This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

Python module to handle bytecode

Project Description

bytearound is a module for assembling and disassembling CPython 2.7.11 bytecode. It provides a representation of bytecode that is easier to modify, create, and inspect than CPython’s internal representation and functionality for going back and forth between this representation and CPython code objects.

An example of how to create code:

from bytearound import ByteAround, Instruction, ops

ba = ByteAround([
    ops.LOAD_CONST('Hello World!'),

And a simple modification:

from bytearound import ByteAround

def f():
    print 'Hello World!'

ba = ByteAround.from_code(f.func_code)
for instr in ba:
    if instr.oparg == 'Hello World!':
        instr.oparg = 'Goodbye World!'
f.func_code = ba.to_code()

Design and limitations

I designed and wrote bytearound to ensure that co == ByteAround.from_code(co).to_code() always holds–that is, converting a Python code object to the bytearound representation and back should give an identical code object. Ensuring that this invariant holds makes it easier to test the code for correctness. The function debug.check() exists to check this invariant.

Unfortunately, there are a number of quirks in the way that CPython generates code objects that turn out to be hard to replicate. To replicate some of these, I added a pessimize= argument to ByteAround.to_code that attempts to faithfully replicate CPython even when not doing so would be a little more efficient, and I created a custom comparison function that ignores a few other known issues. However, it may not turn out to be possible to remove all minor differences using these approaches. Known issues include:

  • CPython computes some parts of the code object before it runs the peephole optimizer, which can cause co_stacksize to be too high (because the peephole optimizer can turn a series of opcodes building a tuple into a single LOAD_CONST opcode). The same issue can also affect the ordering of the co_consts field, apparently because the optimizer adds new constants to the end of the list. Similarly, mathematical operations on constants (e.g. 2 ** 32) may be optimized away by the peephole optimizer, possibly leaving behind unnecessary constants.
  • When singleton objects like None and True are used in a function, CPython adds their name to the co_names field (unnecessarily, because the objects are loaded directly with LOAD_CONST) and adds the constants to the end of the co_consts list. (Normally, co_consts includes constants in order of their first appearance in the function.) However, some other usages of None as a constant are placed in co_consts in the right position.
  • The code object for single-line generator expressions like (f(x) for x in y) has a nonempty co_lnotab field, but the co_lnotab for a function defined like “def f(x): print(x)” is empty. In some other circumstances CPython also generates an unnecessary 0 offset entry in co_lnotab.
  • Large opargs (using EXTENDED_ARG) and large line number offsets are not well-tested and have some known issues, noted in the code.

bytearound has been tested only on Python 2.7.11. Previous releases in the 2.7 series should mostly work, but some changes have been made during the series that impact code objects (e.g. issue 21523).

Similar modules

  • byteplay (unmaintained, has bugs)
  • bytecode (Python 3 only, does not implement stack size computation)
Release History

Release History

This version
History Node


History Node


Download Files

Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
bytearound-0.2.tar.gz (12.5 kB) Copy SHA256 Checksum SHA256 Source May 18, 2016

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting