Rest.li Protocol

Contents

• URI Syntax
• Online Documentation
• Content Types
• Rest.li Protocol 2.0 Object and List/Array Representation
• Complex types, complex keys, and compound keys in the Rest.li 2.0 Protocol
• Collection Resources
• Simple Resources
• Association Resources
• Finders
• Actions
• URI Modifiers
• Response Status Codes
• Message Headers
• Request Message Body
• Response Message Body
• Complex Types
• Empty List Parameters

Note - Any time there is a difference between Rest.li protocol 1.0 and Rest.li protocol 2.0 it will be explicitly mentioned. If nothing is said, that means that there is no difference.

URI Syntax

URIs are described using URI templates as defined in the IETF Draft Spec.

Online Documentation

Rest.li provides online documentation for any loaded resource. The documentation shows example request and response for the resource methods, finders, and actions. Use it to document the Rest.li protocols.

Content Types

The content types of Rest.li data are application/json and application/pson. PSON is a compressed version of JSON.

Rest.li Protocol 2.0 Object and List/Array Representation

Definitions

An object, as used in this document, is a map or a dictionary. It is simply a collection of key-value pairs, where the keys are strings and the values are primitives, arrays, or an object. The terms object and map are used interchangeably throughout this document to refer to the same concept.

Consider the following functions:

encoded(v) is defined as follows -

• if v is a primitive then encoded(v) = the URL encoded value for v
• if v is a map then encoded(v) = the URL encoding for a map as described in the Rest.li 2.0 protocol object URL representation
• if v is an array then encoded(v) = the URL encoding for an array as described in the Rest.li 2.0 protocol array URL representation

reducedEncoded(v) is defined as follows -

• if v is a primitive then reducedEncoded(v) = v with the characters ,, (, ), ', and : URL encoded
• if v is a map then reducedEncoded(v) = the HTTP body/header encoding for a map as described in the Rest.li 2.0 protocol object HTTP body and headers representation
• if v is an array then reducedEncoded(v) = the HTTP body/header encoding for an array as described in the Rest.li 2.0 protocol array HTTP body and headers representation

encoded and reducedEncoded will be used in the sections below.

JSON Serialization of Data Schema

Rest.li objects defined using Pegasus Data Schema (PDSC) is serialized as JSON representation for transportation over the wire. For detailed transport serialization, please see How Data is Serialized for Transport.

Rest.li Protocol 2.0 Object Representation

URL Representation

In Rest.li 2.0, the way to represent an object in the URL is using key:value pairs. More concretely -

(encoded(k1):encoded(v1),encoded(k2):encoded(v2),...)

Note that all keys are strings.

HTTP body and headers representation

Header representation

An object can be present as values for the following headers -

• Location - if present here we simply use the Rest.li 2.0 protocol object URL representation
• X-RestLi-Id or X-LinkedIn-Id - we use the Rest.li 2.0 protocol object HTTP body and headers representation

Body representation

If present in the HTTP body a key:value representation is used. More concretely -

(reducedEncoded(k1):reducedEncoded(v1),reducedEncoded(k2):reducedEncoded(v2),...)

Rest.li 2.0 Protocol Array Notation

URL Representation

An array \[a1, a2, a3, …\] is encoded in the URL as -

List(encoded(a1),encoded(a2),encoded(a3),...)

HTTP body and headers representation

Header representation

An array can be present as values for the following headers -

• Location - if present here we simply use the Rest.li 2.0 protocol array URL representation
• X-RestLi-Id or X-LinkedIn-Id - we use the Rest.li 2.0 protocol array HTTP body and headers representation

Body representation

An array [a1, a2, a3, ...] is encoded as follows -

List(reducedEncoded(a1),reducedEncoded(a2),reducedEncoded(a3),...)

