Repo branches


Repo branches

Inspired by Git, repositories support a concept called branches. Branches are useful for many different purposes as illustrated below.


XP provides advanced features such as diffing to see the changes between two branches. Additionally, the API provides features for "pushing" changes from one branch to another. The push operation automatically handles dependencies and and missing parent items to ensure the result is consistent.

From the CMS API, the push operation is known as "publish".

XP repos currently don’t offer conflict resolution or merging functionality. As such, conflict resolution must either be handled by the application itself, or the application must write data in a way that avoids creating conflicts.

Example usage

Consider the 'Oslo' and 'Enonic' nodes below:


There will be two node-versions in the repository, with data persisted in both blobstore and search index:

node versions

A node-version is a representation of a node’s properties. A node-version has no knowledge of name, parent or other meta-data: just the properties of a node. At the same time, the targeted branch (named 'draft' in this example) gets two entries:

branch initial

The node-versions are now a part of a tree-structure, based on the node’s name and parent. If we push the content of branch 'draft' to the default branch 'master', we end up with something like this:

branch push

At the moment, there are two branches pointing to the same node-versions. This means that a single node version can exist in several branches with different structures. Now, consider that the 'oslo' - node is updated and stored to the 'draft'-branch, resulting in a new node-version with the same id and an updated pointer from the branch:

branch diff

The two branches now point to different node-versions of the 'oslo' node. Again, doing a push-operation from 'draft' to 'master' will result in both nodes pointing to the same node-versions:

branch push 2