HistoryDict

HistoryDict is a light weight mapping backed by sqlite which remember the history of values. The intent is for tracking relatively complex configurations.

Conda Recipes

Install the most recent tagged build: conda install historydict -c conda-forge

conda-forge feedstock: https://github.com/conda-forge/historydict-feedstock

Examples

The interface is very simple and can be accessed either via [] like a dictionary

In [1]: from historydict import HistoryDict

In [2]: fn = '/tmp/testing'

In [3]: hist = HistoryDict(fn)

In [4]: hist['key'] = 'foo'

In [5]: hist['key']
Out[5]: 'foo'

or via past

In [7]: hist.past('k')
Out[7]: 'bar'

Using past you can extract past values

In [8]: hist['key'] = 'baz'

In [9]: hist['key'] = 'buz'

In [10]: hist['key'] = 'aardvark'

In [11]:

In [11]: hist['key']
Out[11]: 'aardvark'

In [12]: hist.past('key', 1)
Out[12]: 'buz'

In [13]: hist.past('key', 2)
Out[13]: 'baz'

In [14]: hist.past('key', 3)
Out[14]: 'foo'

All MutableMapping methods work.

There is also a flush method, which ensures that any mutable values such as dicts or lists are synced with the underlying file. This method is automatically called when a HistoryDict object is deleted, but it can be called manually be the user at any time.

Known limitation

The keys are stored in the underlying sqlite table via the hash of their string representation. This means things like:

hist[{'a': 123, 'b': 456}] = 'cat'

will work without error, but in python 3 will only get the same hash half of the time. Only use keys which have stable and unique string representations.

The values are stored via a json blob, thus only values which can be converted to json can be stored (no numpy arrays currently).