Garden Schema

The Garden Schema is a simple data validation and cleaning library based on OpenAPI 3.0 Schema.

Features

• Define the data structures of PHP arrays of any depth, and validate them.

• Validated data is cleaned and coerced into appropriate types.

• The schema defines a whitelist of allowed data and strips out all extraneous data.

• The Schema class understands a subset of data in OpenAPI Schema format. We will add more support for the built-in JSON schema validation as time goes on.

• Developers can use a shorter schema format in order to define schemas in code rapidly. We built this class to be as easy to use as possible. Avoid developer groans as they lock down their data.

• Add custom validator callbacks to support practically any validation scenario.

• Override the validation class in order to customize the way errors are displayed for your own application.

Uses

Garden Schema is meant to be a generic wrapper for data validation. It should be valuable when you want to bullet-proof your code against user-submitted data. Here are some example uses:

• Check the data being submitted to your API endpoints. Define the schema at the beginning of your endpoint and validate the data before doing anything else. In this way you can be sure that you are using clean data and avoid a bunch of spaghetti checks later in your code. This was the original reason why we developed the Garden Schema.

• Clean user input. The Schema object will cast data to appropriate types and gracefully handle common use-cases (ex. converting the string "true" to true for booleans). This allows you to use more "===" checks in your code which helps avoid bugs in the longer term.

• Validate data before passing it to the database in order to present human-readable errors rather than cryptic database generated errors.

• Clean output before returning it. A lot of database drivers return data as strings even though it's defined as different types. The Schema will clean the data appropriately which is especially important for consumption by the non-PHP world.

Basic Usage

To validate data you first create an instance of the Schema class and then call its validate() method.

namespace Garden\Schema;

$schema = Schema::parse([...]);
try {
    $valid = $schema->validate($data);
} catch (ValidationException $ex) {
    ...
}

In the above example a Schema object is created with the schema definition passed to its constructor (more on that later). Data to be validated can then be passed to the validate() method. If the data is okay then a clean version is returned, otherwise a ValidationException is thrown.

Defining Schemas

The Schema class is instantiated with an array defining the schema. The array can be in OpenAPI 3.0 Schema format or it can be in custom short format. It is recommended you define your schemas in the OpenAPI format, but the short format is good for those wanting to write quick prototypes. The short format will be described in this section.

By default the schema is an array where each element of the array defines an object property. By "object" we mean javascript object or PHP array with string keys. There are several ways a property can be defined:

[
    '<property>', // basic property, can be any type
    '<property>?', // optional property
    '<property>:<type>?', // optional property with specific type

    '<property>:<type>?' => 'Description', // optional, typed property with description
    '<property>?' => ['type' => '<type'>, 'description' => '...'], // longer format

    '<property>:o' => [ // object property with nested schema
        '<property>:<type>' => '...',
        ...
    ],
    '<property>:a' => '<type>', // array property with element type
    '<property>:a' => [ // array property with object element type
        '<property>:<type>' => '...',
        ...
    ]
]

You can quickly define an object schema by giving just as much information as you need. You can create a schema that is nested as deeply as you want in order to validate very complex data. This short schema is converted into a JSON schema compatible array internally and you can see this array with the jsonSerialize() method.

We provide first-class support for descriptions because we believe in writing readable code right off the bat. If you don't like this you can just leave the descriptions out and they will be left empty in the schema.

Types and Short Types

The Schema class supports the following types. Each type has one or more aliases. You can use an alias for brevity when defining a schema in code and it gets converted to the proper type internally, including when used in errors.

Type	Aliases	Notes
boolean	b, bool
string	s, str, dt	The "dt" alias adds a format of "date-time" and validates to DateTimeInterface instances
integer	i, int, ts	The "ts" alias adds a format of "timestamp" and will convert date strings into integer timestamps on validation.
number	f, float
array	a
object	o

Arrays and Objects

The array and object types are a bit special as they contain several elements rather than a single value. Because of this you can define the type of data that should be in those properties. Here are some examples:

$schema = Schema::parse([
    'items:a', // array of any type
    'tags:a' => 's', // array of strings

    'attributes:o', // object of any type
    'user:o' => [ // an object with specific properties
        'name:s',
        'email:s?'
    ]
]);

Non-Object Schemas

By default, schemas define an object because that is the most common use for a schema. If you want a schema to represent an array or even a basic type you define a single field with no name. The following example defines an array of objects (i.e. the output of a database query).

