How are Tezos operations encoded?

Question

Given a transaction in its serialized format, how do you parse it to get the destination, the amount, and the other fields?

I am not asking for library or RPC endpoint to achieve this but for an explanation of how each field is encoded in the serialized raw hex.

Here is an example of a raw hex:

008f1d96e2783258ff663f03dacfe946c026a5d194c73d1987b3da73fadea7d46c008cb5baedee4dc3ec261dfcf57a9600bb0a8e26c0f00bdd85a0018452ac02e0a712000153957451d3cc83a71e26b65ea2391a1b16713d2d009595facf847a72b4c3fe231c0e4185e68e9b2875aa3c639382c86bcf0af23699f47fe66a6550ade936a5b59d5919ad20703885750314e0c368b277de39e7d10a

Answer 1

The structure of an encoded operation group is the following:

<branch : 32 bytes>
<tag#0 : nat> | <content#0>
... 
<tag#n : nat> | <content#n>
<signature (optional) : 64 bytes>

Extracting branch:

1. Append 0x0134 prefix
2. Base58 encode with checksum

Extracting signature:

1. Append 0x04822b prefix
2. Base58 encode with checksum

Operation tag by kind (as of Carthage):

• 0 for endorsement
• 5 for proposal
• 6 for ballot
• 1 for seed_nonce_revelation
• 2 for double_endorsement_evidence
• 3 for double_baking_evidence
• 4 for activate_account
• 107 for reveal
• 108 for transaction
• 109 for origination
• 110 for delegation

Content encoding depends on the operation kind, in your question you were likely referring to "transaction", so let's take it as an example:

<source : key_hash>
<fee : nat>
<counter : nat>
<gas_limit : nat>
<storage_limit : nat>
<amount : nat>
<destination : address>
<has_parameters? : bool>
<parameters (optional)>

"has_parameters" flag is False if at least one of the conditions is met:

• Transaction content does not contain "parameters" field
• "parameters" field has value {"entrypoint": "default", "value": {"prim": "Unit"}}

Parameters are encoded as follows:

<entrypoint tag : 1 byte> | <entrypoint name (optional)>
<encoded value length : 4 bytes, big endian> | <value : micheline>

Entrypoint tag depends on the entrypoint name (can be reserved):

• 0x00 for default
• 0x01 for root
• 0x02 for do
• 0x03 for set_delegate
• 0x04 for remove_delegate
• 0xff otherwise

Entrypoint name is encoded as <encoded name length : 1 byte> | <utf-8 encoded name>

Out of scope of this answer:

• Forging primitives: nat, key_hash, address, bool
• Forging Micheline values

---

Breakdown for the sample opg hex:

branch 
008f1d96e2783258ff663f03dacfe946c026a5d194c73d1987b3da73fadea7d4
BKiXcfN1ZTXnNNbTWSRArSWzVFc6om7radWq5mTqGX6rY4P2Uhe

tag
6c
transaction

source
008cb5baedee4dc3ec261dfcf57a9600bb0a8e26c0
tz1YU2zoyCkXPKEA4jknSpCpMs7yUndVNe3S

fee
f00b
1520

counter
dd85a001
2622173

gas_limit
8452
10500

storage_limit
ac02
300

amount
e0a712
300000

destination
000153957451d3cc83a71e26b65ea2391a1b16713d2d
tz2FwBnXhuXvPAUcr1aF3uX84Z6JELxrdYxD

has_parameters
00
false

signature
9595facf847a72b4c3fe231c0e4185e68e9b2875aa3c639382c86bcf0af23699f47fe66a6550ade936a5b59d5919ad20703885750314e0c368b277de39e7d10a
sighZMqWz5G8drK1VTsmTnQBFEQ9kxQQxL88NFh8UaqDEJ3R3mzgR3g81azadZ9saPwsWga3kEPsyfbzrXm6ueuDvx3pQ5Q9