阅读中文版本

Build a JSON Schema Inferer

JSON Schema Inferer — a complete walkthrough of schema inference, covering sample reader, type inferer, structure merger, and testing.

json_schema_infer/
├── src/index.ts
├── src/sample-reader.ts
├── src/type-inferer.ts
├── tests/json_schema_infer.test.ts
└── package.json

// Sample reader: schema inference core
export function samplereader(input: SamplereaderInput): SamplereaderResult {
  // Build the sample reader that anchors schema inference and defines the core data shape.
  return processSamplereader(input);
}

// Type inferer — handles schema inference Add the type inferer so
export class Typeinferer {
  constructor(private opts: TypeinfererOptions) {}
  async run(ctx: Context): Promise<TypeinfererOutput> {
    return this.execute(ctx);
  }
}

// Structure merger: schema inference stage
export const structuremerger = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

// Constraint collector for schema inference
export async function constraintcollector(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

// Schema emitter: schema inference finalization
export function schemaemitter(state: State): Result {
  validate(state);
  return finalize(state);
}

Project overview

This project builds a schema inference system whose reason for existing is deriving a JSON Schema from a sample of documents so untyped data can be validated going forward. The temptation in schema inference is to bolt the interesting part onto a working prototype and discover the hard cases in production; this build inverts that order and treats a sample that is unrepresentative and produces a schema that rejects valid documents in production as the first thing designed, not the last.

You will implement Sample reader, Type inferer, Structure merger, Constraint collector, and Schema emitter, each as a module with one responsibility and a test that proves it in isolation. The schema inference pipeline is then the wiring that composes them, and the integration test is where the composition is shown to survive a sample that is unrepresentative and produces a schema that rejects valid documents in production.

The walkthrough is written for a developer who has used a schema inference system but has not had to make one honest. It favors explicit data shapes over clever abstractions, and it names the failure each design choice prevents rather than presenting the choice as a style preference.

Each phase ends with something observable — a sample reader that returns the right shape, a type inferer that rejects bad input, an integration test that goes green — so you can stop at any point and still have a schema inference system that works for the cases covered so far.

What you will build

The finished schema inference system exposes five capabilities, each with a focused test:

• Sample reader
• Type inferer
• Structure merger
• Constraint collector
• Schema emitter

Sample reader is the foundation that every later stage assumes; Type inferer is where schema inference correctness lives; Structure merger handles a sample that is unrepresentative and produces a schema that rejects valid documents in production; Constraint collector makes the system observable; and Schema emitter is the graceful-degradation layer that keeps overload from becoming corruption.

The build keeps the schema inference logic in ordinary modules you can read in one sitting, and pulls in dependencies only for the parts that are genuinely tedious to hand-roll. That keeps the diff between the tutorial and a real codebase small, and keeps the failure modes visible in code rather than hidden behind a framework.

Architecture and data model

The system is organized around Sample reader, Type inferer, and Structure merger, with Constraint collector tying them together. Data flows as a merged type tree inferred from many samples, with constraints like required keys and string formats added where evidence supports them.

Each component is a module with a pure interface where practical: the same input produces the same output, and side effects are pushed to the edges. That discipline is what makes the phases independently testable, because a test for Structure merger does not need to stand up Sample reader and Type inferer to exercise the interesting branch.

The data model is deliberately explicit. A merged type tree inferred from many samples, with constraints like required keys and string formats added where evidence supports them Keeping the shape written down — in a schema, a type, or a fixture — means a change to the shape is a visible diff rather than a silent runtime shift, and it means the integration test can assert against a concrete shape instead of guessing.

Configuration and code are kept apart: the configuration says what schema inference should do, the modules say how. A reviewer can change a threshold or a policy without reading the implementation, which keeps the risky changes small and lets the diff stand as the review.

The failure path through Structure merger is designed before the happy path is finished, because a sample that is unrepresentative and produces a schema that rejects valid documents in production is the case that distinguishes a schema inference system from a demo. Every component declares what it does when its inputs are wrong, and the integration test asserts each of those declarations.

