This presentation about "modern SQL" is interesting:
It reminded me how much I hated using SQL back when I did, and made me feel slightly optimistic that they are finally solving some of its most obnoxious failings, and yet made me super sad that we still don't have anything better than this.
I'm not a big functional language person, but honestly, a real functional query language would be so much better than this bunk. It's embarrassing that any time I want to do a slightly different type of query they have to standardize a new language feature. FETCH NEXT 10 ROWS ONLY indeed.
And then we think we're all special with the mapreduces, but we just went too far the other way. A lot of big data analysis forgoes indexes completely. It's so easy to spin up CPU cores now, but so hard to define an index, and so slow to do random disk seeks, that we often find ourselves grunging linearly through terabytes of data only to throw 99.9% of it away.
There must be a better way.