Monads have taken the world by storm, and are supported by do-notation (at least in Haskell). Programmers are increasingly waking up to the usefulness and ubiquity of Applicatives, but they have so far been hampered by the absence of supporting notation. In this paper we show how to re-use the very same do-notation to work for Applicatives as well, providing efficiency benefits for some types that are both Monad and Applicative, and syntactic convenience for those that are merely Applicative. The result is fully implemented in GHC, and is in use at Facebook to make it easy to write highly-parallel queries in a distributed system.
We propose an extension to list comprehensions that makes it easy to express the kind of queries one would write in SQL using ORDER BY, GROUP BY, and LIMIT. Our extension adds expressive power to comprehensions, and generalises the SQL constructs that inspired it. Moreover, it is easy to implement, using simple desugaring rules.
For example, consider this SQL querySELECT dept, SUM(salary)The GROUP BY clause groups records together; the ORDER BY sorts the departments in order of salary bill; and the LIMIT clause picks just the first five records. This support for grouping and sorting is extremely useful in practice, but is not available in list comprehensions.
GROUP BY dept
ORDER BY SUM(salary) DESCENDING
In this paper we propose an extension to list comprehensions that makes it easy to express the kind of queries one would write in SQL using ORDER BY, GROUP BY, and LIMIT. Here, for example, is how the above SQL query would be rendered in our extension.[ (the dept, sum salary)Moreover, our extensions are significantly more general than SQL's facilities.
| (name, dept, salary) <- employees
, group by dept
, order by Down (sum salary)
, order using take 5 ]