Discriminated Unions

Learning Objectives

After completing this lesson, you will be able to:

• Explain Discriminated Unions in terms of literal discriminants, control-flow narrowing, and never-based exhaustiveness.
• Model enrollment workflows where loading, success, rejection, and failure carry different data without weakening the contract to any.
• Trace what the compiler checks and what JavaScript remains at runtime.
• Recognize and correct this recurring failure mode: using optional fields in one broad interface and permitting impossible combinations.
• Defend when to use Discriminated Unions and when a simpler design is clearer.
• Distinguish the compile-time guarantees of Discriminated Unions from runtime behavior.
• Read and explain compiler diagnostics related to Discriminated Unions.
• Choose a simpler alternative when Discriminated Unions would add unnecessary complexity.
• Apply Discriminated Unions without weakening untrusted input to any.
• Review Discriminated Unions for maintainability in a multi-team codebase.
• Test both accepted and intentionally rejected type scenarios.
• Identify the trust boundaries around code that uses Discriminated Unions.
• Evaluate checker, build, bundle, and runtime costs separately.
• Explain the security limitations of erased TypeScript types.
• Use Discriminated Unions in a production-oriented TechLearningPro design.

Introduction

A growing TechLearningPro codebase must support enrollment workflows where loading, success, rejection, and failure carry different data. Copying loosely related types makes valid changes expensive and lets assumptions drift between the UI, application services, and API adapters. The team needs a design that expresses the relationship explicitly while remaining understandable to reviewers.

This lesson approaches Discriminated Unions as an engineering decision rather than syntax to memorize. You will connect the developer experience to the TypeScript compiler, emitted JavaScript, production boundaries, and the maintenance costs paid by a team over time.

What Is This Concept?

In simple language: A discriminated union gives each variant a shared literal tag that identifies its valid fields.

Professional explanation: Discriminated Unions is a compile-time modeling technique based on literal discriminants, control-flow narrowing, and never-based exhaustiveness. It lets the checker preserve domain relationships, reject inconsistent programs, and communicate intent without claiming that a TypeScript type validates values at runtime.

Why Do We Need It?

Without Discriminated Unions
        │
        ▼
Ambiguous intent and defects discovered late
        │
        ▼
TypeScript models the contract
        │
        ▼
Earlier feedback, safer change, clearer design

• It makes the relationship behind enrollment workflows where loading, success, rejection, and failure carry different data visible in the program.
• It moves many integration mistakes into editor and CI feedback.
• It reduces duplicated contracts that can drift during refactoring.
• It gives maintainers a precise vocabulary for reviewing design changes.
• It supports the key engineering decision: encode mutually exclusive domain states explicitly even when a flat interface looks shorter.

Real-World Analogy

A passport's document type determines which fields are valid and which processing lane it enters.

How It Works

Compile time

The checker applies literal discriminants, control-flow narrowing, and never-based exhaustiveness, resolves the resulting relationships, and reports assignments or operations that violate them. These checks happen during editing or compilation and are erased from ordinary JavaScript output.

Runtime

At runtime, Discriminated Unions has no independent type-level behavior: emitted JavaScript follows ordinary JavaScript semantics. External data still requires runtime validation.

1. 1. Identify the invariant in the requirement: enrollment workflows where loading, success, rejection, and failure carry different data.
2. 2. Represent only the information the compiler needs to preserve that invariant.
3. 3. Apply literal discriminants, control-flow narrowing, and never-based exhaustiveness and inspect inference rather than guessing it.
4. 4. Compile under strict mode and test both accepted and rejected calls.
5. 5. Inspect emitted JavaScript when runtime behavior matters.
6. 6. Validate unknown input before it enters the trusted typed core.

Architecture / Flow Diagram

Domain requirement: enrollment workflows where loading, success, rejection, and failure carry different data
        │
        ▼
Type model: Discriminated Unions
        │  compiler applies literal discriminants, control-flow narrowing, and never-based exhaustiveness
        ▼
Accepted program ──or── precise diagnostic
        │
        ▼
Emitted JavaScript (types erased)
        │
        ▼
Runtime validation at every untrusted boundary

Code Examples

type Result = { kind: "ok"; value: string } | { kind: "error"; message: string };

type Enrollment =
  | { status: "pending"; requestedAt: Date }
  | { status: "active"; enrolledAt: Date }
  | { status: "rejected"; reason: string };

function assertNever(value: never): never { throw new Error("Unexpected variant"); }
function label(state: Enrollment): string {
  switch (state.status) {
    case "pending": return "Awaiting review";
    case "active": return state.enrolledAt.toISOString();
    case "rejected": return state.reason;
    default: return assertNever(state);
  }
}

Enterprise Example

TechLearningPro uses Discriminated Unions while implementing enrollment workflows where loading, success, rejection, and failure carry different data. A boundary adapter first validates HTTP or queue payloads as unknown. The application layer then relies on the static contract, and the domain layer stays independent of transport details. Reviewers can distinguish a compile-time guarantee from authorization, validation, and other runtime controls.

Student
   │
   ▼
React / Angular UI
   │ typed command
   ▼
Application service
   │ validated DTO
   ▼
API client ─────► Runtime schema at trust boundary
   │
   ▼
Backend API

Deep Dive

literal discriminants, control-flow narrowing, and never-based exhaustiveness is useful because it preserves a relationship rather than merely replacing a long annotation with a short name. If no meaningful relationship is being enforced, the abstraction may be ceremony.

The principal design risk is using optional fields in one broad interface and permitting impossible combinations. A strong design keeps diagnostics readable, exposes a small public surface, and documents the invariant in domain language.

