docs/src/dev/io/graphbinary.asciidoc

////
Licensed to the Apache Software Foundation (ASF) under one or more
contributor license agreements.  See the NOTICE file distributed with
this work for additional information regarding copyright ownership.
The ASF licenses this file to You under the Apache License, Version 2.0
(the "License"); you may not use this file except in compliance with
the License.  You may obtain a copy of the License at

  http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

////

[[graphbinary]]
= GraphBinary

GraphBinary is a binary serialization format suitable for object trees, designed to reduce serialization
overhead on both the client and the server, as well as limiting the size of the payload that is transmitted over the
wire.

It describes arbitrary object graphs with a fully-qualified format:

[source]
----
{type_code}{type_info}{value_flag}{value}
----

Where:

* `{type_code}` is a single unsigned byte representing the type number.
* `{type_info}` is an optional sequence of bytes providing additional information of the type represented. This is
specially useful for representing complex and custom types.
* `{value_flag}` is a single byte providing information about the value. Flags have the following meaning:
** `0x01` The value is `null`. When this flag is set, no bytes for `{value}` will be provided.
* `{value}` is a sequence of bytes which content is determined by the type.

All encodings are big-endian.

Quick examples, using hexadecimal notation to represent each byte:

- `01 00 00 00 00 01`: a 32-bit integer number, that represents the decimal number 1. It’s composed by the
type_code `0x01`, and empty flag value `0x00` and four bytes to describe the value.
- `01 00 00 00 00 ff`: a 32-bit integer, representing the number 256.
- `01 01`: a null value for a 32-bit integer. It’s composed by the type_code `0x01`, and a null flag value `0x01`.
- `02 00 00 00 00 00 00 00 00 01`: a 64-bit integer number 1. It’s composed by the type_code `0x02`, empty flags and
eight bytes to describe the value.

== Version 1.0

=== Forward Compatibility

The serialization format supports new types being added without the need to introduce a new version.

Changes to existing types require new revision.

=== Data Type Codes

==== Core Data Types

- `0x01`: Int
- `0x02`: Long
- `0x03`: String
- `0x04`: Date
- `0x05`: Timestamp
- `0x06`: Class
- `0x07`: Double
- `0x08`: Float
- `0x09`: List
- `0x0a`: Map
- `0x0b`: Set
- `0x0c`: UUID
- `0x0d`: Edge
- `0x0e`: Path
- `0x0f`: Property
- `0x10`: TinkerGraph
- `0x11`: Vertex
- `0x12`: VertexProperty
- `0x13`: Barrier
- `0x14`: Binding
- `0x15`: Bytecode
- `0x16`: Cardinality
- `0x17`: Column
- `0x18`: Direction
- `0x19`: Operator
- `0x1a`: Order
- `0x1b`: Pick
- `0x1c`: Pop
- `0x1d`: Lambda
- `0x1e`: P
- `0x1f`: Scope
- `0x20`: T
- `0x21`: Traverser
- `0x22`: BigDecimal
- `0x23`: BigInteger
- `0x24`: Byte
- `0x25`: ByteBuffer
- `0x26`: Short
- `0x27`: Boolean
- `0x28`: TextP
- `0x29`: TraversalStrategy
- `0x2a`: BulkSet
- `0x2b`: Tree
- `0x2c`: Metrics
- `0x2d`: TraversalMetrics
- `0xfe`: Unspecified null object
- `0x00`: Custom

==== Extended Types

- `0x80`: Char
- `0x81`: Duration
- `0x82`: InetAddress
- `0x83`: Instant
- `0x84`: LocalDate
- `0x85`: LocalDateTime
- `0x86`: LocalTime
- `0x87`: MonthDay
- `0x88`: OffsetDateTime
- `0x89`: OffsetTime
- `0x8a`: Period
- `0x8b`: Year
- `0x8c`: YearMonth
- `0x8d`: ZonedDateTime
- `0x8e`: ZoneOffset

=== Null handling

The serialization format defines two ways to represent null values:

- Unspecified null object
- Fully-qualified null

When a parent type can contain any subtype e.g., a object collection, a `null` value must be represented using the
"Unspecified Null Object" type code and the null value flag.

