pyrazobill

Python client library for the Razobill telemetry decomposition engine. It splits unstructured log lines into structured templates, numbers, and words, and restores original strings bit-for-bit. Built with a high-performance C++20 core and includes a resilient pure-Python fallback.

Lossless invertibility

Traditional log parsers discard formatting, alter timestamps, or leak unseen variable tokens into cluster definitions. Razobill enforces an algebraic invariant across every parse:

$$\forall m \in \Sigma^, \quad \text{RzJoin}(\text{RzSplit}(m)) \equiv m$$

Every byte of whitespace, delimiter punctuation, and parameter data is preserved with zero information destruction.

Quick installation

Install the precompiled binary wheel from PyPI:

pip install pyrazobill

Pre-built binary wheels support Python 3.9 through 3.13 on Linux (manylinux), macOS (arm64, x86_64), and Windows (amd64).

Public API reference

The pyrazobill module exports five primary functions with zero memory leaks:

Usage

Splitting and joining lines

import pyrazobill as rz

# 1. Parse a log message
line = "Accepted publickey for ubuntu from 192.168.1.100 port 54322 ssh2"
record = rz.RzSplit(line)

print(record.template)
# 'Accepted publickey for {s} from {s} port {i} {s}'

print(record.ints)
# [54322]

print(record.strs)
# ['ubuntu', '192.168.1.100', 'ssh2']

# 2. Compute deterministic 64-bit template hash
print(rz.RzHash(record.template))
# 64-bit integer matching ClickHouse xxHash64() exactly

# 3. Restore the exact original text
restored = rz.RzJoin(*record)
assert restored == line

Batch processing with multi-threading

Use RzSplitMany to parse collections of messages in parallel:

import pyrazobill as rz

lines = [
    "Worker 1 finished job in 12 ms",
    "Worker 2 finished job in 19 ms",
    "Worker 3 finished job in 8 ms",
]

records = rz.RzSplitMany(lines)
for rec in records:
    print(rec.template, rec.ints)

RzSplitMany releases the Python GIL, allowing background worker threads to parse chunks simultaneously.

Scientific provenance

Razobill is grounded in information-theoretic invariants established in:

Atta, Moustafa Mahmoud (2026). Latitude and Doubt: An exact ledger of meaning in messages with measurements on the genetic code and on English. Zenodo. DOI: 10.5281/zenodo.23287286

Documentation

• Python developer guide
• Python API reference
• Lossless invertibility
• ClickHouse & chDB guide