Serialization Format

[Source]

This page describes the XRP Ledger's canonical binary format for transactions and other data. This binary format is necessary to create and verify digital signatures of those transactions' contents, and is also used in other places. The rippled APIs typically use JSON to communicate with client applications. However, JSON is unsuitable as a format for serializing transactions for being digitally signed, because JSON can represent the same data in many different but equivalent ways.

The process of serializing a transaction from JSON or any other representation into their canonical binary format can be summarized with these steps:

  1. Make sure all required fields are provided, including any required but "auto-fillable" fields.

    The Transaction Formats Reference defines the required and optional fields for XRP Ledger transactions.

    Note: The SigningPubKey must also be provided at this step. When signing, you can derive this key from the secret key that is provided for signing.

  2. Convert each field's data into its "internal" binary format.

  3. Sort the fields in canonical order.

  4. Prefix each field with a Field ID.

  5. Concatenate the fields (including prefixes) in their sorted order.

The result is a single binary blob that can be signed using well-known signature algorithms such as ECDSA (with the secp256k1 elliptic curve) and Ed25519. For purposes of the XRP Ledger, you must also hash the data with the appropriate prefix (0x53545800 if single-signing, or 0x534D5400 if multi-signing). After signing, you must re-serialize the transaction with the TxnSignature field included.

Note: The XRP Ledger uses the same serialization format to represent other types of data, such as ledger objects and processed transactions. However, only certain fields are appropriate for including in a transaction that gets signed. (For example, the TxnSignature field, containing the signature itself, should not be present in the binary blob that you sign.) Thus, some fields are designated as "Signing" fields, which are included in objects when those objects are signed, and "non-signing" fields, which are not.

Examples

Both signed and unsigned transactions can be represented in both JSON and binary formats. The following samples show the same signed transaction in its JSON and binary formats:

JSON:

{
  "Account": "rMBzp8CgpE441cp5PVyA9rpVV7oT8hP3ys",
  "Expiration": 595640108,
  "Fee": "10",
  "Flags": 524288,
  "OfferSequence": 1752791,
  "Sequence": 1752792,
  "SigningPubKey": "03EE83BB432547885C219634A1BC407A9DB0474145D69737D09CCDC63E1DEE7FE3",
  "TakerGets": "15000000000",
  "TakerPays": {
    "currency": "USD",
    "issuer": "rvYAfWj5gh67oV6fW32ZzP3Aw4Eubs59B",
    "value": "7072.8"
  },
  "TransactionType": "OfferCreate",
  "TxnSignature": "30440220143759437C04F7B61F012563AFE90D8DAFC46E86035E1D965A9CED282C97D4CE02204CFD241E86F17E011298FC1A39B63386C74306A5DE047E213B0F29EFA4571C2C",
  "hash": "73734B611DDA23D3F5F62E20A173B78AB8406AC5015094DA53F53D39B9EDB06C"
}

Binary (represented as hexadecimal):

120007220008000024001ABED82A2380BF2C2019001ABED764D55920AC9391400000000000000000000000000055534400000000000A20B3C85F482532A9578DBB3950B85CA06594D165400000037E11D60068400000000000000A732103EE83BB432547885C219634A1BC407A9DB0474145D69737D09CCDC63E1DEE7FE3744630440220143759437C04F7B61F012563AFE90D8DAFC46E86035E1D965A9CED282C97D4CE02204CFD241E86F17E011298FC1A39B63386C74306A5DE047E213B0F29EFA4571C2C8114DD76483FACDEE26E60D8A586BB58D09F27045C46

Sample Code

The serialization processes described here are implemented in multiple places and programming languages:

These implementations are all provided with permissive open-source licenses, so you can import, use, or adapt the code for your needs in addition to using it alongside these documents for learning purposes.

Internal Format

Each field has an "internal" binary format used in the rippled source code to represent that field when signing (and in most other cases). The internal formats for all fields are defined in the source code of SField.cpp . (This file also includes fields other than transaction fields.) The Transaction Format Reference also lists the internal formats for all transaction fields.

For example, the Flags common transaction field becomes a UInt32 (32-bit unsigned integer).