In contrast, when the parent type contains a type parameter that must be specified, a `null` value is represented using
a fully-qualified object using the appropriate type code and type information.

=== Data Type Formats

==== Int

Format: 4-byte two's complement integer.

Example values:

- `00 00 00 01`: 32-bit integer number 1.
- `00 00 01 01`: 32-bit integer number 256.
- `ff ff ff ff`: 32-bit integer number -1.
- `ff ff ff fe`: 32-bit integer number -2.

==== Long

Format: 4-byte two's complement integer.

Example values

- `00 00 00 00 00 00 00 01`: 64-bit integer number 1.
- `ff ff ff ff ff ff ff fe`: 64-bit integer number -2.

==== String

Format: `{length}{text_value}`

Where:

- `{length}` is an `Int` describing the byte length of the text. Length is a positive number or zero to represent
the empty string.
- `{text_value}` is a sequence of bytes representing the string value in UTF8 encoding.

Example values

- `00 00 00 03 61 62 63`: the string 'abc'.
- `00 00 00 04 61 62 63 64`: the string 'abcd'.
- `00 00 00 00`: the empty string ''.

==== Date

Format: An 8-byte two's complement signed integer representing a millisecond-precision offset from the unix epoch.

Example values

- `00 00 00 00 00 00 00 00`: The moment in time 1970-01-01T00:00:00.000Z.
- `ff ff ff ff ff ff ff ff`: The moment in time 1969-12-31T23:59:59.999Z.

==== Timestamp

Format: The same as `Date`.

==== Class

Format: A `String` containing the fqcn.

==== Double

Format: 8 bytes representing IEEE 754 double-precision binary floating-point format.

Example values

- `3f f0 00 00 00 00 00 00`: Double 1
- `3f 70 00 00 00 00 00 00`: Double 0.00390625
- `3f b9 99 99 99 99 99 9a`: Double 0.1

==== Float

Format: 4 bytes representing IEEE 754 single-precision binary floating-point format.

Example values

- `3f 80 00 00`: Float 1
- `3e c0 00 00`: Float 0.375

==== List

An ordered collection of items.

Format: `{length}{item_0}...{item_n}`

Where:

- `{length}` is an `Int` describing the length of the collection.
- `{item_0}...{item_n}` are the items of the list. `{item_i}` is a fully qualified typed value composed of
`{type_code}{type_info}{value_flag}{value}`.

==== Set

A collection that contains no duplicate elements.

Format: Same as `List`.

==== Map

A dictionary of keys to values.

Format: `{length}{item_0}...{item_n}`

Where:

- `{length}` is an `Int` describing the length of the map.
- `{item_0}...{item_n}` are the items of the map. `{item_i}` is sequence of 2 fully qualified typed values one
representing the key and the following representing the value, each composed of
`{type_code}{type_info}{value_flag}{value}`.

==== UUID

A 128-bit universally unique identifier.

Format: 16 bytes representing the uuid.

Example

- `00 11 22 33 44 55 66 77 88 99 aa bb cc dd ee ff`: Uuid 00112233-4455-6677-8899-aabbccddeeff.

==== Edge

Format: `{id}{label}{inVId}{inVLabel}{outVId}{outVLabel}{parent}{properties}`

Where:

- `{id}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{label}` is a `String` value.
- `{inVId}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{inVLabel}` is a `String` value.
- `{outVId}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{outVLabel}` is a `String` value.
- `{parent}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which contains
the parent `Vertex`. Note that as TinkerPop currently send "references" only, this value will always be `null`.
- `{properties}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which contains
the properties for the edge. Note that as TinkerPop currently send "references" only this value will always be `null`.

==== Path

Format: `{labels}{objects}`

Where:

- `{labels}` is a `List` in which each item is a `Set` of `String`.
- `{objects}` is a `List` of fully qualified typed values.

==== Property

Format: `{key}{value}{parent}`

Where:

- `{key}` is a `String` value.
- `{value}`  is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{parent}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which is either
an `Edge` or `VertexProperty`. Note that as TinkerPop currently sends "references" only this value will always be
`null`.