Example

Consider the object, which we will call exampleObject, expressed in a JSON notation here -

{
  "k1": "v1",
  "k2": "value with spaces",
  "k3": [1, 2, 3],
  "k4": "value:with:reserved:char"
  "k5":
  {
    "k51": "v51",
    "k52": "v52"
  }
}

Here is how exampleObject would look if present in the URL -

(k1:v1,k2:value%20with%20spaces,k3:List(1,2,3),k4:value%3Awith%3Areserved%3Achar,k5:(k51:v51,k52:v52))

Here is how exampleObject would look if present in the HTTP headers -

Location: ...(k1:v1,k2:value%20with%20spaces,k3:List(1,2,3),k4:value%3Awith%3Areserved%3Achar,k5:(k51:v51,k52:v52))
X-RestLi-Id: (k1:v1,k2:value with spaces,k3:List(1,2,3),k4:value%3Awith%3Areserved%3Achar,k5:(k51:v51,k52:v52))

If we were doing a BATCH_GET request and exampleObject was one of the keys requested here is how the HTTP body would look -

...
"entities": {
  "(k1:v1,k2:value with spaces,k3:List(1,2,3),k4:value%3Awith%3Areserved%3Achar,k5:(k51:v51,k52:v52))": {
    ...
  }
  ...
}
...

Complex types, complex keys, and compound keys in the Rest.li 2.0 Protocol

A complex type is simply a map.

A complex key is a key made up of two parts, key and $params, each of which is a complex type.

A compound key is a complex type with the restriction that all the values are primitives. A compound key cannot have maps or arrays as values for the keys making up its map.

Complex and compound keys have a similar structure: they are essentially a collection of key-value pairs. Because of this similarity in structure we decided to represent both using the Rest.li 2.0 object notation. Any list present as a value in the complex key uses the Rest.li 2.0 list notation. Details can be found in the Association keys and complex keys sections.

Collection Resources

The URI templates below assume variables with types as follows:

collection : simple string or "complex key"
entity_id : simple string
ids : list
finder : simple string
params : associative array

Collection URIs

Simple Resources

The URI templates below assume variables with types as follows:

simple : simple string
params : associative array

Simple URIs

Association Resources

Associations contain entities referenced by compound keys, referred to here as assockeys

The URI templates below assume variables with types as follows:

firstkey : associative array containing a single element
keys : associative array
association : simple string
assockey : simple string conforming to the assockey syntax described below
assockeys : list of strings conforming to the assockey syntax
finder : simple string
params : associative array

Association Keys

Association keys are composed of one or more named assocKey parts.

In protocol 1.0 the key is represented over the wire in the form:

{firstkey}{&keys*}

In protocol 2.0 the key is represented using the protocol 2.0 object notation, with each assocKey being a key in the map.

For example, a two part key identifying an edge in a following graph in protocol 1.0 might be:

followerID=1&followeeID=3

In protocol 2.0 the key would be

(followerID:1,followeeID:3)

Here’s an example association GET request/response. In protocol 1.0:

GET /associations/src=KEY1&desk=KEY2
{
    "message": "Hi!",
    "id": "1"
}

In protocol 2.0:

GET /associations/(src:KEY1,desk:KEY2)
{
    "message": "Hi!",
    "id": "1"
}

In finders, only some keys from the full association key might be required. For example, in protocol 1.0:

followerID=1

In protocol 2.0:

(followerID:1)

All string values for association keys are url encoded. E.g. for the association key composed of code=“1=2b” and widget=“xyz widget”, a GET request in protocol 1.0 using the key would be:

GET /resourceName/code=1%32b&widget=xyz%20widget

In protocol 2.0 it would be:

GET /resourceName/(code:1%32b,widget:xyz%20widget)

When association keys are used in a batch operation, each key is url encoded. For protocol 1.0 the form is:

ids=urlencoded(associationKey1)&ids=urlencoded(associationKey2)...

For protocol 2.0 is ids use the protocol 2.0 array notation.

ids=List((encoded(associationKey1)),(encoded(associationKey2)),...)

For example, in protocol 1.0 a batch get for the keys: src=KEY1\&dest=KEY3 and src=KEY1\&dest=KEY2, would be:

GET /associations?ids=src%3DKEY1%26dest%3DKEY2&ids=src%3DKEY1%26dest%3DKEY3

{
  "errors": {},
    "results": {
        "dest=KEY3&src=KEY1": {
            "message": "Hi!",
            "id": "1"
        },
        "dest=KEY2&src=KEY1": {
            "message": "Hello!",
            "id": "2"
        }
    }
}

In protocol 2.0 a batch get for the keys: (src:KEY1,dest:KEY3) and (src:KEY1,dest:KEY2), would be:

GET /associations?ids=List((src:KEY1,dest:KEY3),(src:KEY1,dest:KEY2))

{
  "errors": {},
    "results": {
        "(dest:KEY3,src:KEY1)": {
            "message": "Hi!",
            "id": "1"
        },
        "(dest:KEY2,src:KEY1)": {
            "message": "Hello!",
            "id": "2"
        }
    }
}

Here’s the basic form of a batch update request using association keys. In protocol 1.0:

PUT /resourceName?ids=urlencoded(key1=urlencoded(value)&key2=urlencoded(value)&...)&ids=...

