JSON:API — Latest Specification (v1.1) (original) (raw)

Status

This page presents the latest published version of JSON:API, which is currently version 1.1. New versions of JSON:API will always be backwards compatible using a never remove, only add strategy. Additions can be proposed in our discussion forum.

If you catch an error in the specification’s text, or if you write an implementation, please let us know by opening an issue or pull request at ourGitHub repository.

Introduction

JSON:API is a specification for how a client should request that resources be fetched or modified, and how a server should respond to those requests. JSON:API can be easily extended with extensions and profiles.

JSON:API is designed to minimize both the number of requests and the amount of data transmitted between clients and servers. This efficiency is achieved without compromising readability, flexibility, or discoverability.

JSON:API requires use of the JSON:API media type (application/vnd.api+json) for exchanging data.

Semantics

All document members, query parameters, and processing rules defined by this specification are collectively called “specification semantics”.

Certain document members, query parameters, and processing rules are reserved for implementors to define at their discretion. These are called “implementation semantics”.

All other semantics are reserved for potential future use by this specification.

Conventions

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described inBCP 14[RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.

The JSON:API media type isapplication/vnd.api+json.

Media Type Parameters

The JSON:API specification supports two media type parameters: ext andprofile, which are used to specify extensions and profiles, respectively.

Note: A media type parameter is an extra piece of information that can accompany a media type. For example, in the headerContent-Type: text/html;charset="utf-8", the media type is text/html andcharset is a parameter.

Extensions

Extensions provide a means to “extend” the base specification by defining additional specification semantics.

Extensions cannot alter or remove specification semantics, nor can they specify implementation semantics.

Profiles

Profiles provide a means to share a particular usage of the specification among implementations.

Profiles can specify implementation semantics, but cannot alter, add to, or remove specification semantics.

Rules for Media Type Parameters

The JSON:API media type MUST NOT be specified with any media type parameters other than ext and profile. The ext parameter is used to supportextensions and the profile parameter is used to support profiles.

Extensions and profiles are each uniquely identified by aURI. Visiting an extension’s or a profile’s URI SHOULD return documentation that describes its usage. The values of the ext and profile parameters MUST equal a space-separated (U+0020 SPACE, “ “) list of extension or profile URIs, respectively.

Note: When serializing the ext or profile media type parameters, the HTTP specification requires that parameter values be surrounded by quotation marks (U+0022 QUOTATION MARK, “"”).

Rules for Extensions

An extension MAY impose additional processing rules or further restrictions and it MAY define new object members as described below.

An extension MUST NOT lessen or remove any processing rules, restrictions or object member requirements defined in this specification or other extensions.

An extension MAY define new members within the document structure defined by this specification. The rules for extension member names are coveredbelow.

An extension MAY define new query parameters. The rules for extension-defined query parameters are covered below.

When an extension defines new query parameters or document members, the extension MUST define a namespace to guarantee that extensions will never conflict with current or future versions of this specification. A namespaceMUST meet all of the following conditions:

An extension MUST NOT define more than one namespace. The namespace used for all query parameters and document members MUST be the same for any given extension.

In the following example, an extension with the namespace version has specified a resource object member version:id to support per-resource versioning. This member might appear as follows:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json;ext="https://jsonapi.org/ext/version"

// ...
{
  "type": "articles",
  "id": "1",
  "version:id": "42",
  "attributes": {
    "title": "Rails is Omakase"
  }
}
// ...

Rules for Profiles

The rules for profile usage are dictated by RFC 6906.

A profile MAY define document members and processing rules that are reserved for implementors.

A profile MUST NOT define any query parameters except implementation-specific query parameters.

A profile MUST NOT alter or remove processing rules that have been defined by this specification or by an extension. However, a profileMAY define processing rules for query parameters whose processing rules have been reserved for implementors to define at their discretion.

For example, a profile could define rules for interpreting the filter query parameter, but it could not specify that relationship names in the include query parameter are space-separated instead of dot-separated.

Unlike extensions, profiles do not need to define a namespace for document members because profiles cannot define specification semantics and thus cannot conflict with current or future versions of this specification. However, it is possible for profiles to conflict with other profiles. Therefore, it is the responsibility of implementors to ensure that they do not support conflicting profiles.

In the following example, a profile has defined a timestamps attribute. According to the profile, the attribute must be an object containing a created member and a modified member and these members’ values must use the RFC 3339 format. With such a profile applied, a response might appear as follows:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json;profile="https://example.com/resource-timestamps"

// ...
{
  "type": "articles",
  "id": "1",
  "attributes": {
    "title": "Rails is Omakase",
    "timestamps": {
      "created": "2020-07-21T12:09:00Z",
      "modified": "2020-07-30T10:19:01Z"
    }
  }
}
// ...

Content Negotiation

Universal Responsibilities

Clients and servers MUST send all JSON:API payloads using the JSON:API media type in the Content-Type header.

Clients and servers MUST specify the ext media type parameter in theContent-Type header when they have applied one or more extensions to a JSON:API document.

Clients and servers MUST specify the profile media type parameters in theContent-Type header when they have applied one or more profiles to a JSON:API document.

Client Responsibilities

When processing a JSON:API response document, clients MUST ignore any parameters other than ext and profile parameters in the server’sContent-Type header.

A client MAY use the ext media type parameter in an Accept header to require that a server apply all the specified extensions to the response document.

A client MAY use the profile media type parameter in an Accept header to request that the server apply one or more profiles to the response document.

Note: A client is allowed to send more than one acceptable media type in theAccept header, including multiple instances of the JSON:API media type. This allows clients to request different combinations of the ext and profilemedia type parameters. A client can use quality valuesto indicate that some combinations are less preferable than others. Media types specified without a qvalue are equally preferable to each other, regardless of their order, and are always considered more preferable than a media type with a qvalue less than 1.

Server Responsibilities

If a request specifies the Content-Type header with the JSON:API media type, servers MUST respond with a 415 Unsupported Media Type status code if that media type contains any media type parameters other than ext orprofile.

