UXTO DB structure

Question

After update Bitcoin Core to version 0.15, chainstate database changed format of data.

Old prefix of key has been 0x63, now 0x43, and key/value are different

Where is the new format described, or can someone have work on this?

UPDATED

Sample

Key:

43000000dcf3713a29f3da53dbde07ea93ace52a4393ba9a009d8ac17c853c18e500

Value:

d42ade0dfd6a93b6ecbc6098336feedaa21b79128e29d5b7ee2d42

---

Prefix BYTE:

0x43

TransactionId:

e5183c857cc18a9d009aba93432ae5ac93ea07dedb53daf3293a71f3dc000000

Index VARINT:

0x00 -> 0x00(0)

---

Code VARINT:

0xD42A -> 0x2AAA(10922)

---

Block Height:

Code >> 1 = 0x1555(5461)

Is Coinbase:

Code & 1 = 0x00(0)

---

Value VARINT:

0xDE0D -> 0x2F8D(12173) -> decompressMSB128 -> 0x21084(135300)

Type VARINT:

0xFD6A -> 0x3F6A(16234)

Answer 1

Each entry in the new 0.15 format is defined as outpoint:coin, and has the following structure:

Outpoint is formed by: key | tx_hash | index.

Where the key corresponds to b'C', or 43 in hex. The transaction hash in encoded in Little endian, and the index is a base128 varint. The corresponding Bitcoin Core source code can be found here.

On the other hand, a coin is formed by: code | value | out_type | script.

Where code encodes the block height and whether the transaction is coinbase or not, as 2*height + coinbase, the value is a txout compressed base128 varint, the out_type is also a base128 varint, and the script is the remaining data. The corresponding Bitcoin Core source code can be found here.

Version 0.15 still used an obfuscation key to avoid triggering anti-virus software, you can check this answer to know how it is stored in the database and how to use it to de-obfuscate you data.

Finally, if you need a decoder for both versions 0.14 (0.8-0.14) and the new 0.15 format, you can check this GitHub repo (particularly functions decode_utxo and decode_utxo_v08_v014 from utils.py).

Disclaimer: I'm the author of the repo.