Definitions File

The following JSON file defines the important constants you need for serializing XRP Ledger data to its binary format and deserializing it from binary:

https://github.com/ripple/ripple-binary-codec/blob/master/src/enums/definitions.json

The following table defines the top-level fields from the definitions file:

Field Contents
TYPES Map of data types to their "type code" for constructing field IDs and sorting fields in canonical order. Codes below 1 should not appear in actual data; codes above 10000 represent special "high-level" object types such as "Transaction" that cannot be serialized inside other objects. See the Type List for details of how to serialize each type.
LEDGER_ENTRY_TYPES Map of ledger objects to their data type. These appear in ledger state data, and in the "affected nodes" section of processed transactions' metadata.
FIELDS A sorted array of tuples representing all fields that may appear in transactions, ledger objects, or other data. The first member of each tuple is the string name of the field and the second member is an object with that field's properties. (See the "Field properties" table below for definitions of those fields.)
TRANSACTION_RESULTS Map of transaction result codes to their numeric values. Result types not included in ledgers have negative values; tesSUCCESS has numeric value 0; tec-class codes represent failures that are included in ledgers.
TRANSACTION_TYPES Map of all transaction types to their numeric values.

For purposes of serializing transactions for signing and submitting, the FIELDS, TYPES, and TRANSACTION_TYPES fields are necessary.

The field definition objects in the FIELDS array have the following fields:

Field Type Contents
nth Number The field code of this field, for use in constructing its Field ID and ordering it with other fields of the same data type.
isVLEncoded Boolean If true, this field is length-prefixed.
isSerialized Boolean If true, this field should be encoded into serialized binary data. When this field is false, the field is typically reconstructed on demand rather than stored.
isSigningField Boolean If true this field should be serialized when preparing a transaction for signing. If false, this field should be omitted from the data to be signed. (It may not be part of transactions at all.)
type String The internal data type of this field. This maps to a key in the TYPES map, which gives the type code for this field.

Field IDs

[Source - Encoding] [Source - Decoding]

When you combine a field's type code and field code, you get the field's unique identifier, which is prefixed before the field in the final serialized blob. The size of the Field ID is one to three bytes depending on the type code and field codes it combines. See the following table:

Type Code < 16 Type Code >= 16
Field Code < 16 1 byte: high 4 bits define type; low 4 bits define field. 2 bytes: low 4 bits of the first byte define field; next byte defines type.
Field Code >= 16 2 bytes: high 4 bits of the first byte define type; low 4 bits of first byte are 0; next byte defines field 3 bytes: first byte is 0x00, second byte defines type; third byte defines field

When decoding, you can tell how many bytes the field ID is by which bits of the first byte are zeroes. This corresponds to the cases in the above table:

High 4 bits are nonzero High 4 bits are zero
Low 4 bits are nonzero 1 byte: high 4 bits define type; low 4 bits define field. 2 bytes: low 4 bits of the first byte define field; next byte defines type
Low 4 bits are zero 2 bytes: high 4 bits of the first byte define type; low 4 bits of first byte are 0; next byte defines field 3 bytes: first byte is 0x00, second byte defines type; third byte defines field

Caution: Even though the Field ID consists of the two elements that are used to sort fields, you should not sort by the serialized Field ID itself, because the byte structure of the Field ID changes the sort order.

Length Prefixing

Some types of variable-length fields are prefixed with a length indicator. Blob fields (containing arbitrary binary data) are one such type. For a list of which types are length-prefixed, see the Type List table.

Note: Some types of fields that vary in length are not length-prefixed. Those types have other ways of indicating the end of their contents.

The length prefix consists of one to three bytes indicating the length of the field immediately after the type prefix and before the contents.

  • If the field contains 0 to 192 bytes of data, the first byte defines the length of the contents, then that many bytes of data follow immediately after the length byte.

  • If the field contains 193 to 12480 bytes of data, the first two bytes indicate the length of the field with the following formula:

    193 + ((byte1 - 193) * 256) + byte2
    
  • If the field contains 12481 to 918744 bytes of data, the first three bytes indicate the length of the field with the following formula:

    12481 + ((byte1 - 241) * 65536) + (byte2 * 256) + byte3
    
  • A length-prefixed field cannot contain more than 918744 bytes of data.

When decoding, you can tell from the value of the first length byte whether there are 0, 1, or 2 additional length bytes:

  • If the first length byte has a value of 192 or less, then that's the only length byte and it contains the exact length of the field contents in bytes.
  • If the first length byte has a value of 193 to 240, then there are two length bytes.
  • If the first length byte has a value of 241 to 254, then there are three length bytes.

Canonical Field Order

All fields in a transaction are sorted in a specific order based first on the field's type (specifically, a numeric "type code" assigned to each type), then on the field itself (a "field code"). (Think of it as sorting by family name, then given name, where the family name is the field's type and the given name is the field itself.)

Type Codes

Each field type has an arbitrary type code, with lower codes sorting first. These codes are defined in SField.h .

