@hlship any plans to support cljs for client part of lacinia?

hi @martinklepsch, sorry i'm a bit behind but should have time to look into the issue this weekend.

Would you consider the following to be a reasonable (i.e. readable, performant, and idiomatic) implementation of "longest common prefix" for two strings?

(defn- longest-common-prefix [a b]
  (->> (map vector a b)
       (take-while  (fn [[char-a char-b]] (= char-a char-b)))
       count))

@pyr no worries, I think it’s probably not even related to unilog but I don’t know enough about this stuff to really tell...

@jumar I think it depends on how similar the strings are in general. I could imagine that using a bi-sectioning algorithm may be faster.

yeah, that's true. Let's say we have no idea how the strings look like and that performance is slightly important, but not the most important factor. Or we can just ignore it for a while and focus on "readable" and "idiomatic" 🙂.

😛

You implementation is fine, you can optimize by replacing the anonymous fn with =:

(->> ["abcdef" "abdd"] (map vector) (take-while =) count)

@ordnungswidrig I don't think that works.

(map vector [a b]) is not the same as (map vector a b)

@rauh you’re right (->> ["abcdef" "abcefgh"] (apply map vector) (take-while (partial apply =)) count) it is. I was fooled by getting the correct result, 2.

How do you combine multiple lein profiles in combination with the "+" operator that makes them add to the default profile?

Is there a function that selects keys according to an example map, like: (f {:a {:b 1 :c 2} :d 2} {:a {:b identity}}) ;;=> {:a {:b 1}}

Maybe Specter can do this? (cc @nathanmarz )

There's select-keys, although that's only one level deep

This seems useful hmm https://github.com/alandipert/intension

I'm having trouble using java.util.ArrayList in Clojure, specifically the remove(int index) method which is supposed to remove the item at given index. It always returns false (meaning item not removed). There is also a remove(Object o) method which removes the given object from the list, and this method works as expected. Here's the code to reproduce:

(import '[java.util ArrayList])
(def foo (new ArrayList [\a \b \c \d]))
(.remove foo \a)        ; true
(.remove foo (int 1))   ; false

@ihabunek (.remove ^ArrayList foo 1)

@borkdude thanks, that works

how come (.remove foo \a) works?

As a thought, perhaps turn on *warn-on-reflection*, which may have thrown up a warning when you tried to run compile it

@lsenjov yes, that threw a warning > call to method remove can't be resolved (target class is unknown)

interesting

i don't have much experience with java interop, didn't know warn-on-reflection was a thing

Aye, the type hinting borkdude said resolved it

I didn't for a long time either

If you're using lein, you can set it in your project.clj

already on it

:global-vars {*warn-on-reflection* true}

thx

@ihabunek Without the type hint, it probably sees (.remove foo (int 1))) as a call to remove the object (int 1)

@ihabunek which it hasn’t, so it returns false

is my guess

i started with (.remove foo 1), which doesn't work, then tried using (int 1)because of what the docs say

btw, is there a more idiomatic way of doing this: i have a large list (3M+ entries) and i need to (drop-nth) repeatedly. Using vectors, this is slow, because i need to create a new vector each time i drop an element.

@ihabunek You could use a transient

🙂

@ihabunek Although I’m not 100% sure what your problem looks like. Transients are mutable counterparts which are only to be used within a function for optimizations

@borkdude I've thought about having maps implicitly work that way by translating themselves into multi-path + keypath

never had a need for it though, but easy enough to implement yourself by extending ImplicitNav to maps

@nathanmarz I’d like to use it for a light weight graphql like query to trim down a JSON response

also there's no performance cost to using ImplicitNav since it is invoked only on the first time through that code and then cached

should only be a few lines of code, if you need help jump into #specter

also it will naturally work for both querying and transformation

I’m going to look at multi-path and keypath first, thanks!

@borkdude "only within a function" isn't exactly correct. They aren't thread-safe and so you have to do thread management on your own. That's about the extent of it.

but you can use them in stuff like stateful transducers where several functions and function bodies may modify the same transient

@tbaldridge yeah, correct

Just as I thought I’d seen all mutable types already. I used volatile as mutable inside a function, but would using transients be more performant? It’s mainly used for building/comparing sets in memory, which seems exactly where transients are useful for.

