schema-to-yup

Schema to Yup schema

Build a Yup schema from a JSON Schema, GraphQL schema (type definition) or any other similar type/class and field/properties model or schema :)

Update

Release 1.9.10 and later of schema-to-yup fixes a bunch of issues including validation of complex nested schemas and proper handling of required vs notRequired including passing a mode configuration for fine control. It also fixes certain related error message issues for certain constraints (such as required).

The library now also includes a typings file so it should work nicely with TypeScript.

In the near future (2.0.0 release) we will include a separate constraint builder infrastructure and more flexible/configurable infastructure in general. This will decouple the JSON to schema to validation type rules from the instantiation of each type validator, making it trivial to support other validators and maintain/develop each separately.

Schemas

• AJV: JSON Schema keywords
• Learn JsonSchema

The builder currently supports the most commonly used JSON Schema layout and GraphQL type definition exports using graphSchemaToJson (see GraphQL schema).

It also supports some extra convenience schema properties that make it more "smooth" to define validation requirements declaratively (see below).

According to the JSON schema specs, you are free to add extra metadata to the field schema definitions beyond those supported "natively".

Quick start

Install

npm install schema-to-yup -S or yarn add schema-to-yup

Use

const schema = {
  $schema: "http://json-schema.org/draft-07/schema#",
  $id: "http://example.com/person.schema.json",
  title: "Person",
  description: "A person",
  type: "object",
  properties: {
    name: {
      description: "Name of the person",
      type: "string"
    },
    email: {
      type: "string",
      format: "email"
    },
    fooorbar: {
      type: "string",
      matches: "(foo|bar)"
    },
    age: {
      description: "Age of person",
      type: "number",
      exclusiveMinimum: 0,
      required: true
    },
    characterType: {
      enum: ["good", "bad"],
      enum_titles: ["Good", "Bad"],
      type: "string",
      title: "Type of people",
      propertyOrder: 3
    }
  },
  required: ["name", "email"]
};

const config = {
  // for error messages...
  errMessages: {
    age: {
      required: "A person must have an age"
    },
    email: {
      required: "You must enter an email address",
      format: "Not a valid email address"
    }
  }
};

const { buildYup } = require("schema-to-yup");
const yupSchema = buildYup(json, config);
// console.dir(schema)
const valid = await yupSchema.isValid({
  name: "jimmy",
  age: 24
});

console.log({
  valid
});
// => {valid: true}

This would generate the following Yup validation schema:

const schema = yup.object().shape({
  name: yup.string().required(),
  age: yup
    .number()
    .required()
    .positive()
});

Note the "required": true for the age property (not natively supported by JSON schema).

Mode

By default, any property will be explicitly notRequired unless set to be required, either via required: true in the property constraint object or via the required list of properties of the object schema definition (of the property).

You can override the notRequired behavior by setting it on the new mode object of the configuration which can be used to control and fine-tune runtime behaviour.

const jsonSchema = {
  title: "users",
  type: "object",
  properties: {
    username: { type: "string" }
  }
};

const yupSchema = buildYup(jsonSchema, {
  mode: {
    notRequired: true // default setting
  },
});

// will be valid since username is not required by default
const valid = yupSchema.validateSync({
  foo: "dfds"
});

const yupSchema = buildYup(jsonSchema, {
  mode: {
    notRequired: true // default setting
  },
});
// will be invalid since username is required by default when notRequired mode is disabled
const valid = yupSchema.validateSync({
  foo: "dfds"
});

The new run time mode settings are demonstrated under sample-runs/mode

No validation error (prop not required unless explicitly specified):

$ npx babel-node sample-runs/modes/not-required-on.js

Validation error if not valid type:

$ npx babel-node sample-runs/modes/not-required-on.js

Shape

You can access the internal Yup shape, via shapeConfig on the yup schema returned by the buildYup schema builder function. This allows you to easily mix and match to suit more advanced requirements.

const { buildYup } = require("json-schema-to-yup");
const { shapeConfig } = buildYup(json, config);
const schema = yup.object().shape({
  ...shapeConfig,
  ...customShapeConfig
})

