Binary Merkle Tree operations on data
| Feature | Value |
|---|---|
| Technology |     |
| Type Checking |   |
| CI/CD |     |
| Docs |  |
| Package |    |
| Meta |     |
- Install using
pip
pip install bmt_pyUsage
>>> from bmt_py import make_chunk
>>> payload = bytes([1, 2, 3])
>>> chunk = make_chunk(payload)
>>> result = chunk.address()
>>> print(bytes_to_hex(result, 64))
# ca6357a08e317d15ec560fef34e4c45f8f19f01c372aa70f1da72bfa7f1a4338- Chunking with Payload Lesser Than 4KB
from bmt_py import make_chunked_file
payload = bytes([1, 2, 3])
chunked_file = make_chunked_file(payload)
print(len(chunked_file.leaf_chunks()))
# 1
only_chunk = chunked_file.leaf_chunks()[0]
only_chunk.span() == chunked_file.span()
# True
only_chunk.address() == chunked_file.address()
# True- Chunking with Payload Greater Than 4KB
from bmt_py import make_chunked_file, get_span_value, bytes_to_hex
with open("The-Book-of-Swarm.pdf", "rb") as f:
file_bytes = f.read()
chunked_file = make_chunked_file(file_bytes)
print(get_span_value(chunked_file.span()))
# 15726634
tree = chunked_file.bmt()
print(len(tree))
# 3
print(len(tree[2])) # last level only contains the root_chunk
# 1
root_chunk = tree[2][0]
second_level_first_chunk = tree[1][0] # first intermediate chunk on the first intermediate chunk level
root_chunk.payload[:32] == second_level_first_chunk.address()
# True
print(len(second_level_first_chunk.payload))
# 4096
print(bytes_to_hex(chunked_file.address(), 64))
# b8d17f296190ccc09a2c36b7a59d0f23c4479a3958c3bb02dc669466ec919c5ddef test_collect_required_segments_for_inclusion_proof():
with open("carrier-chunk-blob", "rb") as f:
file_bytes = f.read()
chunked_file = make_chunked_file(file_bytes)
file_hash = chunked_file.address()
# Segment to prove
segment_index = (len(file_bytes) - 1) // 32
# Check segment array length for carrierChunk inclusion proof
proof_chunks = file_inclusion_proof_bottom_up(chunked_file, segment_index)
assert len(proof_chunks) == 2 # 1 level is skipped because the segment was in a carrierChunk
def test_get_file_hash(segment_index):
proof_chunks = file_inclusion_proof_bottom_up(chunked_file, segment_index)
prove_segment = file_bytes[segment_index * SEGMENT_SIZE : segment_index * SEGMENT_SIZE + SEGMENT_SIZE]
# Padding
prove_segment += bytearray(SEGMENT_SIZE - len(prove_segment))
# Check the last segment has the correct span value.
file_size_from_proof = get_span_value(proof_chunks[-1].span)
assert file_size_from_proof == len(file_bytes)
return file_address_from_inclusion_proof(proof_chunks, prove_segment, segment_index)
# Edge case
hash1 = test_get_file_hash(segment_index)
assert hash1 == file_hash
hash2 = test_get_file_hash(1000)
assert hash2 == file_hashdef test_collect_required_segments_for_inclusion_proof_2(bos_bytes):
with open("The-Book-of-Swarm.pdf", "rb") as f:
bos_bytes = f.read()
chunked_file = make_chunked_file(bos_bytes)
file_hash = chunked_file.address()
# Segment to prove
last_segment_index = (len(file_bytes) - 1) // 32
def test_get_file_hash(segment_index):
proof_chunks = file_inclusion_proof_bottom_up(chunked_file, segment_index)
prove_segment = file_bytes[segment_index * SEGMENT_SIZE : segment_index * SEGMENT_SIZE + SEGMENT_SIZE]
# Padding
prove_segment += bytearray(SEGMENT_SIZE - len(prove_segment))
# Check the last segment has the correct span value.
file_size_from_proof = get_span_value(proof_chunks[-1].span)
assert file_size_from_proof == len(file_bytes)
return file_address_from_inclusion_proof(proof_chunks, prove_segment, segment_index)
# Edge case
hash1 = test_get_file_hash(last_segment_index)
assert hash1 == file_hash
hash2 = test_get_file_hash(1000)
assert hash2 == file_hash
with pytest.raises(Exception, match=r"^The given segment index"):
test_get_file_hash(last_segment_index + 1)def test_collect_required_segments_for_inclusion_proof_3():
