Clojurians Log v2

OK I’ve fixed the spec issue, the Var issue, and the record issue.

Everything is currently on epsilon I’ll create a release tomorrow.

https://clojars.org/meander/epsilon

@rfhayashi @borkdude Lemme know if this works with babashka @mark340 The record problem has been fixed.

@noprompt I'll take a look later today, thanks for reminding me.

@noprompt Is it necessary that the FAIL sentinel is seqable? https://github.com/noprompt/meander/blob/f4b1a90c75a8428e77f59feb0e152216e0460a82/src/meander/match/runtime/epsilon.cljc#L20

I’ll double check. At one point it was, however, that may no longer be case.

Ok, if not, could make it just an Object

(identical? (js/Object.) (js/Object.)) ;;=> false

Don't worry about releasing, I can test with git dep

Branch also works for me

Cool. I tend to prefer (reify) in the place of Object and js/Object.

Actually, sorry, what’s the issue? 🙂

(reify ISeqable ...)

reify does work for some cases, but I have to make explicit support for it on a class by class basis :/

Since it's .cljc you could also consider a :bb branch with (Object.)

@noprompt Hmm, did you mean (reify) as in without args? That works:

$ bb -e '(identical? (reify) (reify))'
false

TIL

Hehe. So what I should have said is “if removing the seqable stuff works then just leave it as (def FAIL (reify)).”

yes, that would work for CLJ, CLJS and bb

I have this function in plain Clojure. Out of curiosity, what would be the meander equivalent? I feel it should be straight forward, but I have not found a concise solution using meander.

(defn replace-value [data]
  (clojure.walk/prewalk
   #(if (:a %)
      (assoc % :a (:b %))
      %) data))

You could use the strategy namespace combining top-down and rewrite.

(m*/top-down
  (m*/rewrite
   {:a (m/or false nil), :b ?b & ?m}
   {:a ?b & ?m}))

The meander version will likely be slower.

The seqable stuff is not removable it appears, however, I’m happy to support a :bb clause.

This works as well:

user=> (identical? (String. "FAIL") (String. "FAIL"))
false

cljs.user=> (identical? (js/String. "FAIL") (js/String. "FAIL"))
false

unless you need the seqable to return nil on seq

If you need it to be quicker, you’ll need to ditch both prewalk and avoid top-down . These traversals indiscriminately hit everything and 99% of that’s a waste.

I have an idea…

oh wait:

(seq "")
;;=> nil

cljs.user=> (identical? (js/String. "") (js/String. ""))
false

user=> (identical? (list) (list))
true
user=> (identical? (with-meta (list) {}) (list))
false
user=> (identical? (with-meta (list) {}) (with-meta (list) {}))
false

😛

That works too :) You can even put some nice info in that metadata ;)

Like {:lol-is-this-a-hack? 'yes}

😂

:lol.impl.dont-look-here true

Random aside: I wonder if anyone has ever Xibit style put meta on their meta.

(with-meta x (with-meta ,,,))

I think there are metadata values with metadata. E.g. when you datafy something, the original value is stored as metadata

Meta linked list. 🙂

So, yah, the list bit does the trick.

Tests are passing.

OK I’ve pushed it up. Wanna try it?

Aren’t semantics fun?! 😄

I'm now running into another Seqable reference:

user=> (require '[meander.epsilon] :reload)
Could not resolve symbol: clojure.lang.Seqable [at /Users/borkdude/.gitlibs/libs/meander/epsilon/7ed07ce8766aa5fbf7d457bdeffe39e2e1323f73/src/meander/match/ir/epsilon.cljc:851:31]

The code gets rid of predicates - why?

perf?

Yes.

Shaving those checks off can save a lot of time.

This isn't exactly the same though:

#'clojure.core/seqable? clojure.lang.Seqable

I support clojure.lang.IPersistentSet already in bb. I'll add support for checking Seqable as well

The set? one looks like an oversight

Yeah, you already have SetInterface

We should probably tag @jimmy on this bit too. 🙂

Will take a look at fixing those discrepancies

@jimmy No pressure. I’m hacking away at them as @borkdude brings stuff up. I just wanted to tag you on the thread.

The next thing I'm running into:

user=> (require '[meander.epsilon] :reload)
Could not resolve symbol: clojure.lang.RT/iter [at /Users/borkdude/.gitlibs/libs/meander/epsilon/7ed07ce8766aa5fbf7d457bdeffe39e2e1323f73/src/meander/substitute/runtime/epsilon.cljc:11:12]

I don’t think there is a clean mapping to core clojure.

Basically the goal is to make a java.lang.Iterator and this was the easiest way to do that without porting the RT code.

isn't this iterator-seq maybe?

No that’s a different thing.

It might be safe to just call .iterator there.

Eh, no, cause there’s special sauce for strings, etc.

I can port it

I think that would be useful regardless of babashka, as to not rely on Clojure internals too much

After your port I will check if it works and make necessary changes to bb if needed

Done

Basically wrote

(if (instance? java.lang.Iterable coll)
  (.iterator ^java.lang.Iterable coll)
  (.iterator (seq coll)))

Feedback:

(set! *warn-on-reflection* true)

(defn iter [coll]
  (if (instance? java.lang.Iterable coll)
    (.iterator ^java.lang.Iterable coll)
    (.iterator ^java.lang.Iterable (seq coll))))

(prn (iter []))
(prn (iter "x")) ;; needs type hint on seq
(prn (iter nil)) ;; needs fixing

@noprompt Thanks.

Maybe return (iter []) for (iter nil) ?

Oh duh

You still need a type hint here: https://github.com/noprompt/meander/commit/6f7b8265e5102ae98eb304719b884a73aef53fe3#diff-cdb266462b83e64b48c052f2f26fca6ea027e9663cce3ff64b08fc81b840e69bR16

Got it

I’m still trying to figure out the Seqable thing.

Installed bb… looks like it doesn’t like java.lang.Iterable either?

That's easy to fix. Let me get back to you after I finish another (unrelated) refactor

https://github.com/noprompt/meander/pull/154

The test script, however, doesn’t account for babashka yet

You can bang on that branch if you want. I’ve gotta take my hands off of it for a bit.

This is a better way to write it for bb:

(defn iter [coll]
  (if (instance? java.lang.Iterable coll)
    (.iterator ^java.lang.Iterable coll)
    (.iterator ^java.lang.Iterable (seq coll))))

@rfhayashi We’re (@borkdude and I) have been working back and forth on supporting bb. There’s an issue and PR in the works.

else I'm afraid it won't even work :)

I'm almost ready here, let me then push Iterable

I just added a :bb clause without type hints. Feel free to adjust it.

https://github.com/noprompt/meander/pull/154/files#diff-cdb266462b83e64b48c052f2f26fca6ea027e9663cce3ff64b08fc81b840e69bR11

Thank you both so much for that 🙂 I'll try to give it a go tomorrow.

I now pushed a commit to bb master where this should work:

(defn iter [coll]
  (if (instance? java.lang.Iterable coll)
    (.iterator ^java.lang.Iterable coll)
    (let [s (or (seq coll) [])]
      (.iterator ^java.lang.Iterable s))))

(= [1 2 3] (iterator-seq (iter [1 2 3])))

I'll download your branch

iter is fixed now on babashka master. Note that you can download binaries from #babashka_circleci_builds or simply from the builds on Github, if you want to try the binary version. The next issue I'll just mention on Github

The next issue is the 2 arg version of resolve. I'll have to fix that in sci (the interpreter backing babashka).

I'll get back to this, getting late now.