3 ms·

It feels like everyone mixes relational databases with SQL in discussions. Relational databases are fine but SQL is disgusting. It's a query language that's ver

It feels like everyone mixes relational databases with SQL in discussions. Relational databases are fine but SQL is disgusting. It's a query language that's very hard to parse or generate and has multiple incompatible variants. Ironically the Lambda calculus version looks much better, I'd never seen those before. I wish we'd go back to that as default, unify around a standard and expose block text SQL as an extra for users.

jeremiep 9y ago
The one downside I see with SQL is that its a string. Would it be a data structure instead it would be incredibly easier to manipulate, without third party query builders. Even without that its still more convenient than most NoSQL. Still, Datalog as implemented by Datomic is a breath of fresh air.
- zimablue 9y ago
  It's more than just being a string though, it's the multiple incompatible versions and the stunted mini language of it's own. A good solution would be able to run arbitrary expressions in some "real" programming language. It also in my experience doesn't seem to deliver on its central premise of converting declarative expressions to fast operations. So often in a SP I've had to force a write to an intermediate table to jimmy the execution plan. I think it's like excel, it's not considered good because the thing itself is good it seems good because the thing it does is very useful (lightweight interactive dataflow programming, talking to relational databases) and it has a near monopoly. LINQ is really cool but doesn't exist in python (macropython has an unused attempt PINQ I think). Probably because a) SQL is bad b) only MS actually has the definition of the mssql language!
oldandtired 9y ago
Not quite, the confusion is that SQL DBMS's are assumed to be relational when they are not. None of the major players have produced a relational DBMS (there is one proviso, but IBM didn't promote that specific product out of the UK over its DB2 product out of the USA).

jeremiep 9y ago

The one downside I see with SQL is that its a string. Would it be a data structure instead it would be incredibly easier to manipulate, without third party query builders. Even without that its still more convenient than most NoSQL. Still, Datalog as implemented by Datomic is a breath of fresh air.

zimablue 9y ago
It's more than just being a string though, it's the multiple incompatible versions and the stunted mini language of it's own. A good solution would be able to run arbitrary expressions in some "real" programming language. It also in my experience doesn't seem to deliver on its central premise of converting declarative expressions to fast operations. So often in a SP I've had to force a write to an intermediate table to jimmy the execution plan. I think it's like excel, it's not considered good because the thing itself is good it seems good because the thing it does is very useful (lightweight interactive dataflow programming, talking to relational databases) and it has a near monopoly. LINQ is really cool but doesn't exist in python (macropython has an unused attempt PINQ I think). Probably because a) SQL is bad b) only MS actually has the definition of the mssql language!

oldandtired 9y ago

Not quite, the confusion is that SQL DBMS's are assumed to be relational when they are not. None of the major players have produced a relational DBMS (there is one proviso, but IBM didn't promote that specific product out of the UK over its DB2 product out of the USA).