# the file's byte counts will cause carrier chunk in the intermediate BMT level
# 128 * 4096 * 128 = 67108864 <- left tree is saturated on bmt level 1
# 67108864 + 2 * 4096 = 67117056 <- add two full chunks at the end thereby
# the zero level won't have carrier chunk, but its parent will be that.
with open("carrier-chunk-blob-2", "rb") as f:
carrier_chunk_file_bytes_2 = f.read()
assert len(carrier_chunk_file_bytes_2) == 67117056
file_bytes = carrier_chunk_file_bytes_2
chunked_file = make_chunked_file(file_bytes)
file_hash = chunked_file.address()
# segment to prove
last_segment_index = (len(file_bytes) - 1) // 32
def test_get_file_hash(segment_index):
proof_chunks = file_inclusion_proof_bottom_up(chunked_file, segment_index)
prove_segment = file_bytes[segment_index * SEGMENT_SIZE : (segment_index * SEGMENT_SIZE) + SEGMENT_SIZE]
# padding
prove_segment = prove_segment.ljust(SEGMENT_SIZE, b"\0")
# check the last segment has the correct span value.
file_size_from_proof = get_span_value(proof_chunks[-1].span)
assert file_size_from_proof == len(file_bytes)
return file_address_from_inclusion_proof(proof_chunks, prove_segment, segment_index)
# edge case
hash1 = test_get_file_hash(last_segment_index)
assert hash1 == file_hash
hash2 = test_get_file_hash(1000)
assert hash2 == file_hash
with pytest.raises(Exception, match=r"^The given segment index"):
test_get_file_hash(last_segment_index + 1)- More examples are here.
How it works
First, it splits the data into chunks that have maximum 4KB payload by default, but this condition can modified as well as its span byte length (8 bytes) that indicates how long byte payload subsumed under the chunk.
If the payload byte length cannot fit exactly to this chunk division, the rightmost chunk's data will be padded with zeros in order to have fixed length data for the BMT operations.
This basic unit is also required to effectively distribute data on decentralized storage systems with regard to plausible deniability, garbage collection, load balancing and else. For more details, please visit Etherem Swarm webpage that has full implementation of this logic.
The used hashing algorithm is the keccak256 function that results in a 32 bytes long segment.
Performing BMT hashing on the chunk data will define the BMT root hash of the chunk.
Then, for integrity considerations, the BMT root hash is hashed with the chunk's span from the left which takes the Chunk address.
In order to refer files also with a single 32 byte segment, the chunk addresses of the payload have to be hashed in the same way until the File address:
chunks can encapsulate 128 chunk addresses on the subsequent BMT tree level by default. These kind of chunks are called Intermediate chunks
By the properties of a BMT tree, the chunks will end in a Root chunk that refers all chunks below (directly or indirectly) and its address will be the File address as well.
One can realize, the BMT levels can have an orphan chunk on the rightmost-side that cannot be hashed with a neighbour chunk, because it does not have a neighbour chunk (e.g. 129/129 chunk).
When it occurs, it does not have a sense to hash this orphan chunk on every BMT level since it will be BMT hashed with zero data.
That's why the the algorithm handles orphan chunk as Carrier chunk and tries to place it into that BMT tree level where the chunk address can be encapsulated with other addresses.
This BMT hashing of data allows to reference any file with unified 32 bytes unique address which is called content addressing.
Nevertheless, it also allows to perform lightweight compact inclusion proof on data. This proof requires little amount of data to be provided for proving whether any particular segment (32 bytes) of the data is present at a particular offset under the file address.
This feature allows to create a logic around data referenced by 32 bytes file addresses where the data segment values have to meet some conditions.
The first/best use-case for this can happen via smart contracts that implement the validation functions that check the provided inclusion proof segments are indeed subsumed under the commited file references.
To get these inclusion segments, the library collects all required segments from the BMT tree that can be used for input of smart contract validation parameters.
Documentation: https://aviksaikat.github.io/bmt-py/
Source Code: https://github.com/Aviksaikat/bmt-py
Development
We use Hatch to manage the development environment and production build. Ensure it's installed on your system.
You can run all the tests with:
hatch run testExecute the following command to apply linting and check typing:
hatch run lintYou can bump the version, create a commit and associated tag with one command:
hatch version patchhatch version minorhatch version majorYour default Git text editor will open so you can add information about the release.
When you push the tag on GitHub, the workflow will automatically publish it on PyPi and a GitHub release will be created as draft.
You can serve the Mkdocs documentation with:
hatch run docs-serveThis project is licensed under the terms of the BSD-3-Clause license.