Types

Currently the following schema types are supported:

• array
• boolean
• date
• number
• object
• string

Custom type handlers

You can pass any custom typehandlers in a typeHandlers object as part of the config object passes. See setTypeHandlers() in entry.js for how this works internally.

function myCustomStringHandler = (obj, config) => {
  // ... custom handler
  // return yup type schema such as yup.string()
  // with one or more constraints added
}

const yupSchema = buildYup(jsonSchema, {
  typeHandlers: {
    string: myCustomStringHandler
  },
});

This can be used to support special cases, to circumvent a bug or unexpected/unwarranted behaviour for one or more of the built-in type handlers etc. You can then use the built in classes as building blocks.

Release 2.0 will likely include a much improved infrastructure of suitable building blocks to easily customize behavior and extend the functionality etc.

Custom constraint builder

Version 1.9.13 and higher supports using a custom constraint builder to add and build constraints.

class MyConstraintBuilder extends ConstraintBuilder {
  constructor(typeHandler, opts = {}) {
    super(typeHandler, opts);
    // custom instance configuration
  }

  build(propName, opts) {
    /// custom build logic

    // returns new type validation handler (base) with built constraint added
    return newBase
  }

  addConstraint(propName, opts) {
    // custom add constraint logic
    return this.typeHandler
  }

  // custom event handler
  onConstraintAdded({ name, value }) {
    // ...
    return this.typeHandler;
  }
}

const createConstraintBuilder = (typeHandler, opts) => {
  // custom builder logic
  return new MyConstraintBuilder(typeHandler, opts)
}

buildYup(jsonSchema, {
  createConstraintBuilder
})

Mixed (any type)

• strict
• default
• nullable
• required
• notRequired
• oneOf (enum, anyOf)
• notOneOf
• when NEW
• isType NEW
• nullable (isNullable) NEW

Array

• ensure
• compact
• items (of)
• maxItems (max)
• minItems (min)
• itemsOf (of) NEW

Boolean

No keys

Date

• maxDate (max)
• minDate (min)

Number

• integer
• moreThan (exclusiveMinimum)
• lessThan (exclusiveMaximum)
• positive
• negative
• min (minimum)
• max (maximum)
• truncate
• round

Object

• camelCase
• constantCase
• noUnknown (propertyNames)

String

• minLength (min)
• maxLength (max)
• pattern (matches or regex)
• email (format: 'email')
• url (format: 'url')
• lowercase
• uppercase
• trim

Conditional logic

Basic support for when conditions as requested and outlined in this issue is now included.

Work will continue in the when-condition branch.

Sample schema using simple when constraint:

const biggyJson = {
  title: "biggy",
  type: "object",
  properties: {
    isBig: {
      type: "boolean"
    },
    count: {
      type: "number",
      when: {
        isBig: {
          is: true,
          then: {
            min: 5
          }
        }
      }
    }
  }
};

Sample valid and invalid values with respect to biggyJson schema

const bigJson = {
  valid: {
    isBig: true,
    count: 5 // since isBig is set, must be >= 5
  },
  invalid: {
    isBig: true,
    count: 4 // since isBig is set, must be >= 5
  }
};

Currently basic support is included in schema-to-yup@1.9.0 on npmjs

More advanced conditionals support will likely be included the next major release: 2.0.

You are welcome to continue the effort to support more conditional schema logic by continuing on this branch and making PRs.

Support for if then and else conditional JSON schema constraints can likely be added using an approach like the when condition (perhaps by transalating to equivalent: when, then and otherwise).

• if - This keyword's value MUST be a valid JSON Schema
• then - This keyword's value MUST be a valid JSON Schema
• else - This keyword's value MUST be a valid JSON Schema

See also json-schema-spec

Customizing conditional logic

You can now also override, extend or customize the when condition logic by passing in your own factory method for the config object entry createWhenCondition

const myWhenConditionFactoryFn = (opts = {}) => {
  const { type, key, value, when, schema, properties, config } = opts;
  // ...
}

const config = {
  createWhenCondition: myWhenConditionFactoryFn
}
const schema = buildYup(nameJsonSchema, config)

The best and easiest way to do this is to extend the WhenCondition class which contains most of the necessary infrastructure you can further build on.