==== Graph

A collection of vertices and edges. Note that while similar the vertex/edge formats here hold some differences as
compared to the `Vertex` and `Edge` formats used for standard serialization/deserialiation of a single graph element.

Format: `{vlength}{vertex_0}...{vertex_n}{elength}{edge_0}...{edge_n}`

Where:

- `{vlength}` is an `Int` describing the number of vertices.
- `{vertex_0}...{vertex_n}` are vertices as described below.
- `{elength}` is an `Int` describing the number of edges.
- `{edge_0}...{edge_n}` are edges as described below.

Vertex Format: `{id}{label}{plength}{property_0}...{property_n}`

- `{id}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{label}` is a `String` value.
- `{plength}` is an `Int` describing the number of properties on the vertex.
- `{property_0}...{property_n}` are the vertex properties consisting of `{id}{label}{value}{parent}{properties}` as
defined in `VertexProperty` where the `{parent}` is always `null` and `{properties}` is a `List` of `Property` objects.

Edge Format: `{id}{label}{inVId}{inVLabel}{outVId}{outVLabel}{parent}{properties}`

Where:

- `{id}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{label}` is a `String` value.
- `{inVId}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{inVLabel}` is always `null`.
- `{outVId}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{outVLabel}` is always `null`.
- `{parent}` is always `null`.
- `{properties}` is a `List` of `Property` objects.

==== Vertex

Format: `{id}{label}{properties}`

Where:

- `{id}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{label}` is a `String` value.
- `{properties}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which contains
properties. Note that as TinkerPop currently send "references" only, this value will always be `null`.

==== VertexProperty

Format: `{id}{label}{value}{parent}{properties}`

Where:

- `{id}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{label}` is a `String` value.
- `{value}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.
- `{parent}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which contains
the parent `Vertex`. Note that as TinkerPop currently send "references" only, this value will always be `null`.
- `{properties}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` which contains
properties. Note that as TinkerPop currently send "references" only, this value will always be `null`.

==== Barrier

Format: a single `String` representing the enum value.

==== Binding

Format: `{key}{value}`

Where:

- `{key}` is a `String` value.
- `{value}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

==== Bytecode

Format: `{steps_length}{step_0}...{step_n}{sources_length}{source_0}...{source_n}`

Where:

* `{steps_length}` is an `Int` value describing the amount of steps.
* `{step_i}` is composed of `{name}{values_length}{value_0}...{value_n}`, where:
** `{name}` is a String.
** `{values_length}` is an `Int` describing the amount values.
** `{value_i}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}` describing the step argument.
* `{sources_length}` is an `Int` value describing the amount of source instructions.
* `{source_i}` is composed of `{name}{values_length}{value_0}...{value_n}`, where:
** `{name}` is a `String`.
** `{values_length}` is an `Int` describing the amount values.
** `{value_i}`  is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

==== Cardinality

Format: a single `String` representing the enum value.

==== Column

Format: a single `String` representing the enum value.

==== Direction

Format: a single `String` representing the enum value.

==== Operator

Format: a single `String` representing the enum value.

==== Order

Format: a single `String` representing the enum value.

==== Pick

Format: a single `String` representing the enum value.

==== Pop

Format: a single `String` representing the enum value.

==== Lambda

Format: `{language}{script}{arguments_length}`
Where:

- `{language}` is a `String`.
- `{script}` is a `String`.
- `{arguments_length}` is an `Int`.

==== P

Format: `{name}{values_length}{value_0}...{value_n}`

Where:

- `{name}` is a String.
- `{values_length}` is an `Int` describing the amount values.
- `{value_i}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

==== Scope

Format: a single `String` representing the enum value.

==== T

Format: a single `String` representing the enum value.

==== Traverser

Format: `{bulk}{value}`

Where:

- `{bulk}` is an `Int`.
- `{value}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

==== BigDecimal

Represents an arbitrary-precision signed decimal number, consisting of an arbitrary precision integer unscaled value
and a 32-bit integer scale.

Format: `{scale}{unscaled_value}`

Where:

- `{scale}` is an `Int`.
- `{unscaled_value}` is a `BigInteger`.

==== BigInteger

A variable-length two's complement encoding of a signed integer.