For example, UInt32 has type code 2 , so all UInt32 fields come before all Amount fields, which have type code 6 .

The definitions file lists the type codes for each type in the TYPES map.

Field Codes

Each field has a field code, which is used to sort fields that have the same type as one another, with lower codes sorting first. These fields are defined in SField.cpp .

For example, the Account field of a Payment transaction has sort code 1 , so it comes before the Destination field which has sort code 3 .

Field codes are reused for fields of different field types, but fields of the same type never have the same field code. When you combine the type code with the field code, you get the field's unique Field ID.

Type List

Transaction instructions may contain fields of any of the following types:

Type Name Type Code Bit Length Length-prefixed? Description
AccountID 8 160 Yes The unique identifier for an account.
Amount 6 64 or 384 No An amount of XRP or issued currency. The length of the field is 64 bits for XRP or 384 bits (64+160+160) for issued currencies.
Blob 7 Variable Yes Arbitrary binary data. One important such field is TxnSignature, the signature that authorizes a transaction.
Hash128 4 128 No A 128-bit arbitrary binary value. The only such field is EmailHash, which is intended to store the MD-5 hash of an account owner's email for purposes of fetching a Gravatar .
Hash160 17 160 No A 160-bit arbitrary binary value. This may define a currency code or issuer.
Hash256 5 256 No A 256-bit arbitrary binary value. This usually represents the "SHA-512Half" hash of a transaction, ledger version, or ledger data object.
PathSet 18 Variable No A set of possible payment paths for a cross-currency payment.
STArray 15 Variable No An array containing a variable number of members, which can be different types depending on the field. Two cases of this include memos and lists of signers used in multi-signing.
STObject 14 Variable No An object containing one or more nested fields.
UInt8 16 8 No An 8-bit unsigned integer.
UInt16 1 16 No A 16-bit unsigned integer. The TransactionType is a special case of this type, with specific strings mapping to integer values.
UInt32 2 32 No A 32-bit unsigned integer. The Flags and Sequence fields on all transactions are examples of this type.

In addition to all of the above field types, the following types may appear in other contexts, such as ledger objects and transaction metadata:

Type Name Type Code Length-prefixed? Description
Transaction 10001 No A "high-level" type containing an entire transaction.
LedgerEntry 10002 No A "high-level" type containing an entire ledger object.
Validation 10003 No A "high-level" type used in peer-to-peer communications to represent a validation vote in the consensus process.
Metadata 10004 No A "high-level" type containing metadata for one transaction.
UInt64 3 No A 64-bit unsigned integer. This type does not appear in transaction instructions, but several ledger objects use fields of this type.
Vector256 19 Yes This type does not appear in transaction instructions, but the Amendments ledger object's Amendments field uses this to represent which amendments are currently enabled.

AccountID Fields

Fields of this type contain the 160-bit identifier for an XRP Ledger account. In JSON, these fields are represented as base58 XRP Ledger "addresses", with additional checksum data so that typos are unlikely to result in valid addresses. (This encoding, sometimes called "Base58Check", prevents accidentally sending money to the wrong address.) The binary format for these fields does not contain any checksum data nor does it include the 0x00 "type prefix" used in address base58 encoding. (However, since the binary format is used mostly for signed transactions, a typo or other error in transcribing a signed transaction would invalidate the signature, preventing it from sending money.)

AccountIDs that appear as stand-alone fields (such as Account and Destination) are length-prefixed despite being a fixed 160 bits in length. As a result, the length indicator for these fields is always the byte 0x14. AccountIDs that appear as children of special fields (Amount issuer and PathSet account) are not length-prefixed.

Amount Fields

The "Amount" type is a special field type that represents an amount of currency, either XRP or an issued currency. This type consists of two sub-types:

  • XRP

    XRP is serialized as a 64-bit unsigned integer (big-endian order), except that the most significant bit is always 0 to indicate that it's XRP, and the second-most-significant bit is 1 to indicate that it is positive. Since the maximum amount of XRP (1017 drops) only requires 57 bits, you can calculate XRP serialized format by taking standard 64-bit unsigned integer and performing a bitwise-OR with 0x4000000000000000.

  • Issued Currencies

    Issued currencies consist of three segments in order:

    1. 64 bits indicating the amount in the internal currency format. The first bit is 1 to indicate that this is not XRP.
    2. 160 bits indicating the currency code. The standard API converts 3-character codes such as "USD" into 160-bit codes using the standard currency code format, but custom 160-bit codes are also possible.
    3. 160 bits indicating the issuer's Account ID. (See also: Account Address Encoding)

You can tell which of the two sub-types it is based on the first bit: 0 for XRP; 1 for issued currency.

