kizzx2 / sbe-python

Easy-to-use Python SBE decoder and encoder
https://pypi.org/project/sbe
MIT License
19 stars 15 forks source link
fix sbe

sbe-python

Easy to use, fast, pure Python FIX SBE encoder and decoder.

Install

pip install sbe

Usage

Simple Decoding

Decode SBE to Python dictionaries in one line. Good for exploratory data analysis.

import sbe

with open('your-schema.xml', 'r') as f:
  schema = sbe.Schema.parse(f)

wtih open('your-data.sbe', 'rb') as f:
  buf = f.read()

# Get a Python dict in one-line
x = schema.decode(buf)

x.name  # The template message name

x.value
# {'userId': 11,
# 'timestamp': 1598784004840,
# 'orderSize': 0,
# 'price': 5678.0,
# ...

# If you need an initial offset, apply it Pythonically
schema.decode(buf[19:])

# decode_header to filter out messages based on header to avoid decoding
# message bodies that are not needed
schema.decode_header(buf)['templateId']

High Performance Decoding (Wrapping)

This gives you decent performance while still retaining high code readability.

import sbe

with open('your-schema.xml', 'r') as f:
  schema = sbe.Schema.parse(f)

wtih open('your-data.sbe', 'rb') as f:
  buf = f.read()

# Wrap the buffer without decoding it, fields are converted to Python variables
# on demand
x = schema.wrap(buf)

x.header['templateId']
x.body['price']
x.body['someGroup'][2]['price']

Direct Access with Pointers

get_raw_pointer gives you the required information to unpack a variable from memoryview / bytes. This gets you very close to the fastest achievable performance in Python:

import sbe

with open('your-schema.xml', 'r') as f:
  schema = sbe.Schema.parse(f)

header_pointer = schema.header_wrapper.get_raw_pointer('templateId')

# Let's say we are only interested in messages of templateId == 3
price_pointer = schema.message_wrappers[3].get_raw_pointer('price')

wtih open('your-data.sbe', 'rb') as f:
  buf = f.read()

# pass `memoryview` to `unpack` to avoid copying
buf = memoryview(buf)[initial_offset:]
template_id = header_pointer.unpack(buf)  # calls buf[offset:offset+size].cast("I")[0] directly

if template_id == 3:
  print(price_pointer.unpack(buf))

Encoding

import sbe

with open('./your-schema.xml', 'r') as f:
  schema = sbe.Schema.parse(f)

# message_id from the schema you want to encode
message_id = 3

obj = {
  'userId': 11,
  'price': 5678.0,
  # ...
}

# Encode from Python dict in one-line
schema.encode(schema.messages[3], obj)

# You can supply your header values as a dict
schema.encode(schema.messages[3], obj, headers)