Format: `{length}{value}`

Where:

- `{length}` is an `Int`.
- `{value}` is the two's complement of the `BigInteger`.

Example values of the two's complement `{value}`:

- `00`: Integer 0.
- `01`: Integer 1.
- `127`: Integer 7f.
- `00 80`: Integer 128.
- `ff`: Integer -1.
- `80`: Integer -128.
- `ff 7f`: Integer -129.

==== Byte

An unsigned 8-bit integer.

==== ByteBuffer

Format: `{length}{value}`

Where:

- `{length}` is an `Int` representing the amount of bytes contained in the value.
- `{value}` sequence of bytes.

==== Short

Format: 2-byte two's complement integer.

==== Boolean

Format: A single byte containing the value `0x01` when it's `true` and `0` otherwise.

==== TextP

Format: `{predicate}{values_length}{value_0}...{value_n}`

Where:

- `{name}` is a String.
- `{values_length}` is an `Int` describing the amount values.
- `{value_i}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

==== TraversalStrategy

Format: `{strategy_class}{configuration}`

Where:

- `{strategy_class}` is a `Class` that is of type `TraversalStrategy`.
- `{configuration}` is a `Map` of data used to configure the strategy that will be given to a `TraversalStrategy` `create(Configuration)` method.

==== BulkSet

Format: `{length}{item_0}...{item_n}`

Where:

- `{length}` is an `Int` describing the length of the `BulkSet`.
- `{item_0}...{item_n}` are the items of the `BulkSet`. `{item_i}` is a sequence of a fully qualified typed value composed of
`{type_code}{type_info}{value_flag}{value}` followed by the "bulk" which is a `Long` value.

If the implementing language does not have a `BulkSet` object to deserialize into, this format can be coerced to a
`List` and still be considered compliant with Gremlin. Simply "expand the bulk" by adding the item to the `List` the
number of times specified by the bulk.

==== Tree

Format: `{length}{item_0}...{item_n}`

Where:

- `{length}` is an `Int` describing the amount of items.
- `{item_0}...{item_n}` are the items of the `Tree`. `{item_i}` is composed of a `{key}` which is a fully-qualified typed value
followed by a `{Tree}`.

==== Metrics

Format: `{id}{name}{duration}{counts}{annotations}{nested_metrics}`

Where:

- `{id}` is a `String` representing the identifier.
- `{name}` is a `String` representing the name.
- `{duration}` is a `Long` describing the duration in nanoseconds.
- `{counts}` is a `Map` composed by `String` keys and `Long` values.
- `{annotations}` is a `Map` composed by `String` keys and a value of any type.
- `{nested_metrics}` is a `List` composed by `Metrics` items.

==== TraversalMetrics

Format: `{duration}{metrics}`

Where:

- `{duration}` is a `Long` describing the duration in nanoseconds.
- `{metrics}` is a `List` composed by `Metrics` items.

==== Custom

A custom type, represented as a blob value.

Type Info: `{name}{custom_type_info}`

Where:

- `{name}` is a `String` containing the implementation specific text identifier of the custom type.
- `{custom_type_info}` is a `ByteBuffer` representing the additional type information, specially useful
for complex custom types.

Value format: `{blob}`

Where:

- `{blob}` is a `ByteBuffer`.

==== Unspecified Null Object

A `null` value for an unspecified Object value.

It's represented using the null `{value_flag}` set and no sequence of bytes.

==== Char

Format: one to four bytes representing a single UTF8 char, according to the Unicode standard.

For characters `0x00`-`0x7F`, UTF-8 encodes the character as a single byte.

For characters `0x80`-`0x07FF`, UTF-8 uses 2 bytes: the first byte is binary `110` followed by the 5 high bits of the
character, while the second byte is binary 10 followed by the 6 low bits of the character.

The 3 and 4-byte encodings are similar to the 2-byte encoding, except that the first byte of the 3-byte encoding starts
with `1110` and the first byte of the 4-byte encoding starts with `11110`.

Example values (hex bytes)

- `97`: Character 'a'.
- `c2 a2`: Character '¢'.
- `e2 82 ac`: Character '€'

==== Duration

A time-based amount of time.