The following diagram shows the serialization formats for both XRP amounts and issued currency amounts:

XRP amounts have a "not XRP" bit, a sign bit, and 62 bits of precision. Issued currency amounts consist of a "not XRP" bit, a sign bit, an exponent (8 bits), mantissa (54 bits), currency code (160 bits), and issuer (160 bits).

Array Fields

Some transaction fields, such as SignerEntries (in SignerListSet transactions) and Memos, are arrays of objects (called the "STArray" type).

Arrays contain several object fields in their native binary format in a specific order. In JSON, each array member is a JSON "wrapper" object with a single field, which is the name of the member object field. The value of that field is the ("inner") object itself.

In the binary format, each member of the array has a Field ID prefix (based on the single key of the wrapper object) and contents (comprising the inner object, serialized as an object). To mark the end of an array, append an item with a "Field ID" of 0xf1 (the type code for array with field code of 1) and no contents.

The following example shows the serialization format for an array (the SignerEntries field):

Array field ID, followed by the Field ID and contents of each array element, followed by the "Array end" field ID

Blob Fields

The Blob type is a length-prefixed field with arbitrary data. Two common fields that use this type are SigningPubKey and TxnSignature, which contain (respectively) the public key and signature that authorize a transaction to be executed.

Blob fields have no further structure to their contents, so they consist of exactly the amount of bytes indicated in the variable-length encoding, after the Field ID and length prefixes.

Hash Fields

The XRP Ledger has several "hash" types: Hash128, Hash160, and Hash256. These fields contain arbitrary binary data of the given number of bits, which may or may not represent the result of a hash operation.

All such fields are serialized as the specific number of bits, with no length indicator, in big-endian byte order.

Object Fields

Some fields, such as SignerEntry (in SignerListSet transactions), and Memo (in Memos arrays) are objects (called the "STObject" type). The serialization of objects is very similar to that of arrays, with one difference: object members must be placed in canonical order within the object field, where array fields have an explicit order already.

The canonical field order of object fields is the same as the canonical field order for all top-level fields, but the members of the object must be sorted within the object. After the last member, there is an "Object end" Field ID of 0xe1 with no contents.

The following example shows the serialization format for an object (a single Memo object in the Memos array).

Object field ID, followed by the Object ID and contents of each object member in canonical order, followed by the "Object end" field ID

PathSet Fields

The Paths field of a cross-currency Payment transaction is a "PathSet", represented in JSON as an array of arrays. For more information on what paths are used for, see Paths.

A PathSet is serialized as 1 to 6 individual paths in sequence[Source] . Each complete path is followed by a byte that indicates what comes next:

  • 0xff indicates another path follows
  • 0x00 indicates the end of the PathSet

Each path consists of 1 to 8 path steps in order[Source] . Each step starts with a type byte, followed by one or more fields describing the path step. The type indicates which fields are present in that path step through bitwise flags. (For example, the value 0x30 indicates changing both currency and issuer.) If more than one field is present, the fields are always placed in a specific order.

The following table describes the possible fields and the bitwise flags to set in the type byte to indicate them:

Type Flag Field Present Field Type Bit Size Order
0x01 account AccountID 160 bits 1st
0x10 currency Currency Code 160 bits 2nd
0x20 issuer AccountID 160 bits 3rd

Some combinations are invalid; see Path Specifications for details.

The AccountIDs in the account and issuer fields are presented without a length prefix. When the currency is XRP, the currency code is represented as 160 bits of zeroes.

Each step is followed directly by the next step of the path. As described above, last step of a path is followed by either 0xff (if another path follows) or 0x00 (if this ends the last path).

The following example shows the serialization format for a PathSet:

PathSet is several paths each followed by a continue or end byte; each path is several path steps consisting of a type byte and one or more 160-bit fields based on the type byte

UInt Fields

The XRP Ledger has several unsigned integer types: UInt8, UInt16, UInt32, and UInt64. All of these are standard big-endian binary unsigned integers with the specified number of bits.

When representing these fields in JSON objects, most are represented as JSON numbers by default. One exception is UInt64, which is represented as a string because some JSON decoders may try to represent these integers as 64-bit "double precision" floating point numbers, which cannot represent all distinct UInt64 values with full precision.

Another special case is the TransactionType field. In JSON, this field is conventionally represented as a string with the name of the transaction type, but in binary, this field is a UInt16. The TRANSACTION_TYPES object in the definitions file maps these strings to specific numeric values.