Turns out the implementation of the article is in Haskell, another declarative language.
At work. we use Power Query with it's M language - a declarative language with lazy evaluation.
Is there something about declarative languages that makes them especially suitable for data work?
Part of it is there’s so many different ways to represent data, and even more ways to compute a given quantity — but the quantity itself often has a clear definition (sum this column from all rows where this holds, say)