10 ms·
Yeah - changing date formats, double typing things you're grouping and ordering by, remembering the oddities of each dialect, typing out full join paths - it's
by thingsilearned 7y ago
Yeah - changing date formats, double typing things you're grouping and ordering by, remembering the oddities of each dialect, typing out full join paths - it's a nice experience to have that done automatically.
Right now the SQL we write is very proper - with quotes around all of the column names and table names listed before each column name. It's not what a human would write. I'd love to one day make it a little more human so that the switching into SQL mode will feel even better. It'll be a fun project.
- infinite8s 7y agoWhat do you think of Looker's idea of turning SQL into a higher level analytic DSL (which can be written/implemented by data engineers)? This DSL then makes it easy for end users to do a more complicated analytics. One issue I've had with pure SQL for analytic purposes is that a user can't build a library of higher-order functions that can be parameterized (imagine a window function to do a cumulative returns across various time periods, applied to different underlying source columns), so its difficult to build up a common business domain language without taking the typical approach of constructing SQL through glueing string fragments together typical of most programming language SQL libraries.