No, though I did download Visual studio and the sqlite3 DLL, and got them running together. I am mainly interested in what might be under the hood. I know from reading the sqlite3 newsletter that some very able programmers are on the task.
I am narrowing in on the 3-tuple: operator,keyword,link which I would characterize as the fundamental instruction format for graph traversal. I would hope the ontologies are maximized so that the great bulk of traversals are within one single nested list of these instructions. Thus the sqlite can maintain these nested lists in object stores, and large searches will minimize the disk access waits. It is all about getting subgraphs trimmed such that their object stores meet the -iLog(i) criteria, where i is the probability of access.
So at the heart of the machine is the simple execution of the cross between two instructions, one from the small search ontology and one from the corporate mesh. Instruction by instruction, we are trying to fit the client supplied ontology to the corporate mesh. We traverse the links where the two keywords and two operators result in a conclusion, in a node by node comvolution.
[My ideas here are not original, and I can locate reference work when I get the time]
I should sign up for the sqlite3 news letter again and spy.
No comments:
Post a Comment