Rules engine specification#

The Smithy rules engine provides service owners with a collection of traits and components to define rule sets. Rule sets can be consumed by code generators to provide the rule set's specific functionality.

A rule set defines zero or more parameters and one or more rules. Parameters define the primary set of values that a rule set operates on. Rules are composed of a set of conditions, which determine if a rule should be selected, and a result. Conditions act on the defined parameters, and allow for the modeling of statements.

When a rule’s conditions are evaluated successfully, the rule provides either a result and its accompanying requirements or an error describing the unsupported state. Modeled endpoint errors allow for more explicit descriptions to users, such as providing errors when a service doesn't support a combination of conditions.

`smithy.rules#endpointRuleSet` trait#

Summary: Defines a rule set for deriving service endpoints at runtime.
Trait selector: service
Value type: document

The content of the endpointRuleSet document has the following properties:

Property name	Type	Description
version	`string`	Required. The rule set schema version. This specification covers version 1.0 of the endpoint rule set.
serviceId	`string`	Required. An identifier for the corresponding service.
parameters	`map<string, parameter object>` of Parameter object	Required. A map of zero or more endpoint parameter names to their parameter configuration.
rules	An `array` of one or more of the following types: Endpoint rule object, Error rule object, or Tree rule object	Required. One or more endpoint rule definitions of any rule type.

A rule set defines endpoint parameters using the parameters property, and maps zero or more parameter names to parameter object values.

Finally a rule set defines the rules property, which MUST contain one or more of the following rule types: endpoint rules, error rules, or tree rules. Rules are evaluated in order from their lowest array index position to highest. If the list of rules is exhausted with none evaluated successfully, then an implementation of the rule set engine MUST stop and return an error indicating that rule exhaustion has occurred. Rule authors SHOULD use either an endpoint rule or error rules, with an empty set of conditions to provide a more meaningful default or error depending on the scenario.

Parameter object#

A parameter object contains the following properties:

Property name	Type	Description
type	`string`	Required. MUST be one of `string` or `boolean`.
builtIn	`string`	Specifies a named built-in value that is sourced and provided to the endpoint provider by a caller.
default	`string` or `boolean`	Specifies the default value for the parameter if not set. Parameters with defaults MUST also be marked as `required`. The type of the provided default MUST match `type`.
required	`boolean`	Specifies that the parameter is required to be provided to the endpoint provider.
documentation	`string`	Required. Specifies a string that SHOULD be used to generate API reference documentation for the endpoint parameter.
deprecated	Deprecated object	Specifies whether an endpoint parameter has been deprecated.

The parameter typing is statically analyzed by the rules engine to validate correct usage within the rule set. Rules engine parameter traits allow values to be bound to parameters from other locations in generated clients.

Parameters MAY be annotated with the builtIn property, which designates that the parameter should be bound to a value determined by the built-in’s name. The rules engine contains built-ins and the set is extensible.

The required property is used to validate that a parameter value MUST be set prior to evaluating the rule set's rules. If a required parameter is not set, rule evaluation MUST NOT proceed, and an implementation MUST return an error to the user.

The default property is used to set a default value on a property if a value is not present. A parameter with the default property, MUST also be marked with the required trait.

The following is an example of a parameter that is marked required.

{
    "parameters": {
        "linkId": {
            "type": "string",
            "documentation": "The identifier of the link to target.",
            "required": true
        }
    }
}

Note

Why must required be set when default is present?

The required property exists as sole the determiner for rules code generation of "is the parameter always going to be set?" The default property exists as a signal of where we can get the value from if it's missing. By forcing both to be set, code generators can simplify their handling of required.

Deprecated object#

A deprecated object contains the following properties:

Property name	Type	Description
message	`string`	Specifies an optional message that can be used in documentation to provide recourse options to a user.
since	`string`	A date string that indicates when the parameter field was deprecated.

The following is an example of a parameter that is marked as deprecated.

{
    "parameters": {
        "linkId": {
            "type": "string",
            "deprecated": {
                "message": "This feature has been deprecated, and requests are now directed towards a global endpoint.",
                "since": "2020-07-02"
            },
        }
    }
}

Endpoint rule object#

An endpoint rule object contains the following properties:

Property name	Type	Description
type	`string`	Required. MUST be `endpoint`.
conditions	An `array` of Condition object	Required. Zero or more conditions used to determine whether the endpoint rule should be selected.
endpoint	Endpoint object	Required. The endpoint to return if this rule is selected.
documentation	`string`	A description of the rule.

An endpoint rule MUST contain zero or more conditions. If all condition clauses evaluate successfully, the endpoint rule is selected. If a condition fails, evaluation of the rule MUST be terminated and evaluation proceeds to any subsequent rules.

