We all know that no one really spends much time hand-writing SQL anymore. Whether you think it’s a good idea or not (and we might just get in to that, too) the majority of people out there who aren’t using PHP are writing applications using ORMs. Hibernate, JDO, JPA, SQLAlchemy, ActiveRecord. It’s what people are doing, and it allows them to focus on writing the objects their application needs.
But there is an efficiency problem, and many of the real performance nuts get hot and bothered when confronted with ORM’s generating SQL Queries. SQL isn’t exactly the best medium for serialization of data, and trying to generate optimal sequences isn’t always successful.
What happens when we take the abstraction of ORMs, and match it with databases exposing lower-level interfaces to send queries?
With the NDB API in MySQL Cluster, you get storage engine level calls like “fetch tuple” or “scan index”.
In Drizzle we’re talking about the possibility to pushing parsing to the client and sending serialized parse trees or maybe even query plans to the server directly.null
