clayote / gorm

Outdated repo, see LiSE, subdir ``allegedb``

gorm

Object relational mapper for graphs with in-built revision control.

gorm serves its own special variants on the networkx graph classes: Graph, DiGraph, MultiGraph, and MultiDiGraph. Every change to them is stored in an SQL database.

This means you can keep multiple versions of one set of graphs and switch between them without the need to save, load, or run git-checkout. Just point the ORM at the correct branch and revision, and all of the graphs in the program will change. All the different branches and revisions remain in the database to be brought back when needed.

usage

>>> from gorm import ORM
>>> orm = ORM('sqlite:///test.db')
>>> orm.initdb()  # only necessary the first time you use a particular database
>>> g = orm.new_graph('test')  # also new_digraph, new_multigraph, new_multidigraph
>>> g.add_nodes_from(['spam', 'eggs', 'ham'])
>>> g.add_edge('spam', 'eggs')
>>> g.edge  # strings become unicode because that's the way sqlite3 rolls
{u'eggs': {u'ham': {}, u'spam': {}}, u'ham': {u'eggs': {}}, u'spam': {u'eggs': {}}}
>>> del g
>>> orm.close()
>>> del orm
>>> orm = ORM('sqlite:///test.db')
>>> g = orm.get_graph('test')  # returns whatever graph type you stored by that name
>>> g.edge
{u'eggs': {u'ham': {}, u'spam': {}}, u'ham': {u'eggs': {}}, u'spam': {u'eggs': {}}}
>>> import networkx as nx
>>> red = nx.random_lobster(10,0.9,0.9)
>>> blue = orm.new_graph('red', red)  # initialize with data from the given graph
>>> red.edge == blue.edge
True
>>> orm.rev = 1
>>> blue.add_edge(17, 15)
>>> red.edge == blue.edge
False
>>> orm.rev = 0  # undoing what I did when rev=1
>>> red.edge == blue.edge
True
>>> orm.rev = 0
>>> orm.branch = 'test'    # navigating to a branch for the first time creates that branch
>>> orm.rev = 1
>>> red.edge == blue.edge
True
>>> orm.branch = 'master'
>>> red.edge == blue.edge
False

gorm's Issues

Use sqlalchemy

It seems like https://pypi.python.org/pypi/graph-alchemy/ is similar (but not maintained). I wonder if this gorm could take over from that library and also use sqlalchemy to gain more backend support. At least something to think about...