Linting Models#

This guide describes how to apply optional linters to a Smithy model.

Linting overview#

A linter is a special kind of model validator that is configurable. Linter implementations are found in code. The smithy-linters package in Maven Central contains several linters that can be used to apply additional validation to Smithy models.

The following example adds smithy-linters as a dependency to a Smithy project:

Smithy CLI

smithy-build.json#

{
    "...": "...",
    "maven": {
        "dependencies": [
            "software.amazon.smithy:smithy-linters:1.53.0"
        ]
    },
    "...": "..."
}

Gradle

Kotlin

build.gradle.kts#

dependencies {
    ...
    implementation("software.amazon.smithy:smithy-linters:1.53.0")
}

Groovy

build.gradle#

dependencies {
    ...
    implementation 'software.amazon.smithy:smithy-linters:1.53.0'
}

After the dependency is added and available on the Java classpath, validators defined in the package and registered using Java SPI are available for use in Smithy models.

Linters in `smithy-linters`#

For Java developers using Smithy's reference implementation, the following linters (except for UnreferencedShape) are defined in the software.amazon.smithy:smithy-linters Maven package.

UnreferencedShape#

Emits when a shape is not connected to the rest of the model. If no configuration is provided, the linter will check if a shape is connected to the closure of any service shape. A selector can be provided to define a custom set of "root" shapes to customize how the linter determines if a shape is unreferenced. Shapes that are connected through the idRef trait are considered connected.

Rationale

Just like unused variables in code, removing unused shapes from a model makes the model easier to maintain.

Default severity

NOTE

Configuration

Property	Type	Description
rootShapeSelector	`string`	A selector that specifies the root shape(s) from which to detect if other shapes are connected. Defaults to "service", meaning any shape connected to any service shape in the model is considered referenced.

Example:

$version: "2"

metadata validators = [
    {
        // Find shapes that aren't connected to service shapes or shapes
        // marked with the trait, smithy.example#myCustomTrait.
        name: "UnreferencedShape"
        configuration: {
            rootShapeSelector: ":is(service, [trait|smithy.example#myCustomTrait])"
        }
    }
]

Note

For backward compatibility reasons, the UnreferencedShape validator is available in software.amazon.smithy:smithy-model Maven package and does not require additional dependencies.

AbbreviationName#

Validates that shape names and member names do not represent abbreviations with all uppercase letters. For example, instead of using "XMLRequest" or "instanceID", this validator recommends using "XmlRequest" and "instanceId".

Rationale

Using a strict form of camelCase where abbreviations are written just like other words makes names more predictable and easier to work with in tooling. For example, a tool that generates code in Python might wish to represent camelCase words using snake_case; utilizing strict camel casing makes it easier to split words apart.

Default severity

DANGER

Configuration

Property	Type	Description
allowedAbbreviations	[ `string` ]	A case-insensitive list of abbreviations to allow to be all capital letters. Defaults to an empty list.

Example:

$version: "2"

metadata validators = [
    {name: "AbbreviationName"}
]

CamelCase#

Validates that shape names and member names adhere to a consistent style of camel casing. By default, this validator will ensure that shape names use UpperCamelCase, trait shape names use lowerCamelCase, and that member names use either lowerCamelCase or UpperCamelCase (depending on which is currently more prevalent in each service closure).

Rationale

Utilizing a consistent camelCase style makes it easier to understand a model and can lead to consistent naming in code generated from Smithy models.

Default severity

DANGER

Configuration

Property	Type	Description
memberNames	`string`	Specifies the camelCase style of member names. Can be set to either "upper", "lower", or "auto" (the default).

Example:

$version: "2"

metadata validators = [
    {name: "CamelCase"}
]

MissingSensitiveTrait#

This validator scans shape or member names and identifies ones that look like they could contain sensitive information but are not marked with the @sensitive trait. This does not apply to shapes where the @sensitive trait would be invalid. Users may also configure this validator with a custom list of terms, and choose to ignore the built-in defaults. The defaults terms include types of personal information such as 'birth day', 'billing address', 'zip code', or 'gender', as well as information that could be maliciously exploited such as 'password', 'secret key', or 'credit card'.