If a request specifies the Content-Type header with an instance of the JSON:API media type modified by the ext media type parameter and that parameter contains an unsupported extension URI, the server MUST respond with a 415 Unsupported Media Type status code.

Note: JSON:API servers that do not support version 1.1 of this specification will respond with a 415 Unsupported Media Type client error if the ext orprofile media type parameter is present.

If a request’s Accept header contains an instance of the JSON:API media type, servers MUST ignore instances of that media type which are modified by a media type parameter other than ext or profile. If all instances of that media type are modified with a media type parameter other than ext or profile, servers MUST respond with a 406 Not Acceptable status code. If every instance of that media type is modified by the ext parameter and each contains at least one unsupported extension URI, the server MUST also respond with a406 Not Acceptable.

If the profile parameter is received, a server SHOULD attempt to apply any requested profile(s) to its response. A server MUST ignore any profiles that it does not recognize.

Note: The above rules guarantee strict agreement on extensions between the client and server, while the application of profiles is left to the discretion of the server.

Servers that support the ext or profile media type parameters SHOULDspecify the Vary header with Accept as one of its values. This applies to responses with and without any profiles or extensions applied.

Note: Some HTTP intermediaries (e.g. CDNs) may ignore the Vary header unless specifically configured to respect it.

Document Structure

This section describes the structure of a JSON:API document, which is identified by the JSON:API media type. JSON:API documents are defined in JavaScript Object Notation (JSON) [RFC8259].

Although the same media type is used for both request and response documents, certain aspects are only applicable to one or the other. These differences are called out below.

Extensions MAY define new members within the document structure. These members MUST comply with the naming requirements specifiedbelow.

Unless otherwise noted, objects defined by this specification or any applied extensions MUST NOT contain any additional members. Client and server implementations MUST ignore non-compliant members.

Note: These conditions allow this specification to evolve through additive changes.

Top Level

A JSON object MUST be at the root of every JSON:API request and response document containing data. This object defines a document’s “top level”.

A document MUST contain at least one of the following top-level members:

The members data and errors MUST NOT coexist in the same document.

A document MAY contain any of these top-level members:

If a document does not contain a top-level data key, the included memberMUST NOT be present either.

The top-level links object MAY contain the following members:

Note: The self link in the top-level links object allows a client to refresh the data represented by the current response document. The client should be able to use the provided link without applying any additional information. Therefore the link must contain the query parameters provided by the client to generate the response document. This includes but is not limited to query parameters used for inclusion of related resources,sparse fieldsets, sorting,pagination and filtering.

The document’s “primary data” is a representation of the resource or collection of resources targeted by a request.

Primary data MUST be either:

For example, the following primary data is a single resource object:

{
  "data": {
    "type": "articles",
    "id": "1",
    "attributes": {
      // ... this article's attributes
    },
    "relationships": {
      // ... this article's relationships
    }
  }
}

The following primary data is a single resource identifier object that references the same resource:

{
  "data": {
    "type": "articles",
    "id": "1"
  }
}

A logical collection of resources MUST be represented as an array, even if it only contains one item or is empty.

Resource Objects

“Resource objects” appear in a JSON:API document to represent resources.

A resource object MUST contain at least the following top-level members:

Exception: The id member is not required when the resource object originates at the client and represents a new resource to be created on the server. In that case, a client MAY include a lid member to uniquely identify the resource by type locally within the document.

In addition, a resource object MAY contain any of these top-level members:

Here’s how an article (i.e. a resource of type “articles”) might appear in a document:

// ...
{
  "type": "articles",
  "id": "1",
  "attributes": {
    "title": "Rails is Omakase"
  },
  "relationships": {
    "author": {
      "links": {
        "self": "/articles/1/relationships/author",
        "related": "/articles/1/author"
      },
      "data": { "type": "people", "id": "9" }
    }
  }
}
// ...

Identification

As noted above, every resource object MUST contain atype member. Every resource object MUST also contain an id member, except when the resource object originates at the client and represents a new resource to be created on the server. If id is omitted due to this exception, a lid member MAY be included to uniquely identify the resource by type locally within the document. The value of the lid member MUST be identical for every representation of the resource in the document, includingresource identifier objects.

The values of the id, type, and lid members MUST be strings.

Within a given API, each resource object’s type and id pair MUSTidentify a single, unique resource. (The set of URIs controlled by a server, or multiple servers acting as one, constitute an API.)

The type member is used to describe resource objects that share common attributes and relationships.

The values of type members MUST adhere to the same constraints asmember names.

Note: This spec is agnostic about inflection rules, so the value of typecan be either plural or singular. However, the same value should be used consistently throughout an implementation.

Fields

A resource object’s attributes and its relationships are collectively called its “fields”.

Fields for a resource object MUST share a common namespace with each other and with type and id. In other words, a resource can not have an attribute and relationship with the same name, nor can it have an attribute or relationship named type or id.

Attributes

The value of the attributes key MUST be an object (an “attributes object”). Members of the attributes object (“attributes”) represent information about the resource object in which it’s defined.

Attributes may contain any valid JSON value, including complex data structures involving JSON objects and arrays.

Keys that reference related resources (e.g. author_id) SHOULD NOT appear as attributes. Instead, relationships SHOULD be used.

Note: See fields and member names for more restrictions on this container.

Relationships

The value of the relationships key MUST be an object (a “relationships object”). Each member of a relationships object represents a “relationship” from the resource objectin which it has been defined to other resource objects.

Relationships may be to-one or to-many.

A relationship’s name is given by its key. The value at that key MUST be an object (“relationship object”).

A “relationship object” MUST contain at least one of the following:

A relationship object that represents a to-many relationship MAY also containpagination links under the links member, as described below. Anypagination links in a relationship object MUST paginate the relationship data, not the related resources.

Note: See fields and member names for more restrictions on this container.

A “related resource link” provides access to resource objects linkedin a relationship. When fetched, the related resource object(s) are returned as the response’s primary data.

For example, an article’s comments relationship could specify a link that returns a collection of comment resource objectswhen retrieved through a GET request.

