Building columnar compression in a row-oriented database

Re:Building columnar compression in a row-oriented database by mamcx on Thursday October 31, 2019 @06:10PM (#21414287)

I'm building a relational lang (that could feel like a in memory db) and explored the idea of using a columnar backed structures.

How well could be apply the same ideas for in-memory processing?

If I understand correctly, you have something alike:

- Store each column on a array of N=1000 - Store the group of columns in pages, with metadata of ranges of keys to locate rows in the adequate page

[ Parent | on HN ]

Re:Building columnar compression in a row-oriented database by ants_a on Friday November 01, 2019 @04:56AM (#21417146)

This is commonly done in graphics and physics processing engines. In that domain it's the array-of-structures vs. structure-of-arrays design choice.

[ Parent | on HN ]

Re:Building columnar compression in a row-oriented database by wmf on Thursday October 31, 2019 @07:18PM (#21414780)

[ Parent | on HN ]