{
  "entities": {
    "key=urlencoded(value)&key2...": { ... },
    ...
  }
}

Note that in the URL the ids are url encoded AND any strings values for the assocKey parts are double url encoded.

In protocol 2.0 the protocol 2.0 array representation is used for the ids.

For example, for a batch update for the association keys: (code=“1=2b”, name=“xyz widget”) and (code=“567”, name=“rachet”)

The batch update request in protocol 1.0 would be:

PUT /widgets?ids=code%3D1%2532b%26widget%3Dxyz%2520widget&ids=code%3D567%26widget%3Drachet

{
  "entities": {
     "code=1%32b&name=xyz%20widget": {...},
     "code=567&name=rachet": {...}
  }
}

In protocol 2.0 the request would be:

PUT /widgets?ids=List((code:1%202b,name:xyz%20widget),(code:567,name:rachet))

{
  "entities": {
     "(code:1=2b,name:xyz widget)": {...},
     "(code:567,name:rachet)": {...}
  }
}

Association URIs

Finders

The URI templates below assume variables with types as follows:

finder : simple string identifying a finder

Finder URIs

Actions

The URI templates below assume variables with types as follows:

action : simple string

Action URIs

URI Modifiers

The URI templates below assume variables with types as follows:

finder_uri : simple string ...
base_uri : simple string generated via one of the uri templates above
start : simple string
count : simple string
fields : list

URIs

Response Status Codes

Status codes should be interpreted according to the HTTP specification

Common status codes used in Rest.li:

• 200 OK
• 201 Created
• 204 No Content
• 400 Bad Request
• 404 Not Found
• 405 Method Not Allowed
• 500 Internal Server Error

Message Headers

Request Message Body

Single Entity

Single entities are sent as the JSON serialized DataMap representing that entity. The "Content-Type: application/json" header may be included for PUT and POST requests, if omitted, the server will assume the request is content type is "application/json".

Create:

POST /widgets
Content-Type: application/json

{"widgetName":"Lever"}

Read:

for collection and association resources:

GET /widgets/1

for simple resources:

GET /currentWidget

Update:

for collection and association resources:

PUT /widgets/1
Content-Type: application/json

{"widgetName":"Lever"}

for simple resources:

PUT /currentWidget
Content-Type: application/json

{"widgetName":"Lever"}

Delete:

for collection and association resources:

DELETE /widgets/1

for simple resources:

DELETE /currentWidget

Batch Create

A DataMap with field

• “elements”: A JSON array of the resources to batch create/update and the objects are the json serialized values of each resource to create.

E.g.

POST /widgets HTTP/1.1
Content-Type: application/json
X-RestLi-Method: BATCH_CREATE

{
  "elements": [
    {"widgetName":"Ratchet"},
    {"widgetName":"Cog"},
    {"widgetName":"!@&%@$#"}
  ]
}

Response:

{
  "elements": [
    {
      "status": 201,
      "id": "100"
    },
    {
      "status": 201,
      "id": "101"
    },
    {
      "status": 406,
      "error": {
        "status": 406,
        "stackTrace": "...",
        "errorDetails": { ... },
        "serviceErrorCode": 999, 
        "exceptionClass": "...", 
        "message": "..."
      }
  ]
}

Responses are associated to the request by array index. E.g. “Rachet” was assigned id 100, “Cog” was assigned id 101 and “!&%$#” resulted on a 406 response code and was not assigned an id.

Note: Batch create requests must include the HTTP Header:

X-RestLi-Method: BATCH_CREATE

Batch Update

A DataMap with field

• “entities”: A JSON serialized map where the keys are keys of the resources to update and the objects are the json serialized replacement values of each resource to update.

E.g.

PUT /widgets?ids=1&ids=2 HTTP/1.1
Content-Type: application/json
X-RestLi-Method: BATCH_UPDATE

{
  "entities": {
    "1": {"widgetName":"Trebuchet"},
    "2": {"widgetName":"Gear"}
   }
}

Response:

{
  "errors": {},
  "results": {
    "1": {
      "status": 204
    },
    "2": {
      "status": 204
    }
  ]
}

Partial Update

Partial update is a set of operations on data object, which is also an instance of DataMap. Operations are expressed using fields with reserved word names. Every operation relates to the object that contains it, i.e., it’s parent. The following is an example of setting the zipCode of the businessAddress, setting name and homeAddress, and deleting note and birthday of a record.