If present, a related resource link MUST reference a valid URL, even if the relationship isn’t currently associated with any target resources. Additionally, a related resource link MUST NOT change because its relationship’s content changes.

Resource Linkage

Resource linkage in a compound document allows a client to link together all of the included resource objects without having to GET any URLs via links.

Resource linkage MUST be represented as one of the following:

Note: The spec does not impart meaning to order of resource identifier objects in linkage arrays of to-many relationships, although implementations may do that. Arrays of resource identifier objects may represent ordered or unordered relationships, and both types can be mixed in one response object.

For example, the following article is associated with an author:

// ...
{
  "type": "articles",
  "id": "1",
  "attributes": {
    "title": "Rails is Omakase"
  },
  "relationships": {
    "author": {
      "links": {
        "self": "http://example.com/articles/1/relationships/author",
        "related": "http://example.com/articles/1/author"
      },
      "data": { "type": "people", "id": "9" }
    }
  },
  "links": {
    "self": "http://example.com/articles/1"
  }
}
// ...

The author relationship includes a link for the relationship itself (which allows the client to change the related author directly), a related resource link to fetch the resource objects, and linkage information.

The optional links member within each resource object contains linksrelated to the resource.

If present, this links object MAY contain a self link that identifies the resource represented by the resource object.

// ...
{
  "type": "articles",
  "id": "1",
  "attributes": {
    "title": "Rails is Omakase"
  },
  "links": {
    "self": "http://example.com/articles/1"
  }
}
// ...

A server MUST respond to a GET request to the specified URL with a response that includes the resource as the primary data.

Resource Identifier Objects

A “resource identifier object” is an object that identifies an individual resource.

A “resource identifier object” MUST contain a type member. It MUSTalso contain an id member, except when it represents a new resource to be created on the server. In this case, a lid member MUST be included that identifies the new resource.

The values of the id, type, and lid members MUST be strings.

A “resource identifier object” MAY also include a meta member, whose value is a meta object that contains non-standard meta-information.

Compound Documents

Servers MAY allow responses that include related resources along with the requested primary resources. Such responses are called “compound documents”.

In a compound document, all included resources MUST be represented as an array of resource objects in a top-level included member.

Every included resource object MUST be identified via a chain of relationships originating in a document’s primary data. This means that compound documents require “full linkage” and that no resource object can be included without a direct or indirect relationship to the document’s primary data.

The only exception to the full linkage requirement is when relationship fields that would otherwise contain linkage data are excluded due tosparse fieldsets requested by the client.

A complete example document with multiple included relationships:

{
  "data": [{
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "JSON:API paints my bikeshed!"
    },
    "links": {
      "self": "http://example.com/articles/1"
    },
    "relationships": {
      "author": {
        "links": {
          "self": "http://example.com/articles/1/relationships/author",
          "related": "http://example.com/articles/1/author"
        },
        "data": { "type": "people", "id": "9" }
      },
      "comments": {
        "links": {
          "self": "http://example.com/articles/1/relationships/comments",
          "related": "http://example.com/articles/1/comments"
        },
        "data": [
          { "type": "comments", "id": "5" },
          { "type": "comments", "id": "12" }
        ]
      }
    }
  }],
  "included": [{
    "type": "people",
    "id": "9",
    "attributes": {
      "firstName": "Dan",
      "lastName": "Gebhardt",
      "twitter": "dgeb"
    },
    "links": {
      "self": "http://example.com/people/9"
    }
  }, {
    "type": "comments",
    "id": "5",
    "attributes": {
      "body": "First!"
    },
    "relationships": {
      "author": {
        "data": { "type": "people", "id": "2" }
      }
    },
    "links": {
      "self": "http://example.com/comments/5"
    }
  }, {
    "type": "comments",
    "id": "12",
    "attributes": {
      "body": "I like XML better"
    },
    "relationships": {
      "author": {
        "data": { "type": "people", "id": "9" }
      }
    },
    "links": {
      "self": "http://example.com/comments/12"
    }
  }]
}

A compound document MUST NOT include more than one resource object for each type and id pair.

Note: In a single document, you can think of the type and id as a composite key that uniquely references resource objects in another part of the document.

Note: For resources that do not contain an id member but do contain a lid, the lid is sufficient to establish resource identity and thus linkage between resource objects and resource identifier objects throughout the document.

Note: This approach ensures that a single canonical resource object is returned with each response, even when the same resource is referenced multiple times.

Meta Information

Where specified, a meta member can be used to include non-standard meta-information. The value of each meta member MUST be an object (a “meta object”).

Any members MAY be specified within meta objects.

For example:

{
  "meta": {
    "copyright": "Copyright 2015 Example Corp.",
    "authors": [
      "Yehuda Katz",
      "Steve Klabnik",
      "Dan Gebhardt",
      "Tyler Kellen"
    ]
  },
  "data": {
    // ...
  }
}

Where specified, a links member can be used to represent links. The value of this member MUST be an object (a “links object”).

Within this object, a link MUST be represented as either:

A link’s relation type SHOULD be inferred from the name of the link unless the link is a link object and the link object has a rel member.

A link’s context is the top-level object, resource object, or relationship object in which it appears.

In the example below, the self link is a string whereas the related link is a link object. The related link object provides additional information about the targeted related resource collection as well as a schema that serves as a description document for that collection:

"links": {
  "self": "http://example.com/articles/1/relationships/comments",
  "related": {
    "href": "http://example.com/articles/1/comments",
    "title": "Comments",
    "describedby": "http://example.com/schemas/article-comments",
    "meta": {
      "count": 10
    }
  }
}

A “link object” is an object that represents a web link.

A link object MUST contain the following member:

A link object MAY also contain any of the following members:

Note: the type and hreflang members are only hints; the target resource is not guaranteed to be available in the indicated media type or language when the link is actually followed.

JSON:API Object

A JSON:API document MAY include information about its implementation under a top level jsonapi member. If present, the value of the jsonapimember MUST be an object (a “jsonapi object”).