$schema = Schema::parse([
    ':a' => [
        'id:i',
        'name:s',
        'birthday:dt'
    ]
]);

This schema would apply to something like the following data:

[
    ['id' => 1, 'name' => 'George', 'birthday' => '1732-02-22'],
    ['id' => 16, 'name' => 'Abraham', 'birthday' => '1809-02-12'],
    ['id' => 32, 'name' => 'Franklin', 'birthday' => '1882-01-30']
]

Optional Properties and Nullable Properties

When defining an object schema you can use a "?" to say that the property is optional. This means that the property can be completely omitted during validation. This is not the same a providing a null value for the property which is considered invalid for optional properties.

If you want a property to allow null values you can specify the nullable attribute on the property. There are two ways to do this:

[
    // You can specify nullable as a property attribute.
    'opt1:s?' => ['nullable' => true],

    // You can specify null as an optional type in the declaration.
    'opt2:s|n?' => 'Another nullable, optional property.'
]

Default Values

You can specify a default value with the default attribute. If the value is omitted during validation then the default value will be used. Note that default values are not applied during sparse validation.

Validating Data

Once you have a schema you validate data using the validate() or isValid() methods.

The Schema::validate() method

You pass the data you want to validate to Schema::validate() and it it either returns a cleaned copy of your data or throws a ValidationException.

$schema = Schema::parse(['id:i', 'name:s']);
try {
    // $u1 will be ['id' => 123, 'name' => 'John']
    $u1 = $schema->validate(['id' => '123', 'name' => 'John']);

    // This will thow an exception.
    $u2 = $schema->validate(['id' => 'foo']);
} catch (ValidationException $ex) {
    // $ex->getMessage() will be: 'id is not a valid integer. name is required.'
}

Calling validate() on user-submitted data allows you to check your data early and bail out if it isn't correct. If you just want to check your data without throwing an exception the isValid() method is a convenience method that returns true or false depending on whether or not the data is valid.

$schema = Schema::parse(['page:i', 'count:i?']);

if ($schema->isValid(['page' => 5]) {
    // This will be hit.
}

if ($schema->isValid(['page' => 2, 'count' => 'many']) {
    // This will not be hit because the data isn't valid.
}

The ValidationException and Validation Classes

When you call validate() and validation fails a ValidationException is thrown. This exception contains a property that is a Validation object which contains more information about the fields that have failed.

If you are writing an API, you can json_encode() the ValidationException and it should provide a rich set of data that will help any consumer figure out exactly what they did wrong. You can also use various properties of the Validation property to help render the error output appropriately.

The Validation JSON Format

The Validation object and ValidationException both encode to a specific format. Here is an example:

ValidationError = {
    "message": "string", // Main error message.
    "code": "integer", // HTTP-style status code.
    "errors": { // Specific field errors.
        "<fieldRef>": [ // Each key is a JSON reference field name.
            {
                "message": "string", // Field error message.
                "error": "string", // Specific error code, usually a schema attribute.
                "code": "integer" // Optional field error code.
            }
        ]
    }
}

This format is optimized for helping present errors to user interfaces. You can loop through the specific errors collection and line up errors with their inputs on a user interface. For deeply nested objects, the field name is a JSON reference.

Schema References

OpenAPI allows for schemas to be accessed with references using the $ref attribute. Using references allows you to define commonly used schemas in one place and then reference them from many locations.

To use references you must:

1. Define the schema you want to reference somewhere.
2. Reference the schema with a $ref attribute.
3. Add a schema lookup function to your main schema with Schema::setRefLookp()

Defining a Reusable Schema

The OpenAPI specification places all reusable schemas under /components/schemas. If you are defining everything in a big array that is a good place to put them.

$components = [
    'components' => [
        'schemas' => [
            'User' => [
                'type' => 'object',
                'properties' => [
                    'id' => [
                        'type' => 'integer'
                    ],
                    'username' => [
                        'type' => 'string'
                    ]
                ]
            ]
        ]
    ]
]

Referencing Schemas With $ref

Reference the schema's path with keys separated by / characters.

$userArray = [
    'type' => 'array',
    'items' => [
        '$ref' => '#/components/schemas/User'
    ]
]

Using Schema::setRefLookup() to Resolve References

The Schema class has a setRefLookup() method that lets you add a callable that is use to resolve references. The callable should have the following signature:

function(string $ref): array|Schema|null {
   ...
}

The function takes the string from the $ref attribute and returns a schema array, Schema object, or null if the schema cannot be found. Garden Schema has a default implementation of a ref lookup in the ArrayRefLookup class that can resolve references from a static array. This should be good enough for most uses, but you are always free to define your own.