Prerequisites and dependencies

Have these ready before you install anything, so the schema inference setup is not blocked by a missing tool:

• Node.js 20 LTS or later
• Familiarity with schema inference fundamentals
• A local environment to run the test suite

The runtime dependencies are chosen to keep the schema inference datapath legible:

• commander (^12.1.0) — CLI framework
• ajv (^8.17.0) — Schema validation

# setup for json-schema-infer
pnpm init
pnpm add commander ajv
pnpm add -D typescript tsx vitest

Versions are pinned to a major range so the schema inference install is reproducible; commit the lockfile. The sample reader reads its operational parameters from the environment and validates them at startup, so a missing value fails the boot rather than the first schema inference request.

Directory structure

Keep each schema inference component in its own file so it can be tested without starting the others:

json_schema_infer/
├── src/index.ts
├── src/sample-reader.ts
├── src/type-inferer.ts
├── tests/json_schema_infer.test.ts
└── package.json

Phase 1: Sample reader

Build the sample reader that anchors schema inference and defines the core data shape.

// Sample reader: schema inference core
export function samplereader(input: SamplereaderInput): SamplereaderResult {
  // Build the sample reader that anchors schema inference and defines the core data shape.
  return processSamplereader(input);
}

After this phase the sample reader works in isolation and is covered by a unit test. The sample reader defines the data shape that every later stage assumes, so getting it right first means the later phases compile against a stable contract.

Treat the sample reader as the single source of truth for the schema inference data shape. Validate inputs at the boundary and reject early, so a malformed value never reaches Type inferer or Structure merger where its effect would be harder to localize.

Phase 2: Type inferer

Add the type inferer so schema inference can deriving a JSON Schema correctly.

// Type inferer — handles schema inference Add the type inferer so
export class Typeinferer {
  constructor(private opts: TypeinfererOptions) {}
  async run(ctx: Context): Promise<TypeinfererOutput> {
    return this.execute(ctx);
  }
}

After this phase the type inferer works in isolation and is covered by a unit test. This is where schema inference correctness lives: the type inferer is what makes deriving a JSON Schema from a sample of documents so untyped data can be validated going forward true, and its test is the proof.

Keep the type inferer pure where you can. A function that takes an input and returns an output is trivial to test and impossible to mock wrong; a class with hidden state is the opposite. Push side effects to the caller and keep the type inferer about the transformation.

Phase 3: Structure merger

Wire the structure merger to handle a sample that is without losing data.

// Structure merger: schema inference stage
export const structuremerger = (next: Next) => async (req: Request) => {
  const out = await transform(req);
  return next(out);
};

After this phase the structure merger works in isolation and is covered by a unit test. The structure merger is the difference between a schema inference system that works on the happy path and one that survives a sample that is unrepresentative and produces a schema that rejects valid documents in production.

Design the failure before the success. Decide what structure merger does when a sample that is unrepresentative and produces a schema that rejects valid documents in production: reject, retry, shed, or degrade — and write that test first. A schema inference system whose failure behavior is unspecified will pick one at random under pressure, and it is usually the wrong one.

Phase 4: Constraint collector

Integrate the constraint collector to make schema inference observable and testable end to end.

// Constraint collector for schema inference
export async function constraintcollector(stream: AsyncIterable<Chunk>): Promise<Aggregate> {
  let acc = init();
  for await (const c of stream) acc = merge(acc, c);
  return acc;
}

After this phase the constraint collector works in isolation and is covered by a unit test. Observability is not a finishing touch; it is how you know the schema inference system is doing what it claims in production.

Emit one structured event per meaningful action, with enough context to localize a failure without a second round-trip. The constraint collector is what turns “something is slow” into “the type inferer is slow for inputs matching this shape,” which is the difference between an alert and an investigation.

Phase 5: Schema emitter

Finalize the schema emitter so the schema inference system degrades gracefully under overload.

// Schema emitter: schema inference finalization
export function schemaemitter(state: State): Result {
  validate(state);
  return finalize(state);
}

After this phase the schema emitter works in isolation and is covered by a unit test. The schema emitter is what keeps overload from becoming corruption: when the schema inference system is asked for more than it can do, it degrades in a defined way instead of failing in an undefined one.

Pick a degradation strategy for schema emitter — shed, queue, or backpressure — and make it explicit in code and in a test. Leaving the behavior to chance is how a schema inference traffic spike becomes a data loss incident.

Testing

Each phase has a focused test: sample reader, type inferer, and structure merger are exercised in isolation, then an integration test drives the full schema inference pipeline and asserts the a sample that is case is handled.

Start with unit tests that pin each component’s contract: the sample reader returns the documented shape, the type inferer transforms correctly, and the structure merger fails in the documented way. These tests are fast and they localize regressions to the component that broke.

