Clojurians Log v2
clojure
(conj {:a 1 :b 2 :c 3}
{:a 4 :d 5})
;;=> {:a 4, :b 2, :c 3, :d 5}
(merge {:a 1 :b 2 :c 3}
{:a 4 :d 5})
;;=> {:a 4, :b 2, :c 3, :d 5}conj can 'merge' 2 maps. Then, what is the purpose of the merge function?
The implementation of merge is enlightening to this point. Compare for instance their behavior handling an absence of arguments or nil.
(conj {:a 1 :b 2 :c 3} nil)
;;=> {:a 1, :b 2, :c 3}
(merge {:a 1 :b 2 :c 3} nil)
;;=> {:a 1, :b 2, :c 3}
(conj nil nil)
;;=> (nil)
(merge nil nil)
;;=> nilIn conj, nil seems to be treated equivalently to an empty list (skipping the lesson that nil isn't actually an empty list). 🤯
Basically yeah. See also e.g. (conj) or (conj nil {}). Multipurpose tool versus specialized affordances.
okey, thanks!
merge is also much more clear when you have stuff like (merge a x). (conj a x) doesn't tell as much.
(conj nil nil) - this is an example nil punning in Clojure: https://ericnormand.me/article/nil-punning
Not just printing: public abstract class AMapEntry extends APersistentVector
(ins)user=> (conj {:a 0} [:b 1])
{:a 0, :b 1}
(ins)user=> (merge {:a 0} [:b 1])
{:a 0, :b 1}(it's common to represent a map entry as a vector literal, because that's also how they natively print)
(ins)user=> (first {:a 0})
[:a 0]
(ins)user=> (type *1)
clojure.lang.MapEntry
(ins)user=> (vector? (first {:a 0}))
trueHuh, I'm surprised the second doesn't result in {:a 0, 0 :b, 1 1}.
I want to perform side-effects like one would with doseq but I'd rather use eduction as it is more efficient. I've had some pearl clutching at this code:
(reduce (fn [_ [k v]]
(side-effect-fn k v))
[] (eduction coll))coll is a big collection that comes from a database and has a function pipeline applied.
This works well but, since I would like to retain my place in polite society 😅, is there a better option that retains the efficiency / single pass effect that eduction brings? Thanks in advance!Yes, I'm looking forward to my own 🤦🏻
i think run! is exactly this
You got me! 🤦🏻
and one thing i’ve seen is if coll is essentially cursor based results make sure you don’t risk a timeout on a connection held open
thanks, it's all good on that front
(but i’m a big fan of this pattern, but that’s the one thing that can get annoying when it runs for > 5 min in prod and it dies with a “This resultset is closed” error
Yeah I'm using it on a local RocksDB on a Kafka partition. We get passed an Iterable that we can safely access like this:
with-open [iter (.prefixScan store key-prefix (StringSerializer.))]thanks for the tidy up tip and I can rejoin polite society once more 🙏🏼
you never left 🙂
you might be interested in this idea too https://ask.clojure.org/index.php/13153/could-gain-another-arity-cover-common-case-feeding-eduction
run! is just a wrapper around reduce. Instead of reduce+`eduction`, you can use transduce.
In this case, run! is good for indicating side effects ... I'm explicitly not returning anything
reduce-kav would be nicer?
For those interested in the feature @alexmiller mentioned, please do upvote it in Ask Clojure. More votes = higher priority, at least for anything viable.