4 ms·
You could also start from first principles and essential abstractions. If you build a solid key-value store, it is possible to construct higher-order stores on
by bob1029 5y ago
You could also start from first principles and essential abstractions. If you build a solid key-value store, it is possible to construct higher-order stores on top.
Once you can quickly find a thing with some key, any degree of structure and complexity is possible from that point. Column-based, row-based, document-based, etc. All are possible to build on top of K-V. Might not be the most optimal solution, but it is a very accessible path in my experience. You can start defining types like Table, Column, Row, Document, etc. which all leverage this idea that you can trivially refer to another thing by some identifier. For instance, Table.FirstRowId and Table.LastRowId could be used to refer to the logical extents of a collection of rows that might be linked together both ways (e.g. Row.PreviousRowId, Row.NextRowId).
- winrid 5y agoThat's definitely a good approach. I'm fortunate enough that I can learn just by reading, so I like to just lay down and read coding/comp-sci stuff when I can. Like, If I wanted to, I bet I know I could create the server, connection manager, query parser, optimizer, WAL, storage... and maybe I will, but would just be fine to read about someone else doing it. I spend enough time coding as it is. :)
- SilverRed 5y agoThis would be basically doing all of the research and learning on databases from the last few decades from scratch. It probably won't give you a good understanding of how a modern DB works though.