`get_prefix()` hardcodes 2-byte prefix — breaks for multi-byte varint codes

get_prefix() returns multihash[:2], assuming the code and length varints are each exactly 1 byte. For hash codes > 127 (e.g., sha2-224 at 0x1013, blake2b-256 at 0xB220), the varint encoding is multi-byte, so [:2] returns incorrect/incomplete data.

Problem

In multihash/multihash.py:

def get_prefix(multihash):
    if is_valid(multihash):
        return multihash[:2]    # ← Always returns first 2 bytes
    raise ValueError("invalid multihash")

For codes that fit in a single varint byte (0x00–0x7F), this works. But for codes > 0x7F, the varint encoding uses multiple bytes:

The "prefix" of a multihash is the varint-encoded code + varint-encoded length. Its size varies depending on the code and digest length.

Proposed Solution

Replace the hardcoded slice with proper varint decoding:

def get_prefix(multihash):
    if not isinstance(multihash, bytes):
        raise TypeError("multihash should be bytes")
    if not is_valid(multihash):
        raise ValueError("invalid multihash")

    buffer = BytesIO(multihash)
    # Read code varint
    varint.decode_stream(buffer)
    # Read length varint
    varint.decode_stream(buffer)
    # Everything up to current position is the prefix
    return multihash[:buffer.tell()]

Add tests with multi-byte varint codes:

def test_get_prefix_multi_byte_code():
    mh = sum(b"test", Func.sha2_224)
    prefix = get_prefix(mh.encode())
    assert prefix == mh.encode()[:3]  # 2-byte code + 1-byte length

Related

• File: multihash/multihash.py, get_prefix() function
• Affected codes: All codes > 0x7F (sha2-224, sha2-512-224, sha2-512-256, ripemd-, blake2b-, blake2s-, keccak-, skein*, etc.)

Start in multihash/multihash.py at get_prefix() and inspect the existing is_valid() and varint.decode_stream() behavior. Add coverage for multi-byte codes such as sha2-224 and blake2b-256, then verify that the returned prefix includes the complete varint-encoded code and length while existing single-byte cases still pass.