data-${string}`;
type HTTPSUrl = `https://${string}`;
type VersionString = `v${number}.${number}.${number}`;

Index Signatures with Template Literals

// Allow data-* attributes while keeping type safety
interface Checkbox {
  id: string;
  checked: boolean;
  [key: `data-${string}`]: unknown;
}

const check: Checkbox = {
  id: 'subscribe',
  checked: true,
  'data-listIds': 'all-the-lists',  // OK
  value: 'yes',  // Error: not data-* and not known property
};

Parsing with infer

Extract parts of strings using conditional types with infer:

// Extract event name from handler type
type EventName<T> = T extends `on${infer Name}` ? Name : never;

type ClickEvent = EventName<'onClick'>;      // 'Click'
type HoverEvent = EventName<'onMouseEnter'>; // 'MouseEnter'
type BadEvent = EventName<'handleClick'>;   // never

// Extract path parameters
type PathParams<T> = T extends `/users/${infer UserId}/posts/${infer PostId}`
  ? { userId: UserId; postId: PostId }
  : never;

type Params = PathParams<'/users/123/posts/456'>;
// { userId: '123'; postId: '456' }

String Transformations

Build recursive types to transform strings:

// Convert snake_case to camelCase
type CamelCase<S extends string> =
  S extends `${infer Head}_${infer Tail}`
    ? `${Head}${Capitalize<CamelCase<Tail>>}`
    : S;

type T1 = CamelCase<'foo'>;           // 'foo'
type T2 = CamelCase<'foo_bar'>;      // 'fooBar'
type T3 = CamelCase<'foo_bar_baz'>;  // 'fooBarBaz'

// Apply to object keys
type CamelCaseKeys<T> = {
  [K in keyof T as CamelCase<K & string>]: T[K]
};

type SnakeCase = { user_name: string; email_address: string };
type Camel = CamelCaseKeys<SnakeCase>;
// { userName: string; emailAddress: string }

Real-World Example: CSS Selectors

// Enhance querySelector with precise types
type HTMLTag = keyof HTMLElementTagNameMap;

declare global {
  interface ParentNode {
    // Simple tag selector
    querySelector<TagName extends HTMLTag>(
      selector: TagName
    ): HTMLElementTagNameMap[TagName] | null;
    
    // Tag#id selector
    querySelector<TagName extends HTMLTag>(
      selector: `${TagName}#${string}`
    ): HTMLElementTagNameMap[TagName] | null;
  }
}

// Usage
const img = document.querySelector('img#hero');
// Type: HTMLImageElement | null
// Can access img?.src, img?.alt, etc.

const div = document.querySelector('div#container');
// Type: HTMLDivElement | null

Combining with Mapped Types

// Create event handler types from event names
type EventMap = {
  click: MouseEvent;
  keydown: KeyboardEvent;
  submit: SubmitEvent;
};

type EventHandlers<Events extends Record<string, Event>> = {
  [K in keyof Events as `on${Capitalize<K & string>}`]?: 
    (event: Events[K]) => void;
};

type Handlers = EventHandlers<EventMap>;
// {
//   onClick?: (event: MouseEvent) => void;
//   onKeydown?: (event: KeyboardEvent) => void;
//   onSubmit?: (event: SubmitEvent) => void;
// }

Pressure Resistance Protocol

When pressured to use simple string types:

1. Identify patterns: What structure do the strings have?
2. Start simple: Use unions of literal types first
3. Add templates: Use template literals for infinite but structured sets
4. Consider parsing: Use infer to extract information
5. Test edge cases: Ensure your types are accurate, not just precise

Red Flags

Common Rationalizations

"String is good enough"

Reality: Template literals catch typos and invalid formats at compile time. 'user-123' vs 'users-123' can be caught immediately.

"This is too complex"

Reality: Start simple with prefix patterns, then add complexity as needed. Even basic template literals provide value.

"It will hurt performance"

Reality: Template literal types are evaluated at compile time. They have no runtime cost.

Quick Reference

The Bottom Line

Template literal types bring type safety to string-heavy code. Use them to model structured strings, parse DSLs, and transform types. Combined with infer and mapped types, they enable powerful type-level string manipulation.

Reference

• Effective TypeScript, 2nd Edition by Dan Vanderkam
• Item 54: Use Template Literal Types to Model DSLs and Relationships Between Strings