E.g.

POST /widgets/1 HTTP/1.1
Content-Type: application/json

{
  "patch": {
    "businessAddress": {
      "$set": {
        "zipCode": "94086"
      }
    },
    "$set": {
      "name": "John",
      "homeAddress": {
        "street": "10th",
        "city": "Sunnyvale"
      }
    },
    "$delete": ["note", "birthday"]
  }
}

Patch

A patch object describes the operations needed to change one map to another, which are the underlying data structure for record, union and map. A patch object is represented by a map itself. It has three types keys.

• “$set” indicates a set operation. Its value is a map, indicating the keys and values to be set. This is usually used to set fields in a record.
• “$delete” indicates a delete operation. Its value is a string array, indicating the map entries to be deleted. This is usually used to delete fields in a record.
• Other key indicates a patch operation. It must be an existing key, whose value must be a map, of the map to be changed. Its value is another patch object. This is usually used to update a field in a nested record.

Batch Partial Update

See Partial Update and Batch Update above for details, here the two are combined.

E.g.

POST /widgets?ids=1&ids=2 HTTP/1.1
Content-Type: application/json
X-RestLi-Method: BATCH_PARTIAL_UPDATE

{
  "entities": {
    "1": {"patch": { "$set": { "name":"Sam"}}},
    "2": {"patch": { "$delete": ["name"]}}
   }
}

Action

Action params are provided in the request body which must contain a data map keyed by param names. E.g.

POST /widgets?action=purge HTTP/1.1
Content-Type: application/json
{
  "reason": "spam",
  "purgedByAdminId": 1
}

The X-RestLi-Method: ACTION may optionally be included, but is not required because rest.li is able to determine the request is an action based on the “action” query param key.

Response Message Body

Single Entity

Single entities are returned as the JSON serialized DataMap representing that entity

List of Entities

Lists of entities are returned in a com.linkedin.rest.CollectionResponse wrapper. They are returned by finders and getAll.

CollectionResponse fields:

• “elements” : JSON serialized list of entity types
• (optional) “paging” : JSON serialized CollectionMetadata object

E.g.

{
  "elements: [
    { "id": 1, "message": "Good morning!", "tone": "FRIENDLY" }
    // ...
  ],
  "metadata" { // only returned by finders that define a metadata schema as part of the interface
    // ...
  },
  "paging": {
    "count": 10,
    "links": [
      "href": "/greetings?count=10&start=10&q=search",
      "rel": "next",
      "type": "application/json"
    ],
    "start": 0
  }
}

Map of Entities

Maps of entities are returned in a com.linkedin.restli.common.BatchResponse or com.linkedin.restli.client.response.BatchKVResponse wrapper.

BatchResponse fields:

• “results” : JSON object containing name/value pairs
  • name is the string value of each map key. This is serialized according to the protocol version.
  • value is the JSON serialized entity for each map value. This is serialized according to the protocol version.
• “errors”: another map mapping keys which failed to generate a 2xx response to the reason, which is a ErrorResponse

Protocol 1 example -

GET /fortunes?ids=1&ids=2&ids=unacceptableKey&ids=faultyKey
{
  "errors": {
     "unacceptableKey": {
       "status": 416,
       "message": "Not Acceptable"
     },
     "faultyKey": {
       "status": 500,
       "exceptionClass": "org.example.SomeServerException",
       "stacktrace": "SomeServerException at org.example.SomeClass.someMethod(SomeClass.java:1)\n..."
     }
   },
  "results": {
    "1": {...},
    "2": {...}
    }
  }
}

Protocol 2 example -

GET /fortunes?ids=List(1,2,unacceptableKey,faultyKey)
{
  "errors": {
     "unacceptableKey": {
       "status": 416,
       "message": "Not Acceptable"
     },
     "faultyKey": {
       "status": 500,
       "exceptionClass": "org.example.SomeServerException",
       "stacktrace": "SomeServerException at org.example.SomeClass.someMethod(SomeClass.java:1)\n..."
     }
   },
  "results": {
    "1": {...},
    "2": {...}
    }
  }
}

