skip to navigation
skip to content

extpickle 1.20160617.170648

extended pickle

Extends the original Pickler class to be able to pickle some otherwise non-supported types. The emphasis is to be fast and for communication via pipes/sockets with the same Python version on the other end - thus we don’t care that much for compatibility with other versions.

A similar project is dill, which is much bigger though.

This project is registered on PyPI and can be installed via

pip install extpickle

For some usage examples, see the test code.

Over the base Pickler class, it adds pickling support for:

  • types.FunctionType, types.CodeType and cell-types. I.e. you can pickle lamdas and functions. Note this will use the marshal module for the byte-code, so this is not portable across different Python versions.
  • Modules. This is done by referencing it via its name. Note that is has some extra handling for __main__, which is allowed to be a different module.
  • mod.__dict__ where mod is a module is also stored just as a reference to the module, not a copy of the dict.
  • (Python 2) buffer.
  • numpy.ndarray. This is also supported by the base class but our implementation is much faster. Actually this is a bit weird because the default implementation should also be just as fast but it isn’t. Our implementation basically just uses fromstring/tostring.
  • New-style classes. The base class would try to look them up in a module. This will fail whenever it cannot be find in such namespace. Our implementation, if that fails, will actually store the information to construct a new class, i.e. the name, the bases and its dict.
  • (Python 2) Old-style classes. This is the same behavior as for the new-style classes.

This was used in the TaskSystem project.

 
File Type Py Version Uploaded on Size
extpickle-1.20160617.170648-py2.py3-none-any.whl (md5) Python Wheel 2.7 2016-06-17 7KB
extpickle-1.20160617.170648.tar.gz (md5) Source 2016-06-17 5KB