BUT FRANCE HAS A PONY
I really like SQL Server. Most of the time. Okay, so most of the time I like SQL Server most of the time. Don’t get me wrong, if I had to go back through the career-time continuum and pick a RDBMS to work with, I’d probably still choose it over Oracle. Probably. And, because I don’t exclusively grow facial hair from my neck, I wouldn’t be allowed to choose PostgreSQL. They’d kick me off the mailing list.
Just kidding. You’re all handsome rogues. We could have had a nice life together, staring longingly into each other’s shoes and trying to implement parallelism.
I’d have DB2 here, but the cost of entry to the Developer Edition is rather steep. So, you know, I’m sure it’s great! But no. Though I would be really happy if Microsoft implemented ANSI Standard constructs into T-SQL half as fast as IBM does.
I have poked at Oracle and PostgreSQL a bit, and found they have some really cool stuff. Heresy, right?
Check out some of these Oracle gadgets and tell me they wouldn’t make your life a whole lot easier.
In no particular order:
Table restores! Built in! I’m very surprised we never got a feature like this. You can do it with a 3rd party tool like Dell LiteSpeed.
Adaptive Plans! Go to the link and read the second paragraph. Read it twice. Wipe the drool off your face.
UPDATE: Geoff Patterson has created a Connect item to get Adaptive Plans for SQL Server.
In-Database Row Archiving! You know all that stuff you do with partitions that Oracle already does better? Where you’re basically praying for partition elimination to not undo the two weeks of work you put in to setting up this partitioned table that developers are writing horrible MERGE upserts to? Yeah. You can just tell the engine to not pay attention to rows you don’t care about anymore when it accesses the index. Fancy that.
Bitmap Indexes! It’s kind of like a filtered index, except for all values of a highly non-selective column.
Materializing CTEs! Even though it’s undocumented, we use plenty of undocumented stuff in SQL Server to get the job done. This is really cool to me, since I’ve discussed this limitation in CTEs before. I’d love to see a way to do this in SQL with the same behavior; not having to create temp tables. It would be a nice way to get around issues with caching statistics for temp tables, and especially since MS is still fixing bugs around temp tables.
Are there more? Yeah, but this is a blog. Go grab a VirtualBox and read the documentation if you’re interested in learning more.