See the src/conditions/legacy folder for the legacy 1.9.0 logic that works but has limited functionality.

Complex example

Here a more complete example of the variations currently possible

{
  "title": "Person",
  "description": "A person",
  "type": "object",
  "properties": {
    "name": {
      "description": "Name of the person",
      "type": "string",
      "required": true,
      "matches": "[a-zA-Z- ]+",
      "mix": 3,
      "maxLength": 40,
    },
    "age": {
      "description": "Age of person",
      "type": "integer",
      "moreThan": 0,
      "max": 130,
      "default": 32,
      "required": false,
      "nullable": true
    },
    "birthday": {
      "type": "date",
      "min": "1-1-1900",
      "maxDate": "1-1-2015"
    },
    "married": {
      "type": "boolean"
    },
    "boss": {
      "type": "object",
      "noUnknown": [
        "name"
      ],
      "properties": {
        "name": {
          "type": "string",
          "notOneOf": ["Dr. evil", "bad ass"]
        }
      }
    },
    "colleagues": {
      "type": "array",
      "items": {
        "type": "object",
        "propertyNames": [
          "name"
        ],
        "properties": {
          "name": {
            "type": "string"
          }
        }
      }
    },
    "programming": {
        "type": "object",
        "properties": {
          "languages": {
            "type": "array",
            "of": {
              "type": "string",
              "enum": ["javascript", "java", "C#"]
            },
            "min": 1,
            "max": 3
          }
        }
      }
    }
  }
}

Complex/Nested schemas

Nested object schema properties are now finally fully supported.

See test/types/object/complex-schema.test.js

Custom models

This library now also supports non JSON schema models. See the types/defaults mappings.

types/defaults/json-schema.js

module.exports {
  getProps: obj => obj.properties,
  getType: obj => obj.type,
  getName: obj => obj.name || obj.title,
  getConstraints: obj => obj,
  isString: obj => obj.type === "string",
  isArray: obj => obj.type === "array",
  isInteger: obj => obj.type === "integer",
  isBoolean: obj => obj.type === "boolean",
  hasDateFormat: obj => ["date", "date-time"].find(t => t === obj.format),
  isDate: obj => obj.type === "string" && defaults.hasDateFormat(obj.format),
  isNumber: obj => obj.type === "number" || defaults.isInteger(obj.type),
  isObject: obj => obj.type === "object",
  isRequired: obj => obj.required
};

GraphQL schema

To support another model, such as GraphQL schema (type definitions) via graphSchemaToJson

Person:

{
  Person: {
    name: 'Person',
    fields: {
      name: {
        type: 'String',
        directives: {
          constraints: {
            minLength: 2
          }
        },
        isNullable: false,
        isList: false
      }
    }
    directives: {},
    type: 'Object',
    implements: []
  }
}

Create a map of methods to match your model layout:

const typeDefConf = {
  getProps: obj => obj.fields,
  getType: obj => obj.type,
  getName: obj => obj.name,
  getConstraints: obj => (obj.directives || {}).constraints || {},
  isString: obj => obj.type === "String",
  isArray: obj => obj.isList,
  isInteger: obj => obj.type === "Int",
  isBoolean: obj => obj.type === "Boolean",
  isDate: obj => obj.type === "Date" || obj.directives.date,
  isNumber: obj => obj.type === "Int" || obj.type === "Float",
  isObject: obj => obj.type === "Object",
  isRequired: obj => !obj.isNullable
};

Please note that getConstraints can be used to control where the constraints of the field will be taken from (depending on the type of model/schema or your preference).

Pass overrides to match your model in config as follows:

const schema = buildYup(nameJsonSchema, { ...typeDefConf, log: true });

The type definition mappings above are already built-in and available. To switch the schema type, pass schemaType in config as follows.

const schema = buildYup(nameJsonSchema, { schemaType: "type-def", log: true });

Feel free to make PRs to make more common schema models conveniently available!