Then add an integration test that composes the components and drives a realistic schema inference scenario end to end. Assert the happy path first, then deliberately trigger a sample that is unrepresentative and produces a schema that rejects valid documents in production and assert the system degrades rather than corrupts. Because the components are pure where practical, the integration test stays deterministic rather than flaky.

A schema inference suite must also prove the negative: bad input is rejected at the boundary, overload is shed rather than queued indefinitely, and a partial failure does not leave the system in an inconsistent state. These are the tests that catch the regressions a happy-path suite never would.

The entire schema inference suite runs with no external service; every dependency is either in-process or stubbed, so every case is reproducible in CI. That is what lets the structure merger failure-mode assertions live as tests rather than as a manual checklist that drifts out of date.

# run the schema inference test suite
pnpm vitest run

Deployment and operations

Run at least two schema inference instances behind a load balancer and share state in Redis. Set resource limits so schema inference overload surfaces as backpressure rather than failure.

Run at least two stateless instances behind a load balancer, and move any shared state into a shared store so the schema inference behavior is global rather than per instance. Size limits so the system sheds load before it runs out of memory, because a schema inference system that OOMs under load fails in a way that is hard to reason about.

Drain on deploy: stop accepting new work, let in-flight schema inference operations finish, then exit. Because the system holds no state that cannot be reconstructed, a rollout is a matter of cycling instances one at a time while watching the error rate.

When a dependency is unavailable, decide explicitly whether the schema inference system fails closed or degrades. Leaving that choice to chance is how a dependency outage becomes a silent data loss, so make the choice visible in code and in a test.

Security considerations

Validate all schema inference inputs at the boundary, reject oversized payloads before buffering them, and keep secrets in the environment rather than the repository.

The schema inference system is a choke point, so it is the right place to reject malformed input before it reaches type inferer or structure merger. Validate the shape and the size at the edge, and let the inner components assume what they receive is well-formed.

Never log secrets or full payloads; log enough to localize a failure without recording data that should not be on disk. Treat any external call as untrusted and validate its response before using it, because a schema inference system that trusts its inputs blindly is one malformed response away from corruption.

Keep dependencies pinned and run an audit in CI. A transitive dependency with a known vulnerability is still a vulnerability in your schema inference system, and the lockfile is what makes the audit meaningful.

Observability

Emit one structured event per schema inference operation with the inputs’ shape, the outcome, and a correlation id. Track error rate and latency per component and alert on a step change rather than a single spike.

The correlation id is the thread that ties a user-visible symptom to the schema inference decision and to the component log entry for the same operation. Without it, an investigation starts from a timestamp and a guess.

Watch the structure merger failure rate separately from the overall error rate. A jump in structure merger failures usually means a sample that is unrepresentative and produces a schema that rejects valid documents in production, not a bug in type inferer, and the two have different owners and different fixes.

Chart the share of schema inference operations that shed, queue, or succeed. A change in the mix points at the cause faster than an aggregate latency can, because each class has a different root.

Troubleshooting

A sample that is unrepresentative and produces a schema that rejects valid documents in production. Reproduce with a minimal fixture, then add a guard test so the schema inference system rejects the condition loudly.

Performance regresses under schema inference load. Profile with the sampling analyzer, locate the hot path in type inferer, and add a targeted benchmark to lock the regression.

Keep the schema inference dashboards separate from dependency dashboards. When an operation fails, the correlation id lets you decide in seconds whether the fault is sample reader, type inferer, or structure merger, instead of guessing from a timeout.

Limitations and next steps

• Single-region; multi-region replication is out of scope.
• The sample reader is in-process and not horizontally shared by default.
• No auth layer; assume deployment behind an edge that enforces identity.

The schema inference system is single-region by design; extending to multi-region means making sample reader shareable across regions, which is a real change to the consistency model rather than a config flag. The structure merger is in-process, which is correct for one instance but must move to a shared store before scaling out.

There is no authentication or authorization layer; the system assumes it is deployed behind an edge that enforces identity. Adding one is straightforward but out of scope for this build, and doing it here would distract from the schema inference correctness that is the point.

The natural next step, once the schema inference pipeline is trusted in production, is to add a short-TTL cache for idempotent reads and a richer constraint collector that exports histograms rather than counters. Both are additive and neither changes the contracts this build establishes.