README.rst

Fixle: fast, fixed-length, integer-keyed db via tokyo-cabinet

Author:	Brent Pedersen (brentp)
Email:	bpederse@gmail.com
License:	LGPL3

Contents

• Implementation
• Usage
• speed
• testing

Implementation

uses cython to wrap the fixed length database api in tokyo-cabinet .

Usage

acts like a list, except for deletions.

>>> from fixle import Fixle
>>> import os

>>> fdb = Fixle('t.fdb', mode='w')
>>> for i in range(10):
...     fdb[i] = str(i)
>>> fdb.append('100')
>>> fdb.extend(['111', '222'])
>>> fdb[1]
'1'

>>> fdb[2:5]
['2', '3', '4']

>>> del fdb[4]
>>> fdb[2:5]
['2', '3']

>>> fdb.items(2, 5)
[(2L, '2'), (3L, '3')]

>>> del fdb

can also be used like a shelve where the values are pickled.

>>> fdb = Fixle('p.fdb', 'w', pickle=True)
>>> for i in range(10):
...     fdb.append({'a': i, 'b': range(i, i + 3)})

>>> fdb[2]
{'a': 2, 'b': [2, 3, 4]}

there is also a class for storing longs/ints:

>>> from fixle import FixleLong
>>> fdb = FixleLong('tl.fdb', 'w')
>>> for i in range(1000):
...     fdb[i] = i * 10000000000

>>> fdb[222]
2220000000000

>>> import os; os.unlink('t.fdb'); os.unlink('tl.fdb')

and another for doubles/floats.

>>> from fixle import FixleDouble

speed

Fixle should be very fast, run

$ python tests/bench.py b

testing

$ python setup.py nosetests

to see it compared to shelve and bsddb. seems to be about 10x faster even when pickling the entries, but NOTE that it must use integer keys.