On Fri, Jan 21, 2011 at 12:29, Richard Quadling <rquadling@xxxxxxxxx> wrote: > Changing data in a database is the role of the database engine. It is > much more efficient to have the cost on the insert than it is on the > select. > Agreed. On insert I could even delegate the operation to another thread which does not timeout with the pageload. > The adjacent list model is very expensive at n-levels for the select, > but trivial cost for the insert. If you are inserting millions of rows > but only occasionally looking at the data, then stick with the > adjacent list model. But if tags and n-levels are regularly accessed > and form a main part to the functionality of the app, then you may > want to reconsider. > I've already reconsidered after some sleep and coffee! > Sure, the insert for the nested set model is more expensive in terms > of the number of rows to amend, but indexing will certainly should > certainly help. If you have tools to help optimize the tag table and > the queries you use, then I'd follow the recommendations (I use MS > SQL, so my Query Optimization tools help me here). The nested set > model is extremely efficient on the select. > Interesting. I am using MySQL for this application, but another hat I wear is learning C# with MS tools and I will have to look into the Query Optimisation. > It is a trade off that you have to decide upon, based upon your data > and needs. If, as I suspect, you are going to be doing a LOT of > selects on the tags and (in the future) to multiple levels, then this > aspect needs to be very efficient. > You suspect correctly. > For me it is well worth the effort of moving from the adjacent list > model to the nested set model. > > Both mechanisms work. In my opinion, the adjacent list model is for > truly simply lookups, not for complicated n-levels. > > One of the changes I made to the nested set model was for a Bill Of > Materials module. The client made complex machinery (industrial > lathes). The sum quantity for all the parts were in the 20,000 region. > Each nut, bolt, screw, etc. Many sub-assemblies. The parent/child node > logic was massive in dealing with retrieving questions like "How many > machines can we build?", "What stock do we need to buy/make to > complete an order of 20 lathes?". Lot's of recursion into each level > to build the list. Getting the results would take 3 or 4 minutes (this > is in a non SQL environment using a peer-to-peer modified D-ISAM > database - it was already slow because of all that). When I moved to > the nested set model, no recursion and 1 query (more or less) and I > have all the results I needed. It was seconds in comparison. > Thanks, I enjoy reading these real-life scenarios. This was a terrific example. -- Dotan Cohen http://gibberish.co.il http://what-is-what.com -- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
