This feed contains pages about my life in obsearch.
Eclipse was seg faulting on me for the entire day yesterday. The reason was unexpected. I was using java 1.6 instead of 1.5. They recommended 1.5, but I did not expect this would have such a great impact. The problem was caused by either subclipse and subversive, when commiting results. The interesting is that I was using SVNKit, a java implementation of SVN (no JavaHL). Now everything is fixed and stable.
Posted Sat 19 Jul 2008 11:32:43 AM JSTStatus:
- I really need now to support all the primitive data-types so I want to reduce the size of the templates to the maximum possible.
- I want to add long support for each object. Now we can have long objects in the DB. [done]
- Re-implement the P+Tree so that the abstraction utility is used along with the bucket implementation. [done] Use buckets to reduce disk access.
- Scrap the jxta index and leave everything ready for the Map-reduce and Hyper-table indexes. [done]
- Change the package name, and remove lots of things that add noise to the project. [done]
- Remove Oracle's code and use instead Java's native byte array operations. [done]
Modification of OBSearch is still underway. Lots of changes have been implemented. Currently none of the previous indexes work. The changes are meant to reduce the amount of code necessary when a new index is introduced. I have also decided to make heavy use of ByteBuffer, a class that allows efficient transfer of primitive values to and from byte arrays.
Posted Thu 03 Jul 2008 07:19:42 PM JSTI am going to do major refactoring on OBSearch.
- I really need now to support all the primitive data-types so I want to reduce the size of the templates to the maximum possible.
- I want to add long support for each object. Now we can have long objects in the DB.
- Re-implement the P+Tree so that the abstraction utility is used along with the bucket implementation. Use buckets to reduce disk access.
- Scrap the jxta index and leave everything ready for the Map-reduce and Hyper-table indexes.
- Change the package name, and remove lots of things that add noise to the project.
- Remove Oracle's code and use instead Java's native byte array operations.