Format: `{seconds}{nanos}`

Where:

- `{seconds}` is a `Long`.
- `{nanos}` is an `Int`.

==== InetAddress

Format: Same as `ByteBuffer`, having only 4 byte or 16 byte sequences allowed.

==== Instant

An instantaneous point on the time-line.

Format: `{seconds}{nanos}`

Where:

- `{seconds}` is a `Long`.
- `{nanos}` is an `Int`.

==== LocalDate

A date without a time-zone in the ISO-8601 calendar system.

Format: `{year}{month}{day}`

Where:

- `{year}` is an `Int` from -999,999,999 to 999,999,999.
- `{month}` is a `Byte` to represent the month, from 1 (January) to 12 (December)
- `{day}` is a `Byte` from 1 to 31.

==== LocalDateTime

Format: `{date}{time}`

Where:

- `{date}` is `LocalDate`.
- `{time}` is a `LocalTime`.

==== LocalTime
A time without a time-zone in the ISO-8601 calendar system.

Format: An 8 byte two's complement long representing nanoseconds since midnight.

Valid values are in the range 0 to 86399999999999

==== MonthDay

A month-day in the ISO-8601 calendar system.

Format: `{month}{day}`

Where:

- `{month}` is `Byte` value from 1 to 12.
- `{day}` is `Byte` value from 1 to 31.

==== OffsetDateTime

A date-time with an offset from UTC/Greenwich in the ISO-8601 calendar system, such as 2007-12-03T10:15:30+01:00.

Format: `{local_date_time}{offset}`

Where:

- `{local_date_time}` is `LocalDateTime`.
- `{offset}` is `ZoneOffset`.

==== OffsetTime

A time with an offset from UTC/Greenwich in the ISO-8601 calendar system, such as 10:15:30+01:00.

Format: `{local_time}{offset}`

Where:

- `{local_time}` is `LocalTime`.
- `{offset}` is `ZoneOffset`.

==== Period

A date-based amount of time in the ISO-8601 calendar system, such as '2 years, 3 months and 4 days'.

Format: `{years}{month}{days}`

Where:

`{years}`, `{month}` and `{days}` are `Int` values.

==== Year

A year in the ISO-8601 calendar system, such as 2018.

Format: An `Int` representing the years.

==== YearMonth

A year-month in the ISO-8601 calendar system, such as 2007-12.

Format: `{year}{month}`

Where:

- `{year}` is an `Int`.
- `{month}` is a `Byte` from 1 to 12.

==== ZonedDateTime

A date-time with a time-zone in the ISO-8601 calendar system.

Format: `{local_date_time}{zone_offset}`

Where:

- `{local_date_time}` is `LocalDateTime`.
- `{zone_offset}` is a `ZoneOffset`.

==== ZoneOffset

A time-zone offset from Greenwich/UTC, such as +02:00.

Format: An `Int` representing total zone offset in seconds.

=== Request and Response Messages

Request and response messages are special container types used to represent messages from client to the server and the
other way around. These messages are independent from the transport layer.

==== Request Message

Represents a message from the client to the server.

Format: `{version}{request_id}{op}{processor}{args}`

Where:

- `{version}` is a `Byte` representing the specification version, with the most significant bit set to one. For this
version of the format, the value expected is `0x81` (`10000001`).
- `{request_id}` is a `UUID`.
- `{op}` is a `String`.
- `{processor}` is a `String`.
- `{args}` is a `Map`.

The total length is not part of the message as the transport layer will provide it. For example: WebSockets,
as a framing protocol, defines payload length.

==== Response Message

Format: `{version}{request_id}{status_code}{status_message}{status_attributes}{result_meta}{result_data}`

Where:

- `{version}` is a `Byte` representing the protocol version, with the most significant bit set to one. For this version
of the protocol, the value expected is `0x81` (`10000001`).
- `{request_id}` is a nullable `UUID`.
- `{status_code}` is an `Int`.
- `{status_message}` is a nullable `String`.
- `{status_attributes}` is a `Map`.
- `{result_meta}` is a `Map`.
- `{result_data}` is a fully qualified typed value composed of `{type_code}{type_info}{value_flag}{value}`.

The total length is not part of the message as the transport layer will provide it.