Jeff Moden (8/24/2014)
I absolutely agree that they'd better be able to pull the proverbial rabbit out of the hat on schema design to enable high performance code to be written but, with the tools they use to do such a thing and the design patterns they follow, they don't actually need to know a thing about how to write the code.I went through that about a year ago. A company wanted me to come in to help them write some code. They introduced me to the DA and he spent about an hour going over the schema, the table keys, and the relationships for a huge database. It was a work of art bordering on genius. I asked him why with such a great design that they needed me to write code. His answer was (and I literally nearly fell backwards in my chair) "I only know how to design the databases... I don't actually know how to write code".
As another example, I went through that at a previous company that I worked for. They built a whole new OSS in Oracle. Again, the design of the database was outstanding in every way I could think of. I'd been assigned a task that I was having problems writing code for (it was early in my experience with Oracle) so I went to the DA that designed it and starting drawing code on his white-board. He asked me to stop because he couldn't help. When I asked "Why not?", his comment was "This is a gibberish to me... I don't know how to do a 3 table join never mind what you've written so far". When I asked him how he designed the database, his reply was "The tools make it easy. I just need to fill in the right blanks and draw the right lines. It even builds the correct indexes for me".
That sounds as if you've met some amazing data analysts - perhaps people who are using tools based on the research by people like Bernstein and Zaniolo and Beeri and Fagin and Raiha and Melkanoff and so on which I thought had been ignored by the industry (except IBM, who published some of it, but as far as I know didn't actually use it) and indeed by most of academia. I wish I'd met some of those DAs and got my hands on some of their tools.
I still reckon a good DA can probably write relational expressions in some notation though, although it may not be SQL so some people won't recognise it as code.
Tom