Custom logs and error handling

You can enable logging py passing a log option in the config argument. If set to true, it will by default assign the internal log function to console.log

const schema = buildYup(nameJsonSchema, { log: true });

You can also pass a log function in the log option to handle log messages and an err option with a custom error handler function.

See Custom errors in Node for how to design custom errors

class ValidationError extends Error {}

const schema = buildYup(nameJsonSchema, {
  log: (name, msg) => console.log(`[${name}] ${msg}`)
  err: (msg) => {
    console.error(`[${name}] ERROR: ${msg}`
    throw new ValidationError(msg)
  })
});

Customization

You can supply a createYupSchemaEntry function as an entry in the config object. This function will then be used to build each Yup Schema entry in the Yup Schema being built.

Use the Yup Type classes such as types.YupArray to act as building blocks or create your own custom logic as you see fit.

Customization example

const { YupSchemaEntry, buildYup, types } = require("schema-to-yup");

class CustomYupArray extends types.YupArray {
  // ...
}

class CustomYupSchemaEntry extends YupSchemaEntry {
  // ...
}

function createYupSchemaEntry(key, value, config) {
  const builder = new CustomYupSchemaEntryBuilder(key, value, config);
  builder.types.array = config => createYupArray(config);
  return builder.toEntry();
}

// use some localized error messages
const messages = i18n.locale(LOCALE);

const yupSchema = buildYup(json, {
  createYupSchemaEntry,
  messages
});

Extend Yup API to bridge other validators

You can use extendYupApi to extend the Yup API with extra validation methods:

const validator = require("validator");
const { extendYupApi } = require("schema-to-yup/validator-bridge");

// by default extends with string format validation methods of validator
// See https://www.npmjs.com/package/validator
extendYupApi({ validator });

You can optionally pass in a custom validator and a constraints map of your choice. You can either extend the default constraints or override them with your own map.

PS: Check out src/validator-bridge for more many options for fine control

const myValidator = new MyValidator();
const constraints = ["creditCard", "currency", { name: "hash", opts: "algo" }];
extendYupApi({ validator: myValidator, override: true, constraints });

const { buildYup } = require("schema-to-yup");
// type def sample schema, using credit-card format validator
const schema = {
  name: "BankAccount",
  fields: {
    accountNumber: {
      type: "String",
      format: "credit-card"
    }
  }
};

// opt in to use generic string format validation, via format: true config option
const yupSchema = buildYup(schema, { format: true, schemaType: "type-def" });
// ...do your validation
const valid = await yupSchema.isValid({
  accountNumber: "123-4567-1828-2929"
});

Now the bridge includes tests. Seems to work ;)

Subclassing

You can sublass YupBuilder or any of the internal classes to create your own custom infrastructure to suit your particular needs, expand with support for extra features etc.

const { YupBuilder } = require("schema-to-yup");

class MyYupBuilder extends YupBuilder {
  // ... custom overrides etc
}

const builder = new MyYupBuilder(schema, config);
const { yupSchema } = builder;
// ...

Error messages

You can pass an errMessages object in the optional config object argument with key mappings for your custom validation error messages.

Internally the validator error messages are resolved with the instance method valErrMessage (from Mixed class)

  notOneOf() {
    const {not, notOneOf} = this.value
    const $oneOf = notOneOf || (not && (not.enum || not.oneOf))
    $oneOf && this
      .base
      .notOneOf($oneOf, this.valErrMessage('notOneOf'))
    return this
  }

The key entries can be either a function, taking a value argument or a static string. Here are some of the defaults that you can override as needed.

export const errValKeys = [
  "oneOf",
  "enum",
  "required",
  "notRequired",
  "minDate",
  "min",
  "maxDate",
  "max",
  "trim",
  "lowercase",
  "uppercase",
  "email",
  "url",
  "minLength",
  "maxLength",
  "pattern",
  "matches",
  "regex",
  "integer",
  "positive",
  "minimum",
  "maximum"
];

export const defaults = {
  errMessages: (keys = errValKeys) =>
    keys.reduce((acc, key) => {
      const fn = ({ key, value }) =>
        `${key}: invalid for ${value.name || value.title}`;
      acc[key] = fn;
      return acc;
    }, {})
};

Custom validation messages using select defaults

const { buildYup, types } = require("schema-to-yup");
const { defaults } = types;

const myErrMessages = require("./err-messages");
const valKeys = ["lowercase", "integer"];

// by default Yup built-in validation error messages will be used if not overridden here
const errMessages = {
  ...defaults.errMessages(valKeys),
  myErrMessages
};

const yupSchema = buildYup(json, {
  errMessages
});

Adding custom constraints

See the number type for the current best practice to add type constraints.

