After searching (thanks to the ACM for the ACM Digital Library) I discovered several variants of the Lineage method (or model) also called Path Enumeration or Materialized Path (already mentioned by DQNOK), which indeed are part of a family of models called Tree Encodings.

Anyway, now my software fly: not tens of times, not hundreds of times, but thousands of times faster. From several minutes to few milliseconds with minimal memory usage. (Hey, this is a web application, memory usage is critical.)

I also learned a bit about some variants of the nested set model, but honestly, maybe I am too stupid but got a headache trying to understand them. Ok, maybe I’m not stupid, but I was in a hurry and the Path Enumeration thing is so simple and intuitive.

Regards,

MV

p.s. David Chandler’s patent is a joke, let me guess it’s an USA patent.

I agree with DQNOK, using patented algorithms is not that simple if you want to avoid any violation.

I searched to check if the patent was not really just an application (meaning no patent was granted) but it seems the patent was granted. The algorithm is so simple, why did they do this?

Urgh.

Anyway, back to the rewrite my code. I think the mptt will be my answer.

Regards,

MV

To update the nested set table directly, one only has to update the left/right indexes appropriately. See this article at mysql.org for more information on how: http://dev.mysql.com/tech-resources/articles/hierarchical-data.html

How can it be an option when it is patented?

I’ve not used patented software before (unless perhaps by accident because the patent office was so obtuse at to grant a patent to something that any reasonable programmer like myself just “came up with”), but it occurs to me that attempting to manage the rights to use it would become an absolute nightmare unless you KNEW it was only going to be used in one app, and you yourself entirely controlled the distribution. Even then, without some automated way of keeping your distribution mechanism synchronized with your patented code use, you still run the risk of being in violation. A violation COULD spell financial ruin for your employer, and ultimately for yourself.

I appreciate your point about it being a valid technical option in certain limited circumstances. But for me, the complexity of introducing patented software into my projects far outweighs any performance benefit I might get. In my app with over a quarter million records, the Lineage method (or “path enumeration method”) yielded queries so fast I didn’t even bother to time them. Virtually instantaneous.

Thanks for the article though. Good explanations and diagrams. I appreciate the work you put in on it.

The two comments I’ve received for this post are essentially examples in which Chandler’s method will not work. That still does not negate the usefulness of it. It is one of many options, including the mptta and the path enumeration model.

This method is effectively the “Lineage” method mentioned elsewhere, but is more limited as the max depth must be known ahead of time. In the Lineage approach, an extremely large VARCHAR can be used which will allow MANY layers, but will only use the amount of storage actually required. This patented approach ALWAYS stores the maximum depth, even when it’s not needed. I considered this method for a BOM (Bill Of Materials) program (before I knew the stupid approach was patented!), but ultimately decided it was way too wasteful, and went with the Lineage approach instead. Presently I’m considering rewriting the app to use the Nested Set (or MPTTA) approach.