The jsonapi object MAY contain any of the following members:

Clients and servers MUST NOT use an ext or profile member for content negotiation. Content negotiation MUST only happen based on media type parameters in Content-Type header.

A simple example appears below:

{
  "jsonapi": {
    "version": "1.1",
    "ext": [
      "https://jsonapi.org/ext/atomic"
    ],
    "profile": [
      "http://example.com/profiles/flexible-pagination",
      "http://example.com/profiles/resource-versioning"
    ]
  }
}

If the version member is not present, clients should assume the server implements at least version 1.0 of the specification.

Note: Because JSON:API is committed to making additive changes only, the version string primarily indicates which new features a server may support.

Member Names

Implementation and profile defined member names used in a JSON:API documentMUST be treated as case sensitive by clients and servers, and they MUSTmeet all of the following conditions:

To enable an easy mapping of member names to URLs, it is RECOMMENDED that member names use only non-reserved, URL safe characters specified in RFC 3986.

Allowed Characters

The following “globally allowed characters” MAY be used anywhere in a member name:

Additionally, the following characters are allowed in member names, except as the first or last character:

Reserved Characters

The following characters MUST NOT be used in implementation andprofile defined member names:

@-Members

Member names MAY also begin with an at sign (U+0040 COMMERCIAL AT, “@”). Members named this way are called “@-Members”. @-Members MAY appear anywhere in a document.

This specification provides no guidance on the meaning or usage of @-Members, which are considered to be implementation semantics. @-MembersMUST be ignored when interpreting this specification’s definitions and processing instructions given outside of this subsection. For example, anattribute is defined above as any member of the attributes object. However, because @-Members must be ignored when interpreting that definition, an @-Member that occurs in an attributes object is not an attribute.

Note: Among other things, “@” members can be used to add JSON-LD data to a JSON:API document. Such documents should be served with an extra headerto convey to JSON-LD clients that they contain JSON-LD data.

Extension Members

The name of every new member introduced by an extension MUST be prefixed with the extension’s namespace followed by a colon (:). The remainder of the name MUST adhere to the rules for implementation defined member names.

Fetching Data

Data, including resources and relationships, can be fetched by sending aGET request to an endpoint.

Responses can be further refined with the optional features described below.

Fetching Resources

A server MUST support fetching resource data for every URL provided as:

For example, the following request fetches a collection of articles:

GET /articles HTTP/1.1
Accept: application/vnd.api+json

The following request fetches an article:

GET /articles/1 HTTP/1.1
Accept: application/vnd.api+json

And the following request fetches an article’s author:

GET /articles/1/author HTTP/1.1
Accept: application/vnd.api+json

Responses

200 OK

A server MUST respond to a successful request to fetch an individual resource or resource collection with a 200 OK response.

A server MUST respond to a successful request to fetch a resource collection with an array of resource objects or an empty array ([]) as the response document’s primary data.

For example, a GET request to a collection of articles could return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "http://example.com/articles"
  },
  "data": [{
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "JSON:API paints my bikeshed!"
    }
  }, {
    "type": "articles",
    "id": "2",
    "attributes": {
      "title": "Rails is Omakase"
    }
  }]
}

A similar response representing an empty collection would be:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "http://example.com/articles"
  },
  "data": []
}

A server MUST respond to a successful request to fetch an individual resource with a resource object or null provided as the response document’s primary data.

null is only an appropriate response when the requested URL is one that might correspond to a single resource, but doesn’t currently.

Note: Consider, for example, a request to fetch a to-one related resource link. This request would respond with null when the relationship is empty (such that the link is corresponding to no resources) but with the single related resource’sresource object otherwise.

For example, a GET request to an individual article could return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "http://example.com/articles/1"
  },
  "data": {
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "JSON:API paints my bikeshed!"
    },
    "relationships": {
      "author": {
        "links": {
          "related": "http://example.com/articles/1/author"
        }
      }
    }
  }
}

If the above article’s author is missing, then a GET request to that related resource would return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "http://example.com/articles/1/author"
  },
  "data": null
}
404 Not Found

A server MUST respond with 404 Not Found when processing a request to fetch a single resource that does not exist, except when the request warrants a200 OK response with null as the primary data (as described above).

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

Fetching Relationships

A server MUST support fetching relationship data for every relationship URL provided as a self link as part of a relationship’s links object.

For example, the following request fetches data about an article’s comments:

GET /articles/1/relationships/comments HTTP/1.1
Accept: application/vnd.api+json

And the following request fetches data about an article’s author:

GET /articles/1/relationships/author HTTP/1.1
Accept: application/vnd.api+json

Responses

200 OK

A server MUST respond to a successful request to fetch a relationship with a 200 OK response.

The primary data in the response document MUST match the appropriate value for resource linkage, as described above forrelationship objects.

The top-level links object MAY contain self and related links, as described above for relationship objects.

For example, a GET request to a URL from a to-one relationship link could return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "/articles/1/relationships/author",
    "related": "/articles/1/author"
  },
  "data": {
    "type": "people",
    "id": "12"
  }
}

If the above relationship is empty, then a GET request to the same URL would return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "/articles/1/relationships/author",
    "related": "/articles/1/author"
  },
  "data": null
}

A GET request to a URL from a to-many relationship link could return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "/articles/1/relationships/tags",
    "related": "/articles/1/tags"
  },
  "data": [
    { "type": "tags", "id": "2" },
    { "type": "tags", "id": "3" }
  ]
}

If the above relationship is empty, then a GET request to the same URL would return:

HTTP/1.1 200 OK
Content-Type: application/vnd.api+json

{
  "links": {
    "self": "/articles/1/relationships/tags",
    "related": "/articles/1/tags"
  },
  "data": []
}
404 Not Found

A server MUST return 404 Not Found when processing a request to fetch a relationship link URL that does not exist.

Note: This can happen when the parent resource of the relationship does not exist. For example, when /articles/1 does not exist, request to/articles/1/relationships/tags returns 404 Not Found.