The following example defines an endpoint rule object that checks a condition and uses a parameter value in a url template:

{
    "documentation": "An endpoint rule description",
    "type": "endpoint",
    "conditions": [
        {
            "fn": "isValidHostLabel",
            "argv": [
                {
                    "ref": "linkId"
                }
            ]
        }
    ],
    "endpoint": {
        "url": "https://{linkId}.service.com"
    }
}

Endpoint object#

An endpoint object contains the following properties:

Property name	Type	Description
url	`string` or Reference object or Function object	Required. The endpoint url. This MUST specify a scheme and hostname and MAY contain port and base path components. A `string` value MAY be a Template string. Any value for this property MUST resolve to a `string`.
properties	`map<string, object>`	A map containing zero or more key value property pairs. Endpoint properties MAY be arbitrarily deep and contain other maps and arrays.
headers	`map<string, array>` where the `array` value is one or more of following types: `string` or Reference object or Function object	A map of transport header names to their respective values. A `string` value in an array MAY be a template string.

An endpoint rule uses an endpoint object to define an endpoint selected based on successful evaluation of rule conditions to that point.

An endpoint MAY return a set of endpoint properties using the properties field. This can be used to provide a grab-bag set of metadata associated with an endpoint that an endpoint resolver implementation MAY use. For example, the authSchemes property is used to specify the priority ordered list of authentication schemes and their configuration supported by the endpoint. Properties MAY contain arbitrary nested maps and arrays of strings and booleans.

Note

To prevent ambiguity, the endpoint properties map MUST NOT contain reference or function objects. Properties MAY contain template string

Endpoint `authSchemes` list property#

The authSchemes property of an endpoint is used to specify the priority ordered list of authentication schemes and their configuration supported by the endpoint. The property is a list of configuration objects that MUST contain at least a name property and MAY contain additional properties. Each configuration object MUST have a unique value for its name property within the list of configuration objects within a given authSchemes property.

If an authSchemes property is present on an Endpoint object, clients MUST resolve an authentication scheme to use via the following process:

Iterate through configuration objects in the authSchemes property.
If the name property in a configuration object contains a supported authentication scheme, resolve this scheme.
If the name is unknown or unsupported, ignore it and continue iterating.
If the list has been fully iterated and no scheme has been resolved, clients MUST return an error.

Adding `authSchemes` configuration validators#

Extensions to the rules engine can provide additional validators for authSchemes configuration objects. No validators are provided by default.

The rules engine is highly extensible through software.amazon.smithy.rulesengine.language.EndpointRuleSetExtension service providers. See the Javadocs for more information.

Error rule object#

An error rule object contains the following properties:

Property name	Type	Description
type	`string`	Required. MUST be `error`.
conditions	An `array` of Condition object	Required. Zero or more conditions used to determine whether the endpoint rule should be selected.
error	`string` or Reference object or Function object	Required. A descriptive message describing the error for consumption by the caller. A `string` value MAY be a Template string. Any value for this property MUST resolve to a `string`.
documentation	`string`	A description of the rule.

An error rule MUST contain zero or more conditions. If all condition clauses evaluate successfully or zero conditions are defined, then the error rule MUST be selected. If a condition fails evaluation, the rule MUST be terminated and evaluation proceeds to any subsequent rules.

The following example defines an error rule object that checks a condition:

{
    "documentation": "An error rule description.",
    "type": "error",
    "conditions": [
        {
            "fn": "not",
            "argv": [
                {
                    "fn": "isValidHostLabel",
                    "argv": [
                        {
                            "ref": "linkId"
                        }
                    ]
                }
            ]
        }
    ],
    "error": "{linkId} must be a valid HTTP host label."
}

Note

In production rule sets, rather than using not, it's more common to see rules where the error rule has no conditions and is only matched after all other rules have failed to match.

Tree rule object#

A tree rule object contains the following properties:

Property name	Type	Description
type	`string`	Required. MUST be `endpoint`.
conditions	An `array` of Condition object	Required. Zero or more conditions used to determine whether the endpoint rule should be selected.
rules	An `array` of one or more of the following types: Endpoint rule object, Error rule object, or Tree rule object	Required. One or more endpoint rule definitions of any rule type.
documentation	`string`	A description of the rule.

A tree rule MUST contain one or more conditions. If all condition clauses evaluate successfully, the tree rule is selected. If a condition fails, evaluation of the rule MUST be terminated and evaluation proceeds to any subsequent rules.

A tree rule MUST have one or more subordinate rules specified using the rules property. A tree rule is equivalent to the following logical expression:

treeCondition1 && ... && treeConditionN && ( subRule1 || ... || subRuleN )

Tree rules are considered terminal branches of the rule set. If a tree rule’s subordinate rules are exhausted with none evaluated successfully, then an implementation of the rules engine MUST stop and return an error indicating that rule exhaustion has occurred.

Rule authors SHOULD use either an Endpoint rule object or Error rule object with an empty set of conditions to provide a more meaningful default or error, depending on the scenario.

The following example is an abbreviated example of a tree rule that consists of a tree, endpoint, and error rule.