Rationale

Sensitive information often incurs legal requirements regarding the handling and logging of it. Mistakenly not marking sensitive data accordingly carries a large risk, and it is helpful to have an automated validator to catch instances of this rather than rely on best efforts.

Default severity

WARNING

Configuration

Property	Type	Description
terms	[ `string` ]	A list of search terms that match shape or member names case-insensitively based on word boundaries (for example, the term "access key id" matches "AccessKeyId", "access_key_id", and "accesskeyid"). See Words boundary matching for details.
excludeDefaults	`boolean`	A flag indicating whether or not to disregard the default set of terms. This property is not required and defaults to false. If set to true, `terms` must be provided.

Example:

$version: "2"

metadata validators = [{
    name: "MissingSensitiveTrait"
    configuration: {
        excludeDefaults: false,
        terms: ["home planet"]
    }
}]

NoninclusiveTerms#

Validates that all text content in a model (i.e. shape names, member names, documentation, trait values, etc.) does not contain words that perpetuate cultural biases. This validator has a built-in set of bias terms that are commonly found in APIs along with suggested alternatives.

Noninclusive terms are case-insensitively substring matched and can have any number of leading or trailing whitespace or non-whitespace characters.

This validator has built-in mappings from noninclusive terms to match model text to suggested alternatives. The configuration allows for additional terms to suggestions mappings to either override or append the built-in mappings. If a match occurs and the suggested alternatives is empty, no suggestion is made in the generated warning message.

Rationale

Intent doesn't always match impact. The use of noninclusive language like "whitelist" and "blacklist" perpetuates bias through past association of acceptance and denial based on skin color. Other words should be used that are not only inclusive, but more clearly communicate meaning. Words like allowList and denyList much more clearly indicate that something is allowed or denied.

Default severity

WARNING

Configuration

Property	Type	Description
terms	{ `keyword` -> [ `alternatives` ] }	A set of noninclusive terms to suggestions to either override or replace the built-in mappings. This property is not required unless `excludeDefaults` is true. The default value is the empty set.
excludeDefaults	`boolean`	A flag indicating whether or not the mappings set specified by `terms` configuration replaces the built-in set or appends additional mappings. This property is not required and defaults to false.

Example:

$version: "2"

metadata validators = [{
    name: "NoninclusiveTerms"
    configuration: {
        excludeDefaults: false,
        terms: {
            mankind: ["humankind"],
            mailman: ["mail carrier", "postal worker"]
        }
    }
}]

ReservedWords#

Validates that shape names and member names do not match a configured set of reserved words.

Rationale

Tools that generate code from Smithy models SHOULD automatically convert reserved words into symbols that are safe to use in the targeted programming language. This validator can be used to warn about these conversions as well as to prevent sensitive words, like internal code-names, from appearing in public artifacts.

Default Severity

DANGER

Configuration

A single key, reserved, is Required in the configuration. Its value is a list of objects with the following properties:

Property	Type	Description
words	[ `string` ]	A list of words that shape or member names MUST not case-insensitively match. Supports a leading and trailing wildcard character of "*". See Wildcards in ReservedWords for details.
terms	[ `string` ]	A list of search terms that match shape or member names case-insensitively based on word boundaries (for example, the term "access key id" matches "AccessKeyId", "access_key_id", and "accesskeyid"). See Words boundary matching for details.
selector	`string`	Specifies a selector of shapes to validate for this configuration. Defaults to validating all shapes, including member names. Note When evaluating member shapes, the member name will be evaluated instead of the shape name.
reason	`string`	A reason to display for why this set of words is reserved.

