validator: introduce experimental node:validator module

[sheplu]

Note: The purpose of this PR is not to reflect what a fully implemented validator would be able to handle. It is intended to open a discussion on whether it makes sense to continue in this direction before investing further in the surface area, feature set, and edge cases.

Introduces a new experimental built-in module node:validator - a synchronous, JSON-Schema-inspired input validator for simple REST-style payload checks.
Gated behind --no-experimental-validator (on by default, opt-out) and emits an ExperimentalWarning on first require.

The purpose of this PR is not reflecting what a fully implemented validator would be able to handle, but to have a discussion to know if this make sense to continue in this line.

Key Features

• Schema class: validate and normalize a schema once, reuse for many validations.
• Supported types: string, number, integer, boolean, object, array, null.
• Constraints:
  • minLength / maxLength / pattern / enum for strings,
  • minimum / maximum / exclusiveMinimum / exclusiveMaximum / multipleOf for numbers,
  • minItems / maxItems / items for arrays,
  • properties / required / additionalProperties for objects.
• Non-throwing validation: schema.validate(data) returns a frozen { valid, errors } result; never throws.
• Machine-readable error codes: 13 codes exposed via the codes export (e.g. INVALID_TYPE, STRING_TOO_SHORT, PATTERN_MISMATCH).
• Defaults: schema.applyDefaults(data) returns a new object with declared defaults applied (input is not mutated).
• Schema composition: schema.toJSON() returns a frozen copy of the definition so schemas can be embedded in other schemas.
• Compile-time validation: invalid schema definitions throw ERR_VALIDATOR_INVALID_SCHEMA at new Schema(...), not at validate time.

Example

Example — validate a user payload

  import { Schema } from 'node:validator';

  const userSchema = new Schema({
    type: 'object',
    required: ['name', 'email'],
    properties: {
      name: { type: 'string', minLength: 1, maxLength: 100 },
      email: { type: 'string', pattern: '^[^@]+@[^@]+$' },
      age: { type: 'integer', minimum: 0, maximum: 150 },
      tags: { type: 'array', items: { type: 'string' }, maxItems: 10 },
    },
  });

  const result = userSchema.validate({
    name: 'Alice',
    email: '[email protected]',
    age: 30,
    tags: ['admin'],
  });

  console.log(result.valid);  // true
  console.log(result.errors); // []

Example — apply defaults

import { Schema } from 'node:validator';

const schema = new Schema({
  type: 'object',
  properties: {
    host: { type: 'string', default: 'localhost' },
    port: { type: 'integer', default: 3000 },
  },
});

const config = schema.applyDefaults({});
console.log(config.host); // 'localhost'
console.log(config.port); // 3000

A bit of the chicken/egg with the linter, so I also opened nodejs/core-validate-commit#147 and for now using the node(validator) commit message

[nodejs-github-bot]

Review requested:

• @nodejs/config
• @nodejs/loaders
• @nodejs/startup

[ljharb]

Unfortunately JSON Schema doesn't support many of the things that JS apps need to validate - functions, regexes, for some examples, so the generic name "validator" doesn't seem like a good fit if it'd only cover JSON-serializable data.

[codecov]

Codecov Report

❌ Patch coverage is 99.35980% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.67%. Comparing base (2428030) to head (6766f00).
⚠️ Report is 44 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/internal/validator/compile.js	98.63%	4 Missing ⚠️
lib/internal/validator/schema.js	99.07%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #62927      +/-   ##
==========================================
+ Coverage   89.63%   89.67%   +0.03%     
==========================================
  Files         706      712       +6     
  Lines      219219   220181     +962     
  Branches    42004    42259     +255     
==========================================
+ Hits       196499   197438     +939     
- Misses      14622    14637      +15     
- Partials     8098     8106       +8     

Files with missing lines	Coverage Δ
lib/internal/bootstrap/realm.js	96.22% <100.00%> (+<0.01%)	⬆️
lib/internal/errors.js	97.65% <100.00%> (+<0.01%)	⬆️
lib/internal/process/pre_execution.js	98.28% <100.00%> (-0.11%)	⬇️
lib/internal/validator/defaults.js	100.00% <100.00%> (ø)
lib/internal/validator/errors.js	100.00% <100.00%> (ø)
lib/internal/validator/validate.js	100.00% <100.00%> (ø)
lib/validator.js	100.00% <100.00%> (ø)
src/node_options.cc	76.53% <100.00%> (-0.08%)	⬇️
src/node_options.h	98.00% <100.00%> (+0.01%)	⬆️
lib/internal/validator/schema.js	99.07% <99.07%> (ø)
... and 1 more

... and 48 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Qard]

A bit of overlap with the ata effort in #62603. Might want to just expose that instead given the measured perf.

[mertcanaltin]

Hey @sheplu, good to see this come up. We've been on the same track.

Context: #62603 is vendoring ata into core for node.config.json right now (two approvals, @lemire as co-author) Once that lands, exposing it to userland is the obvious next thing, and @TheOneTheOnlyJJ and @H4ad both asked for it during that review.

On @ljharb's naming point, agreed, validator reads a bit broad. A few shapes on the table: node:schema, node:jsonschema, or a wider node:json covering parse and validate (ata already vendors simdjson).

Going to open a tracking issue shortly and drop the link here. Syncing with @H4ad later today, happy to pull you in if you want to compare notes.

Appreciate you kicking this off. Rather keep these in sync than run two efforts in parallel.

[sheplu]

Seems like I was bad at searching for something similar! Thanks @Qard

@mertcanaltin definitely, feel free to include me! More that happy to help work on that. And yes the naming was broad :D

[mertcanaltin]

:D The issue being followed is currently open: #62598, I'm awaiting your comments.

[Qard]

@sheplu No worries. I've just been reviewing it so had it already in my mind when I saw this. I actually assumed at first that this was just going to be the work to expose that, since I knew it had not been done yet. 😅

[Delapouite]

I feel like asking input from people who explored this topic along the years would be valuable? Friendly ping to @colinhacks from zod and @ssalbdivad from ArkType who may have a few ideas to share on design of such validation API in Node.

[mcollina]

I'm not convinced this belongs to core. The ecosystem is having constant innovation on this area in the last few years, and solidifying something in core like this would hamper it.