You can put everything together like this:

$sch = new Schema($userArray);
$sch->setRefLookup(new ArrayRefLookup($components));

$valid = $sch->validate(...);

The references are resolved during validation so if there are any mistakes in your references then a RefNotFoundException is thrown during validation, not when you set your schema or ref lookup function.

Schema Polymorphism

Schemas have some support for implementing schema polymorphism by letting you validate an object against different schemas depending on its value.

The discriminator Property

The discriminator of a schema lets you specify an object property that specifies what type of object it is. That property is then used to reference a specific schema for the object. The discriminator has the following format:

{
    "discriminator": {
        "propertyName": "<string>", // Name of the property used to reference a schema.
        "mapping": {
          "<propertyValue1>": "<ref>", // Reference to a schema.
          "<propertyValue>": "<alias>" // Map a value to another value.
        }
    }
}

You can see above that the propertyName specifies which property is used as the discriminator. There is also an optional mapping property that lets you control how schemas are mapped to values. discriminators are resolved int he following way:

1. The property value is mapped using the mapping property.
2. If the value is a valid JSON reference then it is looked up. Only values in mappings can specify a JSON reference in this way.
3. If the value is not a valid JSON reference then it is is prepended with #/components/schemas/ to make a JSON reference.

Here is an example at work:

{
  "discriminator": {
    "propertyName": "petType",
    "mapping": {
      "dog": "#/components/schemas/Dog", // A direct reference.
      "fido": "Dog" // An alias that will be turned into a reference.
    }
  }
}

The oneOf Property

The oneOf property works in conjunction with the discriminator to limit the schemas that the object is allowed to validate against. If you don't specify oneOf then any schemas under #/components/schemas are fair game.

To use the oneOf property you must specify $ref nodes like so:

{
  "oneOf": [
    { "$ref": "#/components/schemas/Dog" },
    { "$ref": "#/components/schemas/Cat" },
    { "$ref": "#/components/schemas/Mouse" },
  ],
  "discriminator": {
    "propertyType": "species"
  }
}

In the above example the "species" property will be used to construct a reference to a schema. That reference must match one of the references in the oneOf property.

If you are familiar with with OpenAPI spec please note that inline schemas are not currently supported for oneOf in Garden Schema.

Validation Options

Both validate() and isValid() can take an additional $options argument which modifies the behavior of the validation slightly, depending on the option.

The request Option

You can pass an option of ['request' => true] to specify that you are validating request data. When validating request data, properties that have been marked as readOnly: true will be treated as if they don't exist, even if they are marked as required.

The response Option

You can pass an option of ['response' => true] to specify that you are validating response data. When validating response data, properties that have been marked as writeOnly: true will be treated as if they don't exist, even if they are marked as required.

The sparse Option

You can pass an option of ['sparse' => true] to specify a sparse validation. When you do a sparse validation, missing properties do not give errors and the sparse data is returned. Sparse validation allows you to use the same schema for inserting vs. updating records. This is common in databases or APIs with POST vs. PATCH requests.

Flags

Flags can be applied a schema to change it's inherit validation.

use Garden\Schema\Schema;
$schema = Schema::parse([]);

// Enable a flag.
$schema->setFlag(Schema::VALIDATE_STRING_LENGTH_AS_UNICODE, true);

// Disable a flag.
$schema->setFlag(Schema::VALIDATE_STRING_LENGTH_AS_UNICODE, false);

// Set all flags together.
$schema->setFlags(Schema::VALIDATE_STRING_LENGTH_AS_UNICODE & Schema::VALIDATE_EXTRA_PROPERTY_NOTICE);

// Check if a flag is set.
$schema->hasFlag(Schema::VALIDATE_STRING_LENGTH_AS_UNICODE); // true

VALIDATE_STRING_LENGTH_AS_UNICODE

By default, schema's validate str lengths in terms of bytes. This is useful because this is the common unit of storage for things like databases.

Some unicode characters take more than 1 byte. An emoji like 😱 takes 4 bytes for example.

Enable this flag to validate unicode character length instead of byte length.

VALIDATE_EXTRA_PROPERTY_NOTICE

Set this flag to trigger notices whenever a validated object has properties not defined in the schema.

VALIDATE_EXTRA_PROPERTY_EXCEPTION

Set this flag to throw an exception whenever a validated object has properties not defined in the schema.

Custom Validation with addValidator()

You can customize validation with Schema::addValidator(). This method lets you attach a callback to a schema path. The callback has the following form:

function (mixed $value, ValidationField $field): bool {
}

The callback should true if the value is valid or false otherwise. You can use the provided ValidationField to add custom error messages.