Example:

$version: "2"

metadata validators = [{
    id: "FooReservedWords"
    name: "ReservedWords"
    configuration: {
        reserved: [
            {
                words: ["Codename"]
                reason: "This is the internal project name."
            }
        ]
    }
}]

Wildcards in ReservedWords#

The ReservedWords validator allows leading and trailing wildcard characters to be specified.

Using both a leading and trailing wildcard indicates that shape or member names match when case-insensitively containing the word. The following table shows matches for a reserved word of *codename*:

Example Result

CreateCodenameInput Match

CodenameResource Match

ReferencedCodename Match

Codename Match
Using a leading wildcard indicates that shape or member names match when case-insensitively ending with the word. The following table shows matches for a reserved word of *codename:

Example Result

CreateCodenameInput No match

CodenameResource No match

ReferencedCodename Match

Codename Match
Using a trailing wildcard indicates that shape or member names match when case-insensitively starting with the word. The following table shows matches for a reserved word of codename*:

Example Result

CreateCodenameInput No match

CodenameResource Match

ReferencedCodename No Match

Codename Match
Using no wildcards indicates that shape or member names match when case-insensitively the same as the word. The following table shows matches for a reserved word of codename:

Example Result

CreateCodenameInput No match

CodenameResource No match

ReferencedCodename No match

Codename Match

Example	Result
CreateCodenameInput	Match
CodenameResource	Match
ReferencedCodename	Match
Codename	Match

Example	Result
CreateCodenameInput	No match
CodenameResource	No match
ReferencedCodename	Match
Codename	Match

Example	Result
CreateCodenameInput	No match
CodenameResource	Match
ReferencedCodename	No Match
Codename	Match

Example	Result
CreateCodenameInput	No match
CodenameResource	No match
ReferencedCodename	No match
Codename	Match

Words boundary matching#

Word boundaries can be used to find terms of interest. Word boundary search text consists of one or more alphanumeric words separated by a single space. When comparing against another string, the contents of the string are separated into words based on word boundaries. Those words are case-insensitively compared against the words in the search text for a match.

Word boundaries are detected when the casing between two characters changes, or the type of character between two characters changes. The following table demonstrates how comparison text is parsed into words.

Comparison text	Parsed words
accessKey	access key
accessKeyID	access key id
accessKeyIDValue	access key id value
accesskeyId	accesskey id
accessKey1	access key 1
access_keyID	access key id

The following table shows matches for a search term of secret id, meaning the word "secret" needs to be followed by the word "id". Word boundary searches also match if the search terms concatenated together with no spaces is considered a word in the search text (for example, secret id will match the word secretid).

Comparison text	Result
SomeSecretId	Match
SomeSecretIDValue	Match
SomeSecret__ID__value	Match
secret_id	Match
secret_id100	Match
secretid	Match
secretid_value	Match
secretidvalue	No Match
SecretThingId	No match
SomeSecretid	No match

Syntax restrictions

Empty search terms are not valid.
Only a single space can appear between words in word boundary patterns.
Leading and trailing spaces are not permitted in word boundary patterns.
Word boundary patterns can only contain alphanumeric characters.

StandardOperationVerb#

Looks at each operation shape name and determines if the first word in the operation shape name is one of the defined standard verbs or if it is a verb that has better alternatives.

Note

Operations names MUST use a verb as the first word in the shape name in order for this validator to properly function.

Rationale

Using consistent verbs for operation shape names helps consumers of the API to more easily understand the semantics of an operation.

Default severity

DANGER

Configuration

Property	Type	Description
verbs	[ `string` ]	The list of verbs that each operation shape name MUST start with.
prefixes	[ `string` ]	A list of prefixes that MAY come before one of the valid verbs. Prefixes are often used to group families of operations under a common prefix (e.g., `batch` might be a common prefix in some organizations). Only a single prefix is honored.
suggestAlternatives	`object`	Used to recommend alternative verbs. Each key is the name of a verb that should be changed, and each value is a list of suggested verbs to use instead.