If a relationship link URL exists but the relationship is empty, then200 OK MUST be returned, as described above.

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

An endpoint MAY return resources related to the primary data by default.

An endpoint MAY also support an include query parameter to allow the client to customize which related resources should be returned.

If an endpoint does not support the include parameter, it MUST respond with 400 Bad Request to any requests that include it.

If an endpoint supports the include parameter and a client supplies it:

The value of the include parameter MUST be a comma-separated (U+002C COMMA, “,”) list of relationship paths. A relationship path is a dot-separated (U+002E FULL-STOP, “.”) list of relationship names. An empty value indicates that no related resources should be returned.

If a server is unable to identify a relationship path or does not support inclusion of resources from a path, it MUST respond with 400 Bad Request.

Note: For example, a relationship path could be comments.author, wherecomments is a relationship listed under a articles resource object, andauthor is a relationship listed under a comments resource object.

For instance, comments could be requested with an article:

GET /articles/1?include=comments HTTP/1.1
Accept: application/vnd.api+json

In order to request resources related to other resources, a dot-separated path for each relationship name can be specified:

GET /articles/1?include=comments.author HTTP/1.1
Accept: application/vnd.api+json

Note: Because compound documents require full linkage (except when relationship linkage is excluded by sparse fieldsets), intermediate resources in a multi-part path must be returned along with the leaf nodes. For example, a response to a request for comments.author should include commentsas well as the author of each of those comments.

Note: A server may choose to expose a deeply nested relationship such ascomments.author as a direct relationship with an alternative name such ascommentAuthors. This would allow a client to request/articles/1?include=commentAuthors instead of/articles/1?include=comments.author. By exposing the nested relationship with an alternative name, the server can still provide full linkage in compound documents without including potentially unwanted intermediate resources.

Multiple related resources can be requested in a comma-separated list:

GET /articles/1?include=comments.author,ratings HTTP/1.1
Accept: application/vnd.api+json

Furthermore, related resources can be requested from a relationship endpoint:

GET /articles/1/relationships/comments?include=comments.author HTTP/1.1
Accept: application/vnd.api+json

In this case, the primary data would be a collection ofresource identifier objects that represent linkage to comments for an article, while the full comments and comment authors would be returned as included data.

Note: This section applies to any endpoint that responds with primary data, regardless of the request type. For instance, a server could support the inclusion of related resources along with a POST request to create a resource or relationship.

Sparse Fieldsets

A client MAY request that an endpoint return only specific fields in the response on a per-type basis by including a fields[TYPE] query parameter.

The value of any fields[TYPE] parameter MUST be a comma-separated (U+002C COMMA, “,”) list that refers to the name(s) of the fields to be returned. An empty value indicates that no fields should be returned.

If a client requests a restricted set of fields for a given resource type, an endpoint MUST NOT include additional fields in resource objects of that type in its response.

If a client does not specify the set of fields for a given resource type, the server MAY send all fields, a subset of fields, or no fields for that resource type.

GET /articles?include=author&fields[articles]=title,body&fields[people]=name HTTP/1.1
Accept: application/vnd.api+json

Note: The above example URI shows unencoded [ and ] characters simply for readability. In practice, these characters should be percent-encoded. See “Square Brackets in Parameter Names”.

Note: This section applies to any endpoint that responds with resources as primary or included data, regardless of the request type. For instance, a server could support sparse fieldsets along with a POST request to create a resource.

Sorting

A server MAY choose to support requests to sort resource collections according to one or more criteria (“sort fields”).

Note: Although recommended, sort fields do not necessarily need to correspond to resource attribute and relationship names.

Note: It is recommended that dot-separated (U+002E FULL-STOP, “.”) sort fields be used to request sorting based upon relationship attributes. For example, a sort field of author.name could be used to request that the primary data be sorted based upon the name attribute of the authorrelationship.

An endpoint MAY support requests to sort the primary data with a sortquery parameter. The value for sort MUST represent sort fields.

GET /people?sort=age HTTP/1.1
Accept: application/vnd.api+json

An endpoint MAY support multiple sort fields by allowing comma-separated (U+002C COMMA, “,”) sort fields. Sort fields SHOULD be applied in the order specified.

GET /people?sort=age,name HTTP/1.1
Accept: application/vnd.api+json

The sort order for each sort field MUST be ascending unless it is prefixed with a minus (U+002D HYPHEN-MINUS, “-“), in which case it MUST be descending.

GET /articles?sort=-created,title HTTP/1.1
Accept: application/vnd.api+json

The above example should return the newest articles first. Any articles created on the same date will then be sorted by their title in ascending alphabetical order.

If the server does not support sorting as specified in the query parametersort, it MUST return 400 Bad Request.

If sorting is supported by the server and requested by the client via query parameter sort, the server MUST return elements of the top-leveldata array of the response ordered according to the criteria specified. The server MAY apply default sorting rules to top-level data if request parameter sort is not specified.

Note: This section applies to any endpoint that responds with a resource collection as primary data, regardless of the request type.

A server MAY choose to limit the number of resources returned in a response to a subset (“page”) of the whole set available.

A server MAY provide links to traverse a paginated data set (“pagination links”).

Pagination links MUST appear in the links object that corresponds to a collection. To paginate the primary data, supply pagination links in the top-level links object. To paginate an included collection returned in a compound document, supply pagination links in the corresponding links object.

The following keys MUST be used for pagination links:

Keys MUST either be omitted or have a null value to indicate that a particular link is unavailable.

Concepts of order, as expressed in the naming of pagination links, MUSTremain consistent with JSON:API’s sorting rules.

The page query parameter family is reserved for pagination. Servers and clients SHOULD use these parameters for pagination operations.

Note: JSON API is agnostic about the pagination strategy used by a server, but the page query parameter family can be used regardless of the strategy employed. For example, a page-based strategy might use query parameters such as page[number] and page[size], while a cursor-based strategy might usepage[cursor].

Note: This section applies to any endpoint that responds with a resource collection as primary data, regardless of the request type.

