got an example?

and do you mean something like (map (juxt f g) collection) so it would yield ( ((f x1) (g x1)) ((f x2) (g x2)) ...)?

sure does. i think i would just make a reduce wrapper that did the normal reduce and also built up the seq of stuff as well

(let [reduce-f +
      reduce-init 0
      f inc]
  (reduce (fn [[acc coll] x]
            [(reduce-f acc x) (conj coll (f x))])
          [reduce-init ()]
          (range 10)))

a reducer that does the normal reduction and also conj's like a mapping. you can make it more tailored or generic as you see fit

there's more elegant solutions out there. i'm playing around with it. perhaps defining map as a reduce and then just doing simultaneous reduces

and i'm doubting that the above is an actual solution

When I give an xf, that is a transducer, to a channel when constructing it, does it run when I put a value into channel (and thus is run on my thread), or when someone takes the value out of channel (and it's run on their thread)?

Also is it just me or is the clojure starting a lot faster on Java12 than Java 8

@roklenarcic seems it's applied when adding items to the channel buffer -> https://github.com/clojure/core.async/blob/a25ae2fb44f0bc2828377866443ed638beeed9c9/src/main/clojure/clojure/core/async/impl/channels.clj#L291

thanks

@roklenarcic @plexus the transducer can be run either by producer or consumer thread, depending on the channel state

performance aside, are map and pmap identical?

they might have different chunking behavior with some inputs (but you shouldn't be using laziness if chunking would effect correctness)

I'm doing some license review prior to distributing a JAR. Our intent is that the JAR is closed-source (for now) but freely available, similar to REBL. Is there a good non-lawyer summary of how compatible LPGL libs might be with this kind of approach?

(the JAR is more application-y than library-y).

@rgm I'm not a lawyer, but Metabase is a) a clojure application b) uses AGPL https://github.com/metabase/metabase/blob/master/LICENSE.txt - might be helpful

oh, interesting, thanks.

does seem like all of the app is there, though, as readable clojure (ie. open source).

how to make spec for number? not generate NaN?

you'd have to filter

found double-in, looks fitting

if you want just doubles, then yes

although I don't want to reject ints..

went with (s/and number? #(not (Double/isNaN %)))

you sure that doesn't throw with non-doubles?

I guess you'll get a conversion

(Double/isNaN 1) => false

holy cow, spec test checking found a bug in a code path I haven't thought about

I just provided a shape for input/output, and it found a bug! mind-blowing!

it works! :)

Has anyone ever had problems importing java classes marked final? My namespace import

(ns com.my.ns
  (:import (org.apache.lucene.search.suggest Input)))

package org.apache.lucene.search.suggest;

import java.util.Set;

import org.apache.lucene.util.BytesRef;

/** corresponds to {@link InputIterator}'s entries */
public final class Input {
  public final BytesRef term;
  public final long v;
  public final BytesRef payload;
  public final boolean hasPayloads;
  public final Set<BytesRef> contexts;
  public final boolean hasContexts;

  public Input(BytesRef term, long v, BytesRef payload) {
    this(term, v, payload, true, null, false);
  }
  
  public Input(String term, long v, BytesRef payload) {
    this(new BytesRef(term), v, payload);
  }
  
  public Input(BytesRef term, long v, Set<BytesRef> contexts) {
    this(term, v, null, false, contexts, true);
  }
  
  public Input(String term, long v, Set<BytesRef> contexts) {
    this(new BytesRef(term), v, null, false, contexts, true);
  }
  
  public Input(BytesRef term, long v) {
    this(term, v, null, false, null, false);
  }
  
  public Input(String term, long v) {
    this(new BytesRef(term), v, null, false, null, false);
  }
  
  public Input(String term, int v, BytesRef payload, Set<BytesRef> contexts) {
    this(new BytesRef(term), v, payload, true, contexts, true);
  }

  public Input(BytesRef term, long v, BytesRef payload, Set<BytesRef> contexts) {
    this(term, v, payload, true, contexts, true);
  }
  

  
  public Input(BytesRef term, long v, BytesRef payload, boolean hasPayloads, Set<BytesRef> contexts, 
      boolean hasContexts) {
    this.term = term;
    this.v = v;
    this.payload = payload;
    this.hasPayloads = hasPayloads;
    this.contexts = contexts;
    this.hasContexts = hasContexts;
  }
  
  public boolean hasContexts() {
    return hasContexts;
  }
  
  public boolean hasPayloads() {
    return hasPayloads;
  }
}

(ns com.my.ns
  (:import (org.apache.lucene.search.suggest Input)))
Execution error (ClassNotFoundException) at java.net.URLClassLoader/findClass (URLClassLoader.java:382).
org.apache.lucene.search.suggest.Input

(ns my.com.ns
  (:import (org.apache.lucene.search.suggest.analyzing AnalyzingInfixSuggester)
           (org.apache.lucene.store NIOFSDirectory)
           (org.apache.lucene.search.suggest InputIterator)
           (org.apache.lucene.util BytesRef))

you can check if the actual class file exists on the classpath with io/resource

ser=> ( "java/util/Set.class")
#object[java.net.URL 0x1fdfafd2 "jar:file:/Library/Java/JavaVirtualMachines/jdk1.8.0_171.jdk/Contents/Home/jre/lib/rt.jar!/java/util/Set.class"]

try that with "org/apache/lucene/search/suggest/Input.class"

you should check the version of the library you are using matches the version of the source you are looking at

for example these docs https://lucene.apache.org/core/4_10_2/suggest/org/apache/lucene/search/suggest/package-summary.html show InputIterator, but no Input

I'm using the latest lucene which, in the same package, it has the class InputIterator

(io/resource "org/apache/lucene/search/suggest/InputIterator.class")
#object[java.net.URL 0x7c4dd006 "jar:file:/Users/magemasher/.m2/repository/org/apache/lucene/lucene-suggest/8.2.0/lucene-suggest-8.2.0.jar!/org/apache/lucene/search/suggest/InputIterator.class"]

(io/resource "org/apache/lucene/search/suggest/Input.class")
nil

(classpath checking was a great idea!)

My deps edn contains

org.apache.lucene/lucene-core               {:mvn/version "8.2.0"}
  org.apache.lucene/lucene-suggest            {:mvn/version "8.2.0"}
  org.apache.lucene/lucene-analyzers-common   {:mvn/version "8.2.0"}
  org.apache.lucene/lucene-queries            {:mvn/version "8.2.0"}

Which is why i'm so confused that the Input class would be missing but the InputIterator from the same package is present.

Figured it out, it's not on the class path or even in the jar because Input is a test class. Sorry for the noise.

(Looking at the lucene source code)

I don't know how to introspect version conflicts in deps.edn, I know you can sort this out with lein using lein deps :why ...

glad you sorted it out

Thanks for the help, I was losing my mind.