Note

At least one verb or one suggestAlternatives key-value pair MUST be provided.

Example:

$version: "2"

metadata validators = [{
    name: "StandardOperationVerb"
    configuration: {
        verbs: ["Register", "Deregister", "Associate"]
        prefixes: ["Batch"]
        suggestAlternatives: {
            "Make": ["Create"]
            "Transition": ["Update"]
        }
    }
}]

RepeatedShapeName#

Validates that Structure member names and Union member names do not case-insensitively repeat their container shape names.

As an example, if a structure named "Table" contained a member named "TableName", then this validator would emit a WARNING event.

Rationale

Repeating a shape name in the members of identifier of the shape is redundant.

Default severity

WARNING

Configuration

Property	Type	Description
exactMatch	`boolean`	If set to true, the validator will only warn if the member name is case-insensitively identical to the containing shape's name.

InputOutputStructureReuse#

Validates that every operation defines a dedicated input and output shape marked with the input trait and output trait.

Rationale

Using the same structure for both input and output can lead to backward-compatibility problems in the future if the members or traits used in input needs to diverge from those used in output. It is always better to use structures that are exclusively used as input or exclusively used as output.
Referencing the same input or output structure from multiple operations can lead to backward-compatibility problems in the future if the inputs or outputs of the operations ever need to diverge. By using the same structure, you are unnecessarily tying the interfaces of these operations together.

Default severity

DANGER

MissingPaginatedTrait#

Checks for operations that look like they should be paginated but do not have the paginated trait.

Rationale

Paginating operations that can return potentially unbounded lists of data helps to maintain a predictable SLA and helps to prevent operational issues in the future.

Default severity

DANGER

Configuration

Property	Type	Description
verbsRequirePagination	[`string`]	Defines the case-insensitive operation verb prefixes for operations that MUST be paginated. A `DANGER` event is emitted for any operation that has a shape name that starts with one of these verbs. Defaults to `["list", "search"]`.
verbsSuggestPagination	[`string`]	Defines the case-insensitive operation verb prefixes for operations that SHOULD be paginated. A `WARNING` event is emitted when an operation is found that matches one of these prefixes, the operation has output, and the output contains at least one top-level member that targets a List. Defaults to `["describe", "get"]`
inputMembersRequirePagination	[`string`]	Defines the case-insensitive operation input member names that indicate that an operation MUST be paginated. A `DANGER` event is emitted if an operation is found to have an input member name that case-insensitively matches one of these member names. Defaults to `["maxresults", "maxitems", "pagesize", "limit", "nexttoken", "pagetoken", "token", "marker"]`
outputMembersRequirePagination	[`string`]	Defines the case-insensitive operation output member names that indicate that an operation MUST be paginated. A `DANGER` event is emitted if an operation is found to have an output member name that case-insensitively matches one of these member names. Defaults to `["nexttoken", "pagetoken", "token", "marker", "nextpage", "nextpagetoken", "position", "nextmarker", "paginationtoken", "nextpagemarker"]`.

Example:

metadata validators = [
    {name: "MissingPaginatedTrait"}
]

ShouldHaveUsedTimestamp#

Looks for shapes that likely represent time, but that do not use a timestamp shape.

The ShouldHaveUsedTimestamp validator checks the following names:

string shape names
short, integer, long, float, and double shape names
structure member names
union member names

The ShouldHaveUsedTimestamp validator checks each of the above names to see if they likely represent a time value. If a name does look like a time value, the shape or targeted shape MUST be a timestamp shape.

A name is assumed to represent a time value if it:

Begins or ends with the word "time"
Begins or ends with the word "date"
Ends with the word "at"
Ends with the word "on"
Contains the exact string "timestamp" or "Timestamp"

For the purpose of this validator, words are matched case insensitively. Words are separated by either an underscore character, or by mixed case characters. For example, "FooBar", "fooBar", "foo_bar", "Foo_Bar", and "FOO_BAR" all contain the same two words, "foo" and "bar".