For simple cases: use addConstraint from the superclass YupMixed

  required() {
    return this.addConstraint("required");
  }

For types with several of these, we should map through a list or map to add them all.

  strict() {
    return this.addValueConstraint("strict");
  }

  required() {
    return this.addConstraint("required");
  }

  notRequired() {
    return this.addConstraint("notRequired");
  }

Can be rewritten to use conventions, iterating a map:

  addMappedConstraints() {
    Object.keys(this.constraintsMap).map(key => {
      const list = constraintsMap[key];
      const fnName = key === 'value' ? 'addValueConstraint' : 'addConstraint'
      list.map(this.[fnName]);
    });
  }

  get constraintsMap() {
    return {
      simple: ["required", "notRequired", "nullable"],
      value: ["default", "strict"]
    };
  }

For more complex contraints that:

• have multiple valid constraint names
• require validation
• optional transformation

You can create a separate Constraint subclass, to offload and handle it all separately. Here is a sample RangeConstraint used by number.

class RangeConstraint extends NumericConstraint {
  constructor(typer) {
    super(typer);
  }

  get $map() {
    return {
      moreThan: ["exclusiveMinimum", "moreThan"],
      lessThan: ["exclusiveMaximum", "lessThan"],
      max: ["maximum", "max"],
      min: ["minimum", "min"]
    };
  }
}

Instead of wrapping a Constraint you can call it directly with a map

// this would be an instance such as YupNumber
// map equivalent to $map in the RangeConstraint
range() {
  return createNumericConstraint(this, map);
}

For the core type constraint class (such as YupNumber) you should now be able to simplify it to:

  get enabled() {
    return ["range", "posNeg", "integer"];
  }

  convert() {
    this.enabled.map(name => this.processConstraint(name));
    super.convert();
    return this;
  }

The following constraint classes are available for use:

• NumericConstraint
• StringConstraint
• RegExpConstraint
• DateConstraint

Currently only YupNumber has been (partly) refactored to take advantage of this new infrastructure. Please help refactor the rest!

YupNumber also has the most unit test coverage, used to test the current infrastructure!

Similar projects

• JSON schema to Elastic Search mapping
• JSON Schema to GraphQL types with decorators/directives
• JSON Schema to Mongoose schema
• JSON Schema to MobX State Tree types
• Convert JSON schema to mongoose 5 schema

The library JSON Schema model builder is a powerful toolset to build a framework to create any kind of output model from a JSON schema.

If you enjoy this declarative/generator approach, try it out!

Testing

Uses jest for unit testing.

• Have unit tests that cover most of the constraints supported.
• Could use some refactoring using the latest infrastructure (see NumericConstraint)
• Please help add more test coverage and help refactor to make this lib even more awesome :)

Development

Current development is taking place on refactoring branch.

On his branch:

• all the code has been converted to TypeScript
• constraint classes for String, Numeric, RegExp etc.
• Validator building has been extracted so you can add support for any Validator, such as Forg
• more...

If you would like to further improved this library or add support for more validators than Yup, please help on this branch. Cheers!

Refactoring

On the master branch we've added a ConstraintBuilder class which should contain all the logic to build the constraints from a set of options in the future. Currently this logic is on the YupMixed class where it really doesn't belong. YupMixed or any other Yup type class should delegate the building of a Yup constraint to a dedicated class.

Ideas and suggestions

Please feel free to come with ideas and suggestions on how to further improve this library.

Author

2018 Kristian Mandrup (CTO@Tecla5)

License

MIT