{
    "conditions": [
        {
            "fn": "isValidHostLabel",
            "argv": [
                {
                    "ref": "linkId"
                }
            ]
        }
    ],
    "type": "tree",
    "rules": [
        {
            "type": "tree",
            "conditions": [
                // Abbreviated for clarity
            ],
            "rules": [
                // Abbreviated for clarity
            ]
        },
        {
            "type": "endpoint",
            "conditions": [
                // Abbreviated for clarity
            ],
            "endpoint": {
                "url": "{linkId}.service.com"
            }
        },
        {
            "type": "error",
            "conditions": [
                // Abbreviated for clarity
            ],
            "error": "An error message."
        }
    ]
}

Condition object#

A condition is specified using an object containing the following properties:

Property	Type	Description
fn	`string`	Required. The name of the function to be executed.
argv	An `array` of one or more of the following types: `string`, `bool`, `array`, Reference object, or Function object	Required. The arguments for the function.
assign	`string`	The destination variable to assign the functions result to.

Conditions are defined within rule objects as requirements for continuing to evaluate the rules within. Conditions are evaluated in-order by their positional index in the array, starting from zero. Conditions represent the logical expression condition1 && ... && conditionN where condition1 through conditionN are each condition objects.

If a condition returns None or False, the condition does not match. A condition that does not match MUST immediately terminate the evaluation of the rule. Processing starts at the next immediate rule to be evaluated. In order for a rule's right-hand-side to be considered, all conditions MUST match. Conditions MAY use references defined in previous conditions in the same rule.

A condition object MAY assign the output result of a function to a named variable using the assign property. Variables created using assign follow the same rules as parameter name identifiers. Variables are scoped to the rule they are defined in, with variables created in a tree rule being visible to all subordinate rules and their children. For endpoint rules or error rules, the variable will fall out of scope once evaluation of the rule has completed. A condition may also reference a variable declared by a prior condition statement within the same rule. Assigned variables are similar to parameters and can be referenced in string templates or passed as values to other functions. Variable names MUST NOT overlap with existing parameter names or variable declarations in scope.

The following example shows using the stringEquals function to compare two arguments as a prerequisite to an endpoint object:

{
    "type": "endpoint",
    "conditions": [
        {
            "documentation": "Use the base endpoint if using the default link",
            "fn": "stringEquals",
            "argv": [
                {
                    "ref": "linkId"
                },
                "default"
            ]
        }
    ],
    "endpoint": {
        "url": "https://service.com"
    }
}

The following example shows the invocation of the parseUrl function, taking a single argument and assigning the result to the url variable. The url variable is then used as part of a template string in an endpoint object.

{
    "type": "endpoint",
    "conditions": [
        {
            "fn": "parseURL",
            "argv": [
                {
                    "ref": "Endpoint"
                }
            ],
            "assign": "url"
        }
    ],
    "endpoint": {
        "url": "https://service.com/{url#path}"
    }
}

Function object#

A function is specified using an object containing the following properties:

Property	Type	Description
fn	`string`	Required. The name of the function to be executed.
argv	An `array` of one or more of the following types: `string`, `bool`, `array`, Reference object, or Function object	Required. The arguments for the function.

All functions have specific type signatures associated with their definition. Static analysis validates that all arguments passed to a function are of the correct type.

The rules engine contains functions and the set is extensible.

Reference object#

A reference is an object containing the following properties:

Property	Type	Description
ref	`string`	Required. The name of the parameter or variable.

References allow for parameter and variable assignments, created by a condition object, to be used in subsequent rules and conditions. A parameter MUST be checked with an isSet condition object to determine whether the parameter is set and can be safely used. The rules engine statically verifies references are valid at time-of-use.

Template string#

String values in rules can use the {parameterName} syntax for defining automatic templating of the named parameter or variable values into a string. For example, if the string {parameterName}.foobar.baz is defined in the rule set and parameterName is bound to the value foo, the resolved string value would be foo.foobar.baz.

Template string parameters follow the same de-referencing rules as outlined in reference object. The rules engine statically verifies referenced parameters in template strings.

Template shorthand#

Template parameters also support a shorthand syntax for accessing nested properties or array arguments using the # character. This syntax is syntactic sugar for using the long-form getAttr function. For example, {parameterName#foo}, with parameterName being a variable containing a JSON object, and foo being a property defined on that object.

The following two expressions are equivalent:

{
    "fn": "stringEquals",
    "argv": [
        "linkId",
        {
            "fn": "getAttr",
            "argv": [
                {
                    "ref": "partResult"
                },
                [
                    "name"
                ]
            ]
        }
    ]
}

{
    "fn": "stringEquals",
    "argv": [
        "linkId",
        "{partResult#name}"
    ]
}

Rules engine specification#

smithy.rules#endpointRuleSet trait#

Parameter object#

Deprecated object#

Endpoint rule object#

Endpoint object#

Endpoint authSchemes list property#

Adding authSchemes configuration validators#

Error rule object#

Tree rule object#

Condition object#

Function object#

Reference object#

Template string#

Template shorthand#

`smithy.rules#endpointRuleSet` trait#

Endpoint `authSchemes` list property#

Adding `authSchemes` configuration validators#