I have a tree structure in my appplication that needs to be persisted to a file locally. I'm looking for some input as to what would be a good way of implementing this. The tree is basically a file tree, with FileNode and FolderNode objects extending a Node object. All Nodes have a reference to their parent, and FolderNodes also have a Set of children. The root is a FolderNode with parent set to null. The tree could have anything between 2 and 100 000s of nodes. Width and depth can be whatever is sensible for a file tree (usually more wide than deep). Application work flow: Load the whole tree into memory (by reading the root from a file and recursively populating the tree) While the tree still has nodes: Process and then remove a leaf node Write the tree to the file (updating the file to reflect the tree in memory) Delete the file As you can tell, the file is only read once, but written after EVERY change, so performance is clearly an issue here. What would be a good solution to this? It feels like JPA with an embedded database like SQLite, Derby or ObjectDB would be the best choice. Is it? And in that case, how should I model the database?

Reputation: 1221

Persisting a tree layout to file. How to implement effectively?

I have a tree structure in my appplication that needs to be persisted to a file locally. I'm looking for some input as to what would be a good way of implementing this.

The tree is basically a file tree, with FileNode and FolderNode objects extending a Node object. All Nodes have a reference to their parent, and FolderNodes also have a Set of children. The root is a FolderNode with parent set to null.

The tree could have anything between 2 and 100 000s of nodes. Width and depth can be whatever is sensible for a file tree (usually more wide than deep).

Application work flow:

Load the whole tree into memory (by reading the root from a file and recursively populating the tree)
While the tree still has nodes:
- Process and then remove a leaf node
- Write the tree to the file (updating the file to reflect the tree in memory)
Delete the file

As you can tell, the file is only read once, but written after EVERY change, so performance is clearly an issue here.

What would be a good solution to this? It feels like JPA with an embedded database like SQLite, Derby or ObjectDB would be the best choice. Is it? And in that case, how should I model the database?

Upvotes: 2

Answers (2)

parasietje

Reputation: 1539

The proper answer is based on what you want to do with the file afterwards. As it is explained here, the best way is to save the thing to /dev/null (NOT save it), because you aren't using it afterwards.

First try and see whether it is really necessary to save it separately. A java.io.File only uses 100 bytes, so it may be worth the effort to just keep the whole thing in memory and only write it out once.

Upvotes: 0

Rajesh Pantula

Reputation: 10241

Actually I should put this in comment bcos this wont be a proper answer to your question, but sorry for putting it in answer.

The tree is basically a file tree, with FileNode and FolderNode objects extending a Node object. All Nodes have a reference to their parent, and FolderNodes also have a Set of children. The root is a FolderNode with parent set to null.

I would suggest you rather than re-inventing the wheel and implementing this yourself. take a look at JCR(Java Content Repository), Apache Jackrabbit is an implementation of JCR, it has the advantages of both a file system and a hierarchy structure, and also has other features like versioning.

Also, it has an inbuilt persistence Manager, which will take care of persisting your tree.

Upvotes: 2

Persisting a tree layout to file. How to implement effectively?

Answers (2)

Related Questions