gitdb2

This is a library that extends sqlalchemy databases to store their content in a directory of files that is managed by git. This makes synchronizing databases much easier.

Installation

gitdb2 uses [libgit2](https://libgit2.github.com/) via [pygit2](http://www.pygit2.org/) to handle the git repository. See [http://www.pygit2.org/install.html] for explanations how to install the most recent version of libgit2 and pygit2. All required python packages are listed in requirements.txt. gitdb2 is installable via pip or python setup.py install.

Example

In this simple example, a gitdb repo for a database with a single table will be constructed.

import sqlalchemy as sa
from sqlalchemy import Column, Integer, String, DateTime, Boolean, ForeignKey, ForeignKeyConstraint
from sqlalchemy.orm import relationship, backref, attributes
import sqlalchemy.ext.declarative
from gitdb2 import *

Base = sqlalchemy.ext.declarative.declarative_base()

class Test(self.Base):
    __tablename__ = 'test'
    id = Column(Integer, primary_key = True)
    foo = Column(String)
gitDBRepo = GitDBRepo.init(Base, 'test')
session = gitDBRepo.session
test = Test()
test.foo = 'probe'
session.add(test)
#The commit will save the object to an file and commit it in git.
session.commit()

More complex example

This advanced example shows that gitdb is able to handle relationships and association_proxies.

import sqlalchemy as sa
from sqlalchemy import Column, Integer, String, DateTime, Boolean, ForeignKey, ForeignKeyConstraint
from sqlalchemy.orm import relationship, backref, attributes
import sqlalchemy.ext.declarative
from gitdb2 import *

Base = sqlalchemy.ext.declarative.declarative_base()
class Test1(Base):
    __tablename__ = 'test1'
    id = Column(Integer, primary_key = True)
    col2 = Column('komischer_name', String)
    test2s = association_proxy('rels', 'test2')

class Test2(Base):
    __tablename__ = 'test2'
    id = Column(Integer, primary_key = True)
    col2 = Column(String)
    test1s = association_proxy('rels', 'test1')

class Rel(Base):
    __tablename__ = 'rels'
    test1_id = Column(Integer, ForeignKey('test1.id'), primary_key=True)
    test2_id = Column(Integer, ForeignKey('test2.id'), primary_key=True)
    test1 = relationship(Test1, backref=backref('rels', cascade="all, delete-orphan"), passive_updates=False)
    test2 = relationship(Test2, backref=backref('rels', cascade="all, delete-orphan"), passive_updates=False)
    def __init__(self, test2=None, test1=None):
        self.test2=test2
        self.test1=test1

gitDBRepo = GitDBRepo.init(Base, 'test')
session = gitDBRepo.session
test1 = Test1()
test1.col2 = 'bla'
test1.test2s.append(Test2())
session.add(test1)
#The commit will save all the objects to files and commit in git.
session.commit()
#change attribute
test1.col2 = 'blub'
session.commit()

#changing primary keys does work as well, the files will be moved.
test1.id = 4
session.commit()

Kown limitations

• In sqlite, one should use "passive_updates=False" for relationships, as sqlite does not cascade primary_key-updates. Also, if the database system does the updates, GitDB probably does not recognizes it (untested).
• Also, many2many relationships via a secondary table do not work in GitDB, as GitDB has no access to the secondary table and thus cannot store it. Use association_proxies instead. Do not forget to set 'cascade="all, delete-orphan"' in order for association_proxy.remove() to work.
• Bulk updates and bulk deletes are not supported at the moment (i.e., Query.update(), Query.delete(). This is because GitDB cannot get the precise rows updated or deleted. GitDB will raise an NotImplementedError if bulk updates or bulk deletes occur in its session. In a later version, this might be overcome by explicitly checking all objects of the respective table and e.g., delete all files without a table row.