It was really interesting to know QlikView's Data Compression logic, in the form of Data table, Symbol table, Bit-stuffed pointers, and Vectors to track current selections, relevant and not-relevant.
As mentioned by HIC, it's really sensible that this data compression logic goes well with redundant data than discrete one. I don't see any data compression (apart from the bit-stuffed pointers, it may be compressed), i was thinking that there was an algorithm followed to do data compression. But still, this bit-stuffed pointer approach is really amazing, and really surprised how come the traditional BI giants not thinking about this approach?
It would be great insight to know in bit detail about: In-memory or Associative (QlikView), Out-Memory or Joins (Direct Discovery) and the Hybrid (Both in-memory and out-memory) data model. Especially, how the hybrid approach, both Associative and Joins handled together? would be a great help. Please, direct me.
Regards,
-Vijay