Collection Metadata

{"type":"record",
 "name":"CollectionMetadata",
 "namespace":"com.linkedin.common.rest",
 "doc":"Metadata and pagination links for this collection",
 "fields":[
 {
   "name":"start",
   "type":"int",
   "doc":"The start index of this collection"
 },{
   "name":"count",
   "type":"int",
   "doc":"The number of elements in this collection segment"
 },{
   "name":"total",
   "type":"int",
   "doc":"The total number of elements in the entire collection (not just this segment)",
   "default":0
 },{
   "name":"links",
   "type":{
     "type":"array",
     "items":{
       "type":"record",
       "name":"Link",
       "doc":"A atom:link-inspired link",
       "fields":[
       {
         "name":"rel",
         "type":"string",
         "doc":"The link relation e.g. 'self' or 'next'"
       },{
         "name":"href",
         "type":"string",
         "doc":"The link URI"
       },{
         "name":"type",
         "type":"string",
         "doc":"The type (media type) of the resource"
       }]
     }
   },
   "doc":"Previous and next links for this collection"
}]}

Error Response

{
  "type":"record",
  "name":"ErrorResponse",
  "namespace":"com.linkedin.common.rest",
  "doc":"A generic ErrorResponse",
  "fields":[
  {
    "name":"status",
    "type":"int",
    "doc":"The HTTP status code"
  },{
    "name":"serviceErrorCode",
    "type":"int",
    "doc":"An service-specific error code (documented in prose)"
  },{
    "name":"message",
    "type":"string",
    "doc":"A human-readable explanation of the error"
  },{
    "name":"exceptionClass",
    "type":"string",
    "doc":"The FQCN of the exception thrown by the server (included the case of a server fault)"
  },{
    "name":"stackTrace",
    "type":"string",
    "doc":"The full (??) stack trace (included the case of a server fault)"
  }]
}

Action Response

Actions may optionally return a response body. If they do, it must contain a data map with a single “value” key, where the value of the key is either a primitive type, e.g.:

{
  "value": 1
}

or complex data type, e.g.:

{
  "value": {
    "firstName": "John",
    "lastName": "Smith"
  }
}

Complex Types

Complex types as keys in protocol 1.0

The serialized form of a complex key uses path keys to specify the values of data elements in a complex data type. For example, given the complex data:

{
  "key": {
    "x": [
      "a1",
      "a2"
    ],
    "y": 123,
    "key.with.dots": "val"
  }
}

Its serialized form is:

// Complex key as a serialized string:
key.x[0]=a1&key.x[1]=a2&key.y=123&key~2Ewith~2Edots=val

If this serialized form is put into a URI (as it usually is), the ‘[’ must be escaped as ‘%5B’ and the ‘]’ must be escaped as ‘%5D’ (URIs require this), so you have the URI form:

// Complex key as a serialized string, escaped for URI:
key.x%5B0%5D=a1&key.x%5B1%5D=a2&key.y=123&key~2Ewith~2Edots=val

Where, in the values of the query params, the chars ‘.[]’ are “~ encoded” to their ascii values. This encoding is the same as “% encoding” except that the escape char is ‘~’ and the only reserved chars are ‘.[]’.

. -> ~2E
[ -> ~5B
] -> ~5D
~ -> ~7E

The Params of a ComplexResourceKey are always prefixed with “$params.” when represented in a URI, e.g.:

$params.x=a1

Complex types as keys in protocol 2.0

The serialized form of a complex key uses the Rest.li 2.0 protocol object notation . For example, given the complex data:

{
  "key": {
    "x": [
      "a1",
      "a2"
    ],
    "y": 123,
    "key.with.dots": "val"
  }
}

Its serialized form is:

// Complex key as a serialized string:
(key:(x:List(a1,a2)),y:123,key.with.dots:val)

The Params of a ComplexResourceKey are always prefixed with “$params.” when represented in a URI, e.g.:

$params:(x:a1)

Complex keys in batch requests in protocol 1.0

If used in batch requests, each key in the batch is represented as a element in an array, the complex data representation is:

[
  { <complekey1> }, { <complexkey2> }
]

And it’s serialized representation is just the list flattened using the same rules as with any complex key, and with the same “~ encoding” applied.

