Hi Everyone, Looking at the j, here's what I'm thinking of: "Why not make the database like kdb on j ?"
We need the special database, not the universal one. Or are there any reefs ? Maybe k was optimized for this way of usage ? Besides, it's hard to change lines to columns. The analysis of multivariates is simple, and when the base contains notes, it's difficult to switch over. Maybe I need to present the data some other way ? Say, we have a chart with the quantity of apples people have (fields: id_person and apples), and a chart that shows how many apples each friend gives to them (id_person, id_friend,apples). We have to find out how many apples everyone is going to have after friends share their apples with these people. If one imagines 'in a matrix way', the friends-chart turns into a matrix with 1 and 0 values. Next, the matrix is multiplied to vector (how many apples each friend gives to each one). In a word, it's simple. However, there's a question: is there any optimization for such actions ? I'm afraid, this matrix will gobble all the memory . Or should I still think 'in a matrix way', not 'with notes'? If I think 'notes', I'll have to go thru the first table, and then to look for the notes in the next table and calculate again. Which way is better and less resource-intensive ? Regards, Nick. ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
