C/C++: How to store data in a file in B tree

Question

It appears to me that one way of storing data in a B-tree as a file can be done efficiently with C using binary file with a sequence (array) of structs, with each struct represe

Answer 1

For implementing B-Trees in a file, you can use the file offset instead of pointers. Also, you can implement a "file memory manager", so that you can re-use deleted items in the file.

In order to fully recover the deleted blocks in a B-Tree file, you will have to recreate the B-Tree in a new file. Also remember the most OSes have no methods for truncating files. A portable method for truncating a file is to write a new file and destroy the old.

Another suggestion is to partition the file into B-Tree partition and data (item) partition. A B-Tree partition will contain the pages. The leaf pages will contain offsets to the data items. The data partition will be a section in the file containing data items. You may end up creating more than one of each partition and the partitions may be interleaved.

I spent much time playing with a file based B-Tree, until I gave up and decided to let a database program (or server) handle the data for me.

Answer 2

You can use Berkley DB as well. It works well with C programs and implements B+ tree.

Answer 3

I did a very quick search and dug up this: http://people.csail.mit.edu/jaffer/WB C source: http://cvs.savannah.gnu.org/viewvc/wb/wb/c/ - it seems to offer disk-based B-tree style databases - although taking a look at "delete.c" it seemed to imply if you delete a node everything down from it would be taken out - if that's the correct behaviour then it sounds like something that might help?

Also - B-trees are often used in filesystems - could you not take a look at some filesystem code?

My own inclination is that of a file-system - if you have a B-tree of fixed-size, whenever you "delete" a node rather than attempting to remove the reference, just set the value to whatever means nothing in your code. Then, have a clean-up thread running that checks if anyone has the file open for reading and if all's quiet blocks the file and tidies up.