Discriminated Unions should end at a trust boundary. Parsed JSON, storage records, environment variables, and third-party SDK values begin as unknown; validation creates runtime evidence before a typed domain value is constructed.

The governing trade-off is encode mutually exclusive domain states explicitly even when a flat interface looks shorter. Prefer the least powerful construct that keeps invalid states unrepresentable and remains easy for another engineer to modify.

Common Mistakes

For each mistake, identify the false assumption and replace it with an explicit contract:

1. 1. Treating Discriminated Unions as runtime validation; types are erased and hostile input is unchanged.
2. 2. Using any to silence a failure instead of understanding literal discriminants, control-flow narrowing, and never-based exhaustiveness.
3. 3. Ignoring the central pitfall: using optional fields in one broad interface and permitting impossible combinations.
4. 4. Adding assertions before proving the asserted fact.
5. 5. Designing from implementation shapes instead of domain invariants.
6. 6. Publishing an abstraction whose diagnostics are harder than the duplicated code.
7. 7. Testing only successful examples and never adding compile-time negative cases.
8. 8. Coupling domain contracts to a framework, transport, or generated client unnecessarily.
9. 9. Assuming a more sophisticated type improves runtime speed; it does not.

Best Practices

• Enable strict mode and keep strictNullChecks on.
• Start with a concrete domain example before extracting an abstraction.
• Name the invariant behind enrollment workflows where loading, success, rejection, and failure carry different data.
• Document why literal discriminants, control-flow narrowing, and never-based exhaustiveness is necessary.
• Prefer unknown to any at untrusted boundaries.
• Validate external values with runtime code or a schema library.
• Keep public contracts smaller than private implementation types.
• Let inference handle local details; annotate exported boundaries.
• Use type tests for both expected success and expected failure.
• Keep compiler diagnostics understandable to the consuming team.
• Avoid assertions unless runtime evidence or construction proves them.
• Inspect generated declarations for library-facing APIs.
• Measure checker latency before blaming an advanced construct.
• Separate domain types from wire-format DTOs.
• Review optionality, mutability, and nullability deliberately.
• Revisit the decision periodically: encode mutually exclusive domain states explicitly even when a flat interface looks shorter.

Performance

Type annotations normally have no direct runtime cost because they are removed from emitted JavaScript. Performance work must separate editor/type-checking cost, compilation cost, bundle output, and actual JavaScript execution.

• Discriminated Unions normally changes checker work, not JavaScript execution speed.
• Deep composition can increase editor and CI type-checking time; measure with compiler diagnostics before simplifying.
• Runtime performance depends on emitted algorithms, allocations, I/O, and validation—not on erased annotations.
• Type-driven refactoring may enable better code, but benchmark the emitted application rather than claiming a type-level speedup.

Security

Static types improve reviewability and make invalid internal states harder to express, but they are not a security boundary. Attackers interact with the emitted JavaScript and network interfaces, not your type declarations.

• Parse untrusted input as unknown and validate structure, ranges, formats, and size at runtime.
• Keep authentication and authorization checks in executable code.
• Do not let an assertion convert attacker-controlled data into a trusted domain value.
• Avoid exposing sensitive fields merely because a projected type hides them; the runtime object may still contain them.
• Use Discriminated Unions to improve reviewability, while treating validation and policy enforcement as separate controls.

Real-World Architecture

Place Discriminated Unions in the narrowest stable layer that owns its invariant. Transport adapters validate data and map DTOs; application services coordinate use cases; domain modules expose purposeful contracts; infrastructure implements those contracts. This direction prevents a clever type from becoming an accidental dependency shared by every layer.

Interview Questions & Answers

Practical Exercise

Problem: Design an exhaustive enrollment state machine and render every state.

Difficulty: Advanced

Requirements

• Use Discriminated Unions to encode the central relationship without any.
• Accept boundary input as unknown and include a minimal runtime validation step.
• Add one valid and two intentionally rejected compile-time examples.
• Explain the emitted JavaScript behavior and one design trade-off.

Expected behavior: The valid path compiles and runs, invalid type combinations fail during checking, malformed external input is rejected by executable validation, and the design remains readable under strict mode.

Hints

• Write the invariant in one sentence before writing a type.
• Begin with the smallest operation that demonstrates literal discriminants, control-flow narrowing, and never-based exhaustiveness.
• Use satisfies or @ts-expect-error where a type test is clearer than an assertion.

The complete solution is intentionally withheld. First model the contract, compile under strict mode, and explain every assertion or escape hatch during review.

Key Takeaways

• Discriminated Unions addresses enrollment workflows where loading, success, rejection, and failure carry different data.
• Its core mechanism is literal discriminants, control-flow narrowing, and never-based exhaustiveness.
• Its guarantees are compile-time guarantees.
• Emitted JavaScript still determines runtime behavior.
• Unknown external values require runtime validation.
• The main hazard is using optional fields in one broad interface and permitting impossible combinations.
• The key trade-off is encode mutually exclusive domain states explicitly even when a flat interface looks shorter.
• Strict mode and negative type tests make the contract more reliable.
• Small public surfaces improve diagnostics and maintainability.
• Type sophistication is valuable only when it preserves a real invariant.
• Security controls and performance claims require runtime evidence.

Summary

Discriminated Unions gives TechLearningPro a precise way to model enrollment workflows where loading, success, rejection, and failure carry different data through literal discriminants, control-flow narrowing, and never-based exhaustiveness. Used with strict checking, boundary validation, and deliberate ownership, it improves change safety without pretending that erased types enforce runtime policy.

Next Lesson Preview

Next, study Narrowing and Type Guards. The next lesson extends this foundation with another production modeling technique.