Filtering Data

You can filter data before it is validating using Schema::addFilter(). This method lets you filter data at a schema path. The callback has the following form:

function (mixed $value, ValidationField $field): mixed {
}

The callback should return the filtered value. Filters are called before validation occurs so you can use them to clean up date you know may need some extra processing.

The Schema::addFilter() also accepts $validate parameter that allows your filter to validate the data and bypass default validation. If you are validating date in this way you can add custom errors to the ValidationField parameter and return Invalid::value() your validation fails.

Format Filters

You can also filter all fields with a particular format using the Schema::addFormatFilter(). This method works similar to Schema::addFilter() but it applies to all fields that match the given format. You can even use format filters to override default format processing.

$schema = new Schema([...]);

// By default schema returns instances of DateTimeImmutable, instead return a string.
$schema->addFormatFilter('date-time', function ($v) {
    $dt = new \DateTime($v);
    return $dt->format(\DateTime::RFC3339);
}, true);

Overriding the Validation Class and Localization

Since schemas generate error messages, localization may be an issue. Although the Garden Schema doesn't offer any localization capabilities itself, it is designed to be extended in order to add localization yourself. You do this by subclassing the Validation class and overriding its translate() method. Here is a basic example:

class LocalizedValidation extends Validation {
    public function translate($str) {
        if (substr($str, 0, 1) === '@') {
            // This is a literal string that bypasses translation.
            return substr($str, 1);
        } else {
            return gettext($str);
        }
    }
}

// Install your class like so:
$schema = Schema::parse([...]);
$schema->setValidationClass(LocalizedValidation::class);

There are a few things to note in the above example:

• When overriding translate() be sure to handle the case where a string starts with the '@' character. Such strings should not be translated and have the character removed.

• You tell a Schema object to use your specific Validation subclass with the setValidationClass(). This method takes either a class name or an object instance. If you pass an object it will be cloned every time a validation object is needed. This is good when you want to use dependency injection and your class needs more sophisticated instantiation.

JSON Schema Support

The Schema object is a wrapper for an OpenAPI Schema array. This means that you can pass a valid JSON schema to Schema's constructor. The table below lists the JSON Schema properties that are supported.

Property	Applies To	Notes
allOf	Schema[]	An instance validates successfully against this keyword if it validates successfully against all schemas defined by this keyword's value.
multipleOf	integer/number	A numeric instance is only valid if division by this keyword's value results in an integer.
maximum	integer/number	If the instance is a number, then this keyword validates only if the instance is less than or exactly equal to "maximum".
exclusiveMaximum	integer/number	If the instance is a number, then the instance is valid only if it has a value strictly less than (not equal to) "exclusiveMaximum".
minimum	integer/number	If the instance is a number, then this keyword validates only if the instance is greater than or exactly equal to "minimum".
exclusiveMinimum	integer/number	If the instance is a number, then the instance is valid only if it has a value strictly greater than (not equal to) "exclusiveMinimum".
maxLength	string	Limit the unicode character length of a string.
minLength	string	Minimum length of a string.
maxByteLength	string	Maximum byte length of the the property.
pattern	string	A regular expression without delimiters. You can add a custom error message with the x-patternMessageCode field.
items	array	Ony supports a single schema.
maxItems	array	Limit the number of items in an array.
minItems	array	Minimum number of items in an array.
uniqueItems	array	All items must be unique.
maxProperties	object	Limit the number of properties on an object.
minProperties	object	Minimum number of properties on an object.
additionalProperties	object	Validate additional properties against a schema. Can also be true to always validate.
required	object	Names of required object properties.
properties	object	Specify schemas for object properties.
enum	any	Specify an array of valid values.
type	any	Specify a type of an array of types to validate a value.
default	object	Applies to a schema that is in an object property.
format	string	Support for date-time, email, ipv4, ipv6, ip, uri.
oneOf	object	Works with the discriminator property to validate against a dynamic schema.

OpenAPI Schema Support

OpenAPI defines some extended properties that are applied during validation.

Property	Type	Notes
nullable	boolean	If a field is nullable then it can also take the value null.
readOnly	boolean	Relevant only for Schema "properties" definitions. Declares the property as "read only". This means that it MAY be sent as part of a response but SHOULD NOT be sent as part of the request. If the property is marked as readOnly being true and is in the required list, the required will take effect on the response only.
writeOnly	boolean	Relevant only for Schema "properties" definitions. Declares the property as "write only". Therefore, it MAY be sent as part of a request but SHOULD NOT be sent as part of the response. If the property is marked as writeOnly being true and is in the required list, the required will take effect on the request only.
discriminator	object	Validate against a dynamic schema based on a property value.