ernie
ernie copied to clipboard
Efficient binary encoding for your data (based on Erlang's External Terms Format)
Ernie
Simple binary data exchange format that is based on External Term Format in Erlang. This is update on BERT which is a little bit outdated right now.
Name
Ernie is fun over the original BERT name, from the Sesame Street Bert & Ernie couple.
RPC
No, this document will not define RPC syntax because of 2 reasons:
- This is out of scope of this document. It is meant to define only data encoding format, not services one will use it for.
- Authors of this document believes that RPC are bad, and you should not use them. Instead use message passing between your services.
Schema/structured data
This is meant to encode free-form data, if you want to encode structured data with schema then we highly suggest you to take look into other formats like:
- ASN.1 and any of it's encodings
- ProtoBuffers
- Apache Thrift
- FlatBuffers
- Cap'n'Proto
Specification
Introduction
- The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14.
Types
-
Integer
Binary representation of integers.
Implementation MUST support at least 8-bit unsigned and 32-bit signed integers with full precision.
-
64-bit Float
IEEE 754 double precision floats.
Implementation MUST support all non-NaN, non-infinite, non-subnormal values described in IEEE 754 binary64 spec. Implementations SHOULD NOT encode these values.
-
Tuple
Fixed length sequence of values.
-
List
Dynamic length sequence of values.
-
Binary
Representation of bytes with arbitrary length. If used to store human-readable strings it MUST be encoded as correct UTF-8 string.
-
Map
List of key-value pairs in arbitrary order. Keys MUST NOT appear twice in the same map.
Encoding
-
Top level
Binary encoded data MUST start with magic byte of decimal value of
131
followed by single byteTag
value followed by arbitrary length data:0 1 2 +--------+-------+-------+------- | 131 | Tag | Data ... +--------+-------+-------+-------
-
Integer
Integer can be encoded in 3 different forms:
-
Values in range 0..255 (aka byte)
0 1 2 +-------+--------+ | 97 | Data | +-------+--------+
Where
Data
is unsigned byte representation of value. -
Values in range -4 294 967 295..4 294 967 294 (signed 32-bit value)
0 1 2 3 4 5 +-------+-------+------+------+------+ | 98 | Data | +-------+-------+------+------+------+
Where
Data
is signed 32-bit representation of integer in U2 big-endian encoding. -
"Short" arbitrary long signed integers (up to 2040-bit)
0 1 2 3 ... N+3 +-------+-------+-------+------+------+ | 110 | N | Sign | Data | +-------+-------+-------+------+------+
N
is big-endian 16-bit unsigned integer.Data
represents little-endian encodedN
byte long integer.Sign
is0
for positive integer and1
for negative one. -
"Long" arbitrary long signed integers (from 2041 to 524 288-bit long)
0 1 2 3 4 ... N+4 +-------+-------+-------+-------+------+------+ | 111 | N | Sign | Data | +-------+-------+-------+-------+------+------+
N
is big-endian 16-bit unsigned integer.Data
represents little-endian encodedN
byte long integer.Sign
is0
for positive integer and1
for negative one.
-
-
Float
0 1 2 3 4 5 6 7 8 9 +-------+-------+--------+-------+-------+-------+-------+-------+-------+ | 70 | Data | +-------+-------+--------+-------+-------+-------+-------+-------+-------+
Where
Data
is IEEE 754 double precision floating-point number encoded as big-endian. -
Tuple
-
Tuples up to 255 elements
0 1 2 ... N+2 +-------+-------+--------+--------+ | 104 | Arity | Data ... | +-------+-------+--------+--------+
-
Tuples up to 4 294 967 295 elements
0 1 2 3 4 5 ... N+5 +-------+-------+--------+--------+--------+--------+--------+ | 105 | Arity | Data ... | +-------+-------+--------+--------+--------+--------+--------+
Where
Arity
is big-endian unsigned integer declaring amount of elements inData
. -
-
List
-
Byte list
0 1 2 3 ... N+3 +-------+-------+-------+-------+-------+ | 107 | N | Data ... | +-------+-------+-------+-------+-------+
This is optimisation for lists that contains only values in range 0..255, inclusive.
-
General list
0 1 2 3 4 5 ... N+5 N+6 +-------+-------+-------+-------+-------+-------+-------+-------+ | 108 | Arity | Data ... | 106 | +-------+-------+-------+-------+-------+-------+-------+-------+
Where
Arity
is big-endian unsigned 32-bit integer declaring amount of elements inData
. -
Empty list
0 1 +-------+ | 106 | +-------+
-
-
Binary
0 1 2 3 4 5 ... N+5 +-------+-------+-------+-------+-------+-------+-------+ | 109 | N | Data ... | +-------+-------+-------+-------+-------+-------+-------+
-
Map
0 1 2 3 4 5 ... N+5 +-------+-------+-------+-------+-------+-------+-------+ | 116 | Arity | Data ... | +-------+-------+-------+-------+-------+-------+-------+
The
Arity
field is an unsigned 32-bit integer in big-endian format that determines the number of key-value pairs in the map. Key and value pairs (Ki => Vi
) are encoded in sectionData
in the following order:K1, V1, K2, V2,..., Kn, Vn
.
Open questions
- Should we support atoms?
- Should we allow defining improper lists?
License
This document is released on Creative Commons Attribution-ShareAlike 4.0 International License.