For example,

[
  { "keypart1":"v1", "keypart2":"v2" }, { "keypart1":"v3", "keypart2":"v4" }
]

It’s serialized form is:

// Complex key as a serialized string:
ids[0].keypart1=v1&ids[0].keypart2=v2&ids[1].keypart1=v3&ids[1].keypart2=v4

If this serialized form is put into a URI (as it usually is), the ‘[’ must be escaped as ‘%5B’ and the ‘]’ must be escaped as ‘%5D’ (URIs require this), so you have the URI form:

// Complex key as a serialized string, escaped for URI:
ids%5B0%5D.keypart1=v1&ids%5B0%5D.keypart2=v2&ids%5B1%5D.keypart1=v3&ids%5B1%5D.keypart2=v4

If $params are in a batch complex key key, they are also prefixed by their key’s position in the ids array, e.g. “ids[0].$params.parmkeypart1=v5”

When complex keys are used in batch requests, they are often included both in the URI and in the json body.

For example, a batch update request has the ids in the URI as well as the “entities” part of the body:

PUT /widgets?ids%5B0%5D.keypart1=v1&ids%5B0%5D.keypart2=v2&ids%5B1%5D.keypart1=v3&ids%5B1%5D.keypart2=v4

{
  "entities": {
    "keypart1=v1&keypart2=v2": { <content to put for v1,v2> }
    "keypart1=v3&keypart2=v4": { <content to put for v3,v4> }
  }
}

Note how the paths for the keys in the URI are prefixed by an ids array position, but the paths for the keys in the JSON body are not.

Complex keys in batch requests in protocol 2.0

If used in batch requests, each key in the batch is represented as an element in the ids array using the protocol 2.0 array notation .

For example,

[
  { "keypart1":"v1", "keypart2":"v2" }, { "keypart1":"v3", "keypart2":"v4" }
]

It’s serialized form is:

// Complex key as a serialized string:
ids=List((keypart1:v1,keypart2:v2),(keypart1:v3,keypart2:v4))

If $params are in a batch complex key key, they included in the same object as their id portion. e.g.

ids=List(($params:(parmkeypart1:v5),keypart1:v1,keypart2:v2),($params:(parmkeypart1:v55),keypart1:v11,keypart2:v22))

When complex keys are used in batch requests, they are often included both in the URI and in the JSON body.

For example, a batch update request has the ids in the URI as well as the “entities” part of the body:

PUT /widgets?ids=List((keypart1:v1,keypart2:v2),(keypart1:v3,keypart2:v4))

{
  "entities": {
    "(keypart1:v1,keypart2:v2)": { <content to put for v1,v2> }
    "(keypart1:v3,keypart2:v4)": { <content to put for v3,v4> }
  }
}

As long as the protocol 2.0 array notation is being adhered to no addional escaping is required for complex keys in a batch request.

Query parameters as complex types in protocol 1.0

Complex types work the same for query params as they do for keys. See the above sections for details. E.g. for a complex type:

{
  "a": 1,
  "b": 2
}

In protocol 1.0 a query param named “param1” of this type would be:

...?param1.a=1&param1.b=2

In protocol 2.0 a query param named “param1” of this type would be:

...?param1=(a:1,b:2)

Empty List Parameters

Suppose you have a finder called search with a parameter called filters which is a list of some type.

In protocol 1, if a client attempted to send an empty list the URL would look as follows -

/resource?q=search

Note that the parameter is lost in this case, and on the server side it would be treated as a null or uninitialized field. This was a bug in protocol 1!

We have fixed this in protocol 2. Here is how the URL looks in protocol 2 when you send in an empty list -

/resource?q=search&filters=List()

In other words, List() denotes an empty list in protocol 2. This applies to all cases in protocol 2 where lists are being used.

Empty map parameters

Suppose you have a finder called search with a parameter called preferences which is a map of some type.

In protocol 1, if a client attempted to send an empty map the URL would look as follows -

/resource?q=search

Note that the parameter is lost in this case, and on the server side it would be treated as a null or uninitialized field. This was a bug in protocol 1!

We have fixed this in protocol 2. Here is how the URL looks in protocol 2 when you send in an empty map -

/resource?q=search&preferences=()

In other words, () denotes an empty map in protocol 2. This applies to all cases in protocol 2 where maps are being used.

Empty string parameters

Empty strings are a special case in Rest.li Protocol 2.0. They must be represented with two single quotes ’’. This applies if the empty string is a value or a map key. In Rest.li 1.0, empty strings would simply appear as nothing.