/usr/lib/python2.7/dist-packages/pymc/database/__init__.py is in python-pymc 2.2+ds-1.1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 | """
==========================
Database backends for PyMC
==========================
A typical MCMC run will generate thousands of samples, and some application
requires well over 100000 iterations. Keeping all this information in memory
can badly strain the performances of PyMC, and users will find their other
applications slowing down. Moreover, we generally wish to store all or part
of the sampled data for future use. To do so, PyMC offers different storing
strategies:
- `no_trace` : don't keep track of the samples,
- `ram` : keep everything in RAM (default),
- `pickle` : put traces in a dictionnary and pickle it once sampling is over.
- `txt` : keep everything in RAM, and dump samples in txt files once
sampling is completed,
- `sqlite` : store data in a sqlite database,
- `hdf5` : store data in a hdf5 file, using pytables.
Although what happens under the hood is very different from one backend to
another, from the user perspective, there is no difference whatsoever. The only
thing that will change is the output file (if any) generated by the backend.
Writing a new backend
---------------------
Each backend is implemented in a file in the database directory. Each one
of these files define two classes: Trace and Database. A new backend
can easily be implemented by defining new classes and saving them in a
file. Look at base.py for skeleton classes, and the other modules for examples.
"""
__modules__ = ['no_trace', 'txt', 'ram', 'pickle', 'sqlite', 'hdf5', 'hdf52', "__test_import__"]
from . import no_trace
from . import txt
from . import ram
from . import pickle
try:
from . import sqlite
except ImportError:
pass
try:
from . import hdf5
except ImportError:
pass
|