Rationale

Smithy tooling can convert timestamp shapes into idiomatic language types that make them easier to work with in client tooling.

Default severity

DANGER

Configuration

Property	Type	Description
additionalPatterns	[ `string` ]	A list of regular expression patterns that identify names that represent time.

MissingClientOptionalTrait#

Allows services to control backward compatibility guarantees for members marked as @required and @default by requiring the application of the @clientOptional trait.

Rationale

Different service providers have different backward compatibility guarantees for @required and @default structure members. Some services wish to reserve the right to remove the @required trait at any time, while others are able to strictly follow the backward-compatibility guarantees of the @required trait. For example, it is considered backward compatible to remove the @required trait from a member and replace it with the @default trait. However, this isn't possible for members that target structure or union shapes because they can have no default value. The risk associated with such members may be unacceptable for some services.

Default severity

DANGER

Configuration

Property	Type	Description
onRequiredOrDefault	`boolean`	Requires that members marked with the `@required` or `@default` trait are also marked with the `@clientOptional` trait.
onRequiredStructureOrUnion	`boolean`	Requires that `@required` members that target structure or union shapes are also marked with the `@clientOptional` trait. `@required` members that target structures and unions are risky because there is no backward compatible way to replace the `@required` trait with the `@default` trait if the member ever needs to be made optional.

The following example requires that @required members that target a structure or union are marked with the @clientOptional trait.

$version: "2"

metadata validators = [
    {
        name: "MissingClientOptionalTrait",

        // Limit validation to a specific set of namespaces.
        namespaces: ["smithy.example"],

        configuration: {
            onRequiredStructureOrUnion: true
        }
    }
]

This validation can be suppressed for any member that the service provider decides is not at risk of ever needing to become optional in the future:

structure Sprocket {
    @required
    @suppress(["MissingClientOptionalTrait"])
    owner: OwnerStructure
}

Writing custom validators#

Custom validators can be written in Java to apply more advanced model validation. Writing a custom validator involves writing an implementation of a Smithy validator in Java, creating a JAR, and making the JAR available on the classpath.

Custom validators are implementations of the software.amazon.smithy.model.validation.Validator interface. Most validators should extend from software.amazon.smithy.model.validation.AbstractValidator.

The following linter emits a ValidationEvent for every shape in the model that is not documented.

package com.example.mypackage;

import java.util.List;
import java.util.stream.Collectors;
import software.amazon.smithy.model.Model;
import software.amazon.smithy.model.traits.DocumentationTrait;
import software.amazon.smithy.model.validation.AbstractValidator;
import software.amazon.smithy.model.validation.ValidationEvent;

public class DocumentationValidator extends AbstractValidator {
    @Override
    public List<ValidationEvent> validate(Model model) {
        return model.shapes()
                .filter(shape -> !shape.hasTrait(DocumentationTrait.class))
                .map(shape -> error(shape, "This shape is not documented!"))
                .collect(Collectors.toList());
    }
}

Validators need to be registered as Java service providers. Add the following class name to a file named software.amazon.smithy.model.validation.Validator found in the src/main/resources/META-INF/services directory of a standard Gradle Java package:

com.example.mypackage.DocumentationValidator

When added to the classpath (typically as a dependency of a published JAR), the custom validator is automatically applied to a model each time the model is loaded.

Writing custom Linters#

Like custom validators, custom linters can be written in Java to apply more advanced model validation.

Custom linters are implementations of the software.amazon.smithy.model.validation.Validator interface. Because linters are configurable, they are created using an implementation of the software.amazon.smithy.model.validation.ValidatorService interface.