Filtering

The filter query parameter family is reserved for filtering data. Servers and clients SHOULD use these parameters for filtering operations.

Note: JSON API is agnostic about the strategies supported by a server.

Creating, Updating and Deleting Resources

A server MAY allow resources of a given type to be created. It MAYalso allow existing resources to be modified or deleted.

A request MUST completely succeed or fail (in a single “transaction”). No partial updates are allowed.

Note: The type member is required in every resource object throughout requests and responses in JSON:API. There are some cases, such as when POSTing to an endpoint representing heterogeneous data, when the type could not be inferred from the endpoint. However, picking and choosing when it is required would be confusing; it would be hard to remember when it was required and when it was not. Therefore, to improve consistency and minimize confusion, type is always required.

Creating Resources

A resource can be created by sending a POST request to a URL that represents a collection of resources. The request MUST include a single resource objectas primary data. The resource object MUST contain at least a type member.

For instance, a new photo might be created with the following request:

POST /photos HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "photos",
    "attributes": {
      "title": "Ember Hamster",
      "src": "http://example.com/images/productivity.png"
    },
    "relationships": {
      "photographer": {
        "data": { "type": "people", "id": "9" }
      }
    }
  }
}

If a relationship is provided in the relationships member of theresource object, its value MUST be a relationship object with a datamember. The value of this key represents the linkage the new resource is to have.

Client-Generated IDs

A server MAY accept a client-generated ID along with a request to create a resource. An ID MUST be specified with an id key, the value of which MUST be a universally unique identifier. The client SHOULD use a properly generated and formatted UUID as described in RFC 4122 [RFC4122].

NOTE: In some use-cases, such as importing data from another source, it may be possible to use something other than a UUID that is still guaranteed to be globally unique. Do not use anything other than a UUID unless you are 100% confident that the strategy you are using indeed generates globally unique identifiers.

For example:

POST /photos HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "photos",
    "id": "550e8400-e29b-41d4-a716-446655440000",
    "attributes": {
      "title": "Ember Hamster",
      "src": "http://example.com/images/productivity.png"
    }
  }
}

A server MUST return 403 Forbidden in response to an unsupported request to create a resource with a client-generated ID.

Responses

201 Created

If the requested resource has been created successfully and the server changes the resource in any way (for example, by assigning an id), the server MUSTreturn a 201 Created response and a document that contains the resource as primary data.

The response SHOULD include a Location header identifying the location of the newly created resource, in order to comply with RFC 7231.

If the resource object returned by the response contains a self key in itslinks member and a Location header is provided, the value of the selfmember MUST match the value of the Location header.

HTTP/1.1 201 Created
Location: http://example.com/photos/550e8400-e29b-41d4-a716-446655440000
Content-Type: application/vnd.api+json

{
  "data": {
    "type": "photos",
    "id": "550e8400-e29b-41d4-a716-446655440000",
    "attributes": {
      "title": "Ember Hamster",
      "src": "http://example.com/images/productivity.png"
    },
    "links": {
      "self": "http://example.com/photos/550e8400-e29b-41d4-a716-446655440000"
    }
  }
}

A server MAY return a 201 Created response with a document that contains no primary data if the requested resource has been created successfully and the server does not change the resource in any way (for example, by assigning anid or createdAt attribute). Other top-level members, such as meta, could be included in the response document.

Note: Only servers that accept Client-Generated IDs can avoid assigning an id to a new resource.

202 Accepted

If a request to create a resource has been accepted for processing, but the processing has not been completed by the time the server responds, the server MUST return a 202 Accepted status code.

204 No Content

If the requested resource has been created successfully and the server does not change the resource in any way (for example, by assigning an id or createdAtattribute), the server MUST return either a 201 Created status code and response document (as described above) or a 204 No Content status code with no response document.

403 Forbidden

A server MAY return 403 Forbidden in response to an unsupported request to create a resource.

404 Not Found

A server MUST return 404 Not Found when processing a request that references a related resource that does not exist.

409 Conflict

A server MUST return 409 Conflict when processing a POST request to create a resource with a client-generated ID that already exists.

A server MUST return 409 Conflict when processing a POST request in which the resource object’s type is not among the type(s) that constitute the collection represented by the endpoint.

A server SHOULD include error details and provide enough information to recognize the source of the conflict.

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

Updating Resources

A resource can be updated by sending a PATCH request to the URL that represents the resource.

The URL for a resource can be obtained in the self link of the resource object. Alternatively, when a GET request returns a single resource object as primary data, the same request URL can be used for updates.

The PATCH request MUST include a single resource object as primary data. The resource object MUST contain type and id members.

For example:

PATCH /articles/1 HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "To TDD or Not"
    }
  }
}

Updating a Resource’s Attributes

Any or all of a resource’s attributes MAY be included in the resource object included in a PATCH request.

If a request does not include all of the attributes for a resource, the serverMUST interpret the missing attributes as if they were included with their current values. The server MUST NOT interpret missing attributes as nullvalues.

For example, the following PATCH request is interpreted as a request to update only the title and text attributes of an article:

PATCH /articles/1 HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "To TDD or Not",
      "text": "TLDR; It's complicated... but check your test coverage regardless."
    }
  }
}

Updating a Resource’s Relationships

Any or all of a resource’s relationships MAY be included in the resource object included in a PATCH request.

If a request does not include all of the relationships for a resource, the serverMUST interpret the missing relationships as if they were included with their current values. It MUST NOT interpret them as null or empty values.

If a relationship is provided in the relationships member of a resource object in a PATCH request, its value MUST be a relationship object with a data member. The relationship’s value will be replaced with the value specified in this member.

For instance, the following PATCH request will update the author relationship of an article:

PATCH /articles/1 HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "articles",
    "id": "1",
    "relationships": {
      "author": {
        "data": { "type": "people", "id": "1" }
      }
    }
  }
}

Likewise, the following PATCH request performs a complete replacement of the tags for an article:

PATCH /articles/1 HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": {
    "type": "articles",
    "id": "1",
    "relationships": {
      "tags": {
        "data": [
          { "type": "tags", "id": "2" },
          { "type": "tags", "id": "3" }
        ]
      }
    }
  }
}

A server MAY reject an attempt to do a full replacement of a to-many relationship. In such a case, the server MUST reject the entire update, and return a 403 Forbidden response.

