Clojurians Log v2
clojure-dev
I’m working on testing concat and lazy-cat for #clojure-test-suite. After reading through the doc strings for both, I’m struggling to understand in which situations I’d use lazy-cat instead of concat. The doc string for lazy-cat says: (lazy-cat xs ys zs) === (concat (lazy-seq xs) (lazy-seq ys) (lazy-seq zs)). Given that concat returns a lazy-seq , as do most of the sequence functions that are likely to form inputs to lazy-cat, under what circumstances is lazy-cat providing me more desirable laziness than I already have with concat? And that’s doubly a question with chunking being part of the equation, since any underlying chunking will undo some of the laziness. Is this mostly a historical artifact of Clojure’s development, or are there real situations today (Clojure 1.12) where lazy-cat is particularly required?
there’s a usage in the tests for for that is interesting:
(defn only
"Returns a lazy seq of increasing ints starting at 0. Trying to get
the nth+1 value of the seq throws an exception. This is meant to
help detecting over-eagerness in lazy seq consumers."
[n]
(lazy-cat (range n)
(throw (Exception. "consumer went too far in lazy seq"))))Thanks. I suspect that test was written a while ago. range is chunked, now, so wrapping it in a lazy-seq doesn’t do much other than put off any initial chunking, which I suspect is already deferred (I’d have to go through the chunking implementation). You’d get the same thing, I think, with (concat (range n) (lazy-seq (throw (Exception. …)))). Based on this, it doesn’t seem like lazy-cat is adding much to the core API. That’s why I’m suspecting that it’s a historical artifact that might not provide much value in modern Clojure.
I’ve always thought of it for “building a lazy seq out of non-lazy collections with different creation costs” as Renzo noted https://clojuredocs.org/clojure.core/lazy-cat#example-5aa01b54e4b0316c0f44f910.