The following validator emits a ValidationEvent for every shape in the model that has documentation that contains a forbidden string.

package com.example.mypackage;

import java.util.List;
import java.util.Optional;
import java.util.stream.Collectors;
import java.util.stream.Stream;
import software.amazon.smithy.model.Model;
import software.amazon.smithy.model.node.NodeMapper;
import software.amazon.smithy.model.shapes.Shape;
import software.amazon.smithy.model.traits.DocumentationTrait;
import software.amazon.smithy.model.validation.AbstractValidator;
import software.amazon.smithy.model.validation.ValidationEvent;
import software.amazon.smithy.model.validation.ValidatorService;

public class ForbiddenDocumentationValidator extends AbstractValidator {

    /**
     * ForbiddenDocumentation configuration settings.
     */
    public static final class Config {
        private List<String> forbid;

        public List<String> getForbid() {
            return forbid;
        }

        public void setForbid(List<String> forbid) {
            this.forbid = forbid;
        }
    }

    // Does the actual work of converting metadata found in a Smithy
    // model into an actual implementation of a Validator.
    public static final class Provider extends ValidatorService.Provider {
        public Provider() {
            super(ForbiddenDocumentationValidator.class, configuration -> {
                // Deserialize the Node value into the Config POJO.
                NodeMapper mapper = new NodeMapper();
                ForbiddenDocumentationValidator.Config config = mapper.deserialize(configuration, Config.class);
                return new ForbiddenDocumentationValidator(config);
            });
        }
    }

    private final List<String> forbid;

    // The constructor is private since the validator is only intended to
    // be created when loading a model via the Provider class.
    private ForbiddenDocumentationValidator(Config config) {
        this.forbid = config.forbid;
    }

    @Override
    public List<ValidationEvent> validate(Model model) {
        // Find every shape that violates the linter and return a list
        // of ValidationEvents.
        return model.shapes()
                .filter(shape -> shape.hasTrait(DocumentationTrait.class))
                .flatMap(shape -> validateShape(shape).map(Stream::of).orElseGet(Stream::empty))
                .collect(Collectors.toList());
    }

    private Optional<ValidationEvent> validateShape(Shape shape) {
        // Grab the trait by type.
        DocumentationTrait trait = shape.expectTrait(DocumentationTrait.class);
        String docString = trait.getValue();

        for (String text : forbid) {
            if (docString.contains(text)) {
                // Emit an event that points at the location of the trait
                // and associates the warning with the shape.
                return Optional.of(warning(shape, trait, "Documentation uses forbidden text: " + text));
            }
        }

        return Optional.empty();
    }
}

Configurable linters need to be registered as Java service providers. Add the following class name to a file named software.amazon.smithy.model.validation.ValidatorService found in the src/main/resources/META-INF/services directory of a standard Gradle Java package:

com.example.mypackage.ForbiddenDocumentationValidator$Provider

When added to the classpath (typically as a dependency of a published JAR), the custom validator is available to be used as a validator. The following example warns each time the word "meow" appears in documentation:

$version: "2"

metadata validators = [
    {
        name: "ForbiddenDocumentation"
        configuration: {
            forbid: ["meow"]
        }
    }
]

Tip

The EmitEachSelector can get you pretty far without needing to write any Java code. For example, the above linter can be implemented using the following Smithy model:

$version: "2"

metadata validators = [
    {
        name: "EmitEachSelector"
        id: "ForbiddenDocumentation"
        message: "Documentation uses forbidden text"
        configuration: {
            selector: "[trait|documentation*='meow']"
        }
    }
]

Linting Models#

Linting overview#

Linters in smithy-linters#

UnreferencedShape#

AbbreviationName#

CamelCase#

MissingSensitiveTrait#

NoninclusiveTerms#

ReservedWords#

Wildcards in ReservedWords#

Words boundary matching#

StandardOperationVerb#

RepeatedShapeName#

InputOutputStructureReuse#

MissingPaginatedTrait#

ShouldHaveUsedTimestamp#

MissingClientOptionalTrait#

Writing custom validators#

Writing custom Linters#

Linters in `smithy-linters`#