Note: Since full replacement may be a very dangerous operation, a server may choose to disallow it. For example, a server may reject full replacement if it has not provided the client with the full list of associated objects, and does not want to allow deletion of records the client has not seen.

Responses

200 OK

If a server accepts an update but also changes the targeted resource in ways other than those specified by the request (for example, updating theupdatedAt attribute or a computed sha), it MUST return a 200 OKresponse and a document that contains the updated resource as primary data.

A server MAY return a 200 OK response with a document that contains no primary data if an update is successful and the server does not change the targeted resource in ways other than those specified by the request. Other top-level members, such as meta, could be included in the response document.

202 Accepted

If an update request has been accepted for processing, but the processing has not been completed by the time the server responds, the server MUSTreturn a 202 Accepted status code.

204 No Content

If an update is successful and the server doesn’t change the targeted resource in ways other than those specified by the request, the serverMUST return either a 200 OK status code and response document (as described above) or a 204 No Content status code with no response document.

403 Forbidden

A server MUST return 403 Forbidden in response to an unsupported request to update a resource or relationship.

404 Not Found

A server MUST return 404 Not Found when processing a request to modify a resource that does not exist.

A server MUST return 404 Not Found when processing a request that references a related resource that does not exist.

409 Conflict

A server MAY return 409 Conflict when processing a PATCH request to update a resource if that update would violate other server-enforced constraints (such as a uniqueness constraint on a property other than id).

A server MUST return 409 Conflict when processing a PATCH request in which the resource object’s type or id do not match the server’s endpoint.

A server SHOULD include error details and provide enough information to recognize the source of the conflict.

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

Updating Relationships

Although relationships can be modified along with resources (as described above), JSON:API also supports updating of relationships independently at URLs from relationship links.

Note: Relationships are updated without exposing the underlying server semantics, such as foreign keys. Furthermore, relationships can be updated without necessarily affecting the related resources. For example, if an article has many authors, it is possible to remove one of the authors from the article without deleting the person itself. Similarly, if an article has many tags, it is possible to add or remove tags. Under the hood on the server, the first of these examples might be implemented with a foreign key, while the second could be implemented with a join table, but the JSON:API protocol would be the same in both cases.

Note: A server may choose to delete the underlying resource if a relationship is deleted (as a garbage collection measure).

Updating To-One Relationships

A to-one relationship can be updated by sending a PATCH request to a URL from a to-one relationship link.

The PATCH request MUST include a top-level member named data containing one of:

For example, the following request updates the author of an article:

PATCH /articles/1/relationships/author HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": { "type": "people", "id": "12" }
}

And the following request clears the author of the same article:

PATCH /articles/1/relationships/author HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": null
}

If the relationship is updated successfully then the server MUST return a successful response.

Updating To-Many Relationships

A to-many relationship can be updated by sending a PATCH, POST, orDELETE request to a URL from a to-many relationship link.

For all request types, the body MUST contain a data member whose value is an empty array or an array of resource identifier objects.

If a client makes a PATCH request to a URL from a to-manyrelationship link, the server MUST either completely replace every member of the relationship, return an appropriate error response if some resources cannot be found or accessed, or return a 403 Forbiddenresponse if complete replacement is not allowed by the server.

For example, the following request replaces every tag for an article:

PATCH /articles/1/relationships/tags HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": [
    { "type": "tags", "id": "2" },
    { "type": "tags", "id": "3" }
  ]
}

And the following request clears every tag for an article:

PATCH /articles/1/relationships/tags HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": []
}

If a client makes a POST request to a URL from arelationship link, the server MUST add the specified members to the relationship unless they are already present. If a given typeand id is already in the relationship, the server MUST NOT add it again.

Note: This matches the semantics of databases that use foreign keys for has-many relationships. Document-based storage should check the has-many relationship before appending to avoid duplicates.

If all of the specified resources can be added to, or are already present in, the relationship then the server MUST return a successful response.

Note: This approach ensures that a request is successful if the server’s state matches the requested state, and helps avoid pointless race conditions caused by multiple clients making the same changes to a relationship.

In the following example, the comment with ID 123 is added to the list of comments for the article with ID 1:

POST /articles/1/relationships/comments HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": [
    { "type": "comments", "id": "123" }
  ]
}

If the client makes a DELETE request to a URL from arelationship link the server MUST delete the specified members from the relationship or return a 403 Forbidden response. If all of the specified resources are able to be removed from, or are already missing from, the relationship then the server MUST return a successful response.

Note: As described above for POST requests, this approach helps avoid pointless race conditions between multiple clients making the same changes.

Relationship members are specified in the same way as in the POST request.

In the following example, comments with IDs of 12 and 13 are removed from the list of comments for the article with ID 1:

DELETE /articles/1/relationships/comments HTTP/1.1
Content-Type: application/vnd.api+json
Accept: application/vnd.api+json

{
  "data": [
    { "type": "comments", "id": "12" },
    { "type": "comments", "id": "13" }
  ]
}

Note: RFC 7231 specifies that a DELETE request may include a body, but that a server may reject the request. This spec defines the semantics of a server, and we are defining its semantics for JSON:API.

Responses

200 OK

If a server accepts an update but also changes the targeted relationship in other ways than those specified by the request, it MUST return a 200 OKresponse and a document that includes the updated relationship data as its primary data.

A server MAY return a 200 OK response with a document that contains no primary data if an update is successful and the server does not change the targeted relationship in ways other than those specified by the request. Other top-level members, such as meta, could be included in the response document.

202 Accepted

If a relationship update request has been accepted for processing, but the processing has not been completed by the time the server responds, the server MUST return a 202 Accepted status code.

204 No Content

If an update is successful and the server doesn’t change the targeted relationship in ways other than those specified by the request, the serverMUST return either a 200 OK status code and response document (as described above) or a 204 No Content status code with no response document.

