Question

Say I have two tables (I am using Django, but this question is mostly language agnostic):

Organization(models.Model):
    name = models.CharField(max_length=100)

Event(models.Model):
    organization = models.ForeignKey(Organization)
    name = models.CharField(max_length=100)  

Users are allowed to create both events and organizations. There is the chance that two separate users create organization objects that are supposed to resemble the same real world organization. When someone notices this problem, they should be able to merge the two objects so there is only one organization.

The question I have is this: How do I merge these two organizations in order to ensure I can “unmerge” them if the user incorrectly merged them? Thus, the simple solution of deleting one of the Organization objects and pointing all Events to the other one is not an option. I am looking for very high level guidelines on best practices here.

A few possible solutions:

• Add another table that joins together organizations that have been “merged” and keep track of merges that way
• Add a foreign key field on Organization to point to an organization it was merged with
• Keep copies of all of the original objects as they existed before a merge, using something like django-reversion

Answer 1

Personally, I would go with a solution which uses something like django-reversion. However, if you want to create something more robust and less dependent on 3rd party logic, add a merged_into field to Organization and merged_from field to Event:

Organization(models.Model):
    name = models.CharField(max_length=100)
    merged_into = models.ForeignKey('self', null=True, blank=True)

Event(models.Model):
    organization = models.ForeignKey(Organization)
    name = models.CharField(max_length=100)  
    merged_from = models.ForeignKey(Organization null=True, blank=True)

On merge, you can choose update the events as well. From now on, be sure to redirect all references of “merged_into” organizations into the new organization.

If you want to allow multiple merges (for example: A + B into C, A+C into D, E+F into G and D+G into H), you can create a new organization instance each time and merge both “parents” into it, copying the events instead of updating them. This keeps the original events intact waiting for a rollback. This also allows merging more than 2 organizations into a new one in one step.