Note: This is the appropriate response to a POST request sent to a URL from a to-many relationship link when that relationship already exists. It is also the appropriate response to a DELETE request sent to a URL from a to-many relationship link when that relationship does not exist.

403 Forbidden

A server MUST return 403 Forbidden in response to an unsupported request to update a relationship.

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

Deleting Resources

A resource can be deleted by sending a DELETE request to the URL that represents the resource:

DELETE /photos/1 HTTP/1.1
Accept: application/vnd.api+json

Responses

200 OK

A server MAY return a 200 OK response with a document that contains no primary data if a deletion request is successful. Other top-level members, such as meta, could be included in the response document.

202 Accepted

If a deletion request has been accepted for processing, but the processing has not been completed by the time the server responds, the server MUSTreturn a 202 Accepted status code.

204 No Content

If a deletion request is successful, the server MUST return either a 200 OK status code and response document (as described above) or a 204 No Contentstatus code with no response document.

404 NOT FOUND

A server SHOULD return a 404 Not Found status code if a deletion request fails due to the resource not existing.

Other Responses

A server MAY respond with other HTTP status codes.

A server MAY include error details with error responses.

A server MUST prepare responses, and a client MUST interpret responses, in accordance withHTTP semantics.

Query Parameters

Query Parameter Families

Although “query parameter” is a common term in everyday web development, it is not a well-standardized concept. Therefore, JSON:API provides its owndefinition of a query parameter.

For the most part, JSON:API’s definition coincides with colloquial usage, and its details can be safely ignored. However, one important consequence of this definition is that a URL like the following is considered to have two distinct query parameters:

/?page[offset]=0&page[limit]=10

The two parameters are named page[offset] and page[limit]; there is no single page parameter.

In practice, however, parameters like page[offset] and page[limit] are usually defined and processed together, and it’s convenient to refer to them collectively. Therefore, JSON:API introduces the concept of a query parameter family.

A “query parameter family” is the set of all query parameters whose name starts with a “base name”, followed by zero or more instances of empty square brackets (i.e. []), square-bracketed legal member names, or square-bracketed dot-separated lists of legal member names. The family is referred to by its base name.

For example, the filter query parameter family includes parameters named:filter, filter[x], filter[], filter[x][], filter[][], filter[x][y],filter[x.y], etc. However, filter[_] is not a valid parameter name in the family, because _ is not a valid member name.

Note: Dot separated lists of legal member names are intended to be used for relationship paths. For example, this allows filtering strategies using relationship paths as defined for sorting in query parameters such as GET /posts?sort=author.name&filter[author.status]=active.

Extension-Specific Query Parameters

The base name of every query parameter introduced by an extension MUST be prefixed with the extension’s namespace followed by a colon (:). The remainder of the base name MUST contain only the characters [a-z] (U+0061 to U+007A, “a-z”).

Implementation-Specific Query Parameters

Implementations MAY support custom query parameters. However, the names of these query parameters MUST come from a familywhose base name is a legal member name and also contains at least one non a-z character (i.e., outside U+0061 to U+007A).

It is RECOMMENDED that a capital letter (e.g. camelCasing) be used to satisfy the above requirement.

If a server encounters a query parameter that does not follow the naming conventions above, or the server does not know how to process it as a query parameter from this specification, it MUST return 400 Bad Request.

Note: By forbidding the use of query parameters that contain only the characters [a-z], JSON:API is reserving the ability to standardize additional query parameters later without conflicting with existing implementations.

Errors

Processing Errors

A server MAY choose to stop processing as soon as a problem is encountered, or it MAY continue processing and encounter multiple problems. For instance, a server might process multiple attributes and then return multiple validation problems in a single response.

When a server encounters multiple problems for a single request, the most generally applicable HTTP error code SHOULD be used in the response. For instance, 400 Bad Request might be appropriate for multiple 4xx errors or 500 Internal Server Error might be appropriate for multiple 5xx errors.

Error Objects

Error objects provide additional information about problems encountered while performing an operation. Error objects MUST be returned as an array keyed by errors in the top level of a JSON:API document.

An error object MAY have the following members, and MUST contain at least one of:

Appendix

Query Parameters Details

Parsing/Serialization

A query parameter is a name–value pair extracted from, or serialized into, a URI’s query string.

To extract the query parameters from a URI, an implementation MUST run the URI’s query string, excluding the leading question mark, through theapplication/x-www-form-urlencoded parsing algorithm, with one exception: JSON:API allows the specification that defines a query parameter’s usage to provide its own rules for parsing the parameter’s value from the value bytes identified in steps 3.2 and and 3.3 of the application/x-www-form-urlencodedparsing algorithm. The resulting value might not be a string.

Note: In general, the query string parsing built in to servers and browsers will match the process specified above, so most implementations do not need to worry about this.

The application/x-www-form-urlencoded format is referenced because it is the basis for the a=b&c=d style that almost all query strings use today.

However, application/x-www-form-urlencoded parsing contains the bizarre historical artifact that + characters must be treated as spaces, and it requires that all values be percent-decoded during parsing, which makes it impossible to use RFC 3986 delimiter charactersas delimiters. These issues motivate the exception that JSON:API defines above.

Similarly, to serialize a query parameter into a URI, an implementation MUSTuse the application/x-www-form-urlencoded serializer, with the corresponding exception that a parameter’s value — but not its name — may be serialized differently than that algorithm requires, provided the serialization does not interfere with the ability to parse back the resulting URI.

Square Brackets in Parameter Names

With query parameter families, JSON:API allows for query parameters whose names contain square brackets (i.e., U+005B “[” and U+005D “]”).

According to the query parameter serialization rules above, a compliant implementation will percent-encode these square brackets. However, some URI producers — namely browsers — do not always encode them. Servers SHOULDaccept requests in which these square brackets are left unencoded in a query parameter’s name. If a server does accept these requests, it MUST treat the request as equivalent to one in which the square brackets were percent-encoded.