If you want to know more about transients: http://hypirion.com/musings/understanding-clojure-transients

@gklijs so whenever you modify a collection, (set, vector, map, etc.) you get back a new copy of the collection. This is relatively fast, but Transients let you say: "I'm just going to throw away each result since I'm building a collection and I'm only interested in the final result". So transients allow you to look up and add new values, but that's about it, since they are mutable.

But transients also support O(1) conversion to and from persistent collections. So think of (into [] ...), this is a perfect use-case for transients as you can make the [] into a transient, add all the items, then make it persistent again.

So I wouldn't say you should use them all the time, but if you want to build a collection and do it very quickly, transients are a good place to start.

if your use case is bulk loading a collection, you should probably use transients

I have a game-state, which I transform into a deftype object, and there are several collection in the deftype object, that are update for each ‘step’ to predict the next state. All I really need from the fuction is the next best move. So none of the intermediate collection really need to be persistant. Can you also do functions like remove on transients?

Looks like there is no remove for a transient, but I could do a reduce with pop to get the same result.

Hi, I got a question about locally installing project with leiningen. I tried to look up on what the standard way of doing this is, but when I do lein install although the dependency does get placed into the default local repo (~/.m2) a new project that wants to import it can not find it. So how am I supposed to do this properly?

lein install should just work

Maybe you have a typo?

under "~/.m2/repository/"i get the directory structure projectname/projectname/0.1.0 so I tried [projectname "0.1.0"] but lein deps gives the error

what error?

are all of projectname's dependencies available either in your .m2/ or from some maven repo lein can pull from?

the error says it could not find artifact in either central or in clojars

which artifact? the one you installed in your m2? a different one?

the newly installed one

i basically did a lein install on it, swtiched to the other, added it to the project.clj and ran lein deps on it

how do you know it is the newly installed one? is it the same coordinates (group id, artifact id, version) ?

project is called foo so the error says can not find artifact foo:foo:jar:0.1.0

maybe the naming is off somehow?

aaaah well got it, glanced over the fact that the version was named 0.1.0-snapshot, whooops

well thanks for the help in any case !

was confused by the fact that the repo showed a 0.1.0 in the directory listing.

Bumping this in-case there's some lein experts around now that weren't here before: https://clojurians.slack.com/archives/C03S1KBA2/p1490351252897485

i.e. I'd like to do something like lein with-profile +foo,lib-v1:foo,lib-v2 test but I get Profiles in with-profile must either all be qualified, or none qualified

can you just include default?

lein with-profile default,foo,lib-v1:foo,lib-v2 test

Is anyone familiar with http://github.com/Gonzih/cljs-electron ?? Long story short I want to use it to build an app but I'm a confused in regards to where my Clojure (aka desktop code) would go and how it would communicate with the ClojureScript/Electron UI code?

i'm not terribly familiar with cljs-electron but very familiar with electron itself

and it uses node so i dont think you'd be using clojure anywhere really

@mattsfrey so I'm probably using the wrong tool for the job. What I'd like to do is have some clojure code that does some networking stuff and I thought it would be cool to use Electron as a nice front-end for that app. But perhaps my thinking of Electron and CLJS as a "View" for a Clojure "Controller" in a MVC type mindset is misguided. Yes?

well

you can have your backend server on the jvm and make rest calls to it from an electron frontend

if you want to launch it like an application you'd need to have electron fire up a jvm and run your server jar etc though

If I were to write the app in a more "traditional" (for me) way I would use Netty for the networking and JavaFX for the UI and it would all be one JAR. But if I wanted to do it the way you're describing the way I should approach it is put together the Aleph Clojure stuff and have it communicate to a separate running instance of Electron right?

yeah basically you'd have electron launch your jar file for the server in a process call

and then your front end could make rest calls to it

Hmm okay. So cljs-electron uses foreman I guess and it starts up electron then figwheel so if I added an additional call to lein run and wrote my app the way I would expect, then what I need to do is do "typical" server/client comm between the electron and my "server" app right?

lein is a dev tool, don’t make an actual app depend on lein being present at runtime

Hmm, yeah I think you're right. I would be better off with CLJ + some JavaFX to properly bundle it rather than figuring out how I'd properly bundle the electron+clj app. Thanks @noisesmith

@mbcev For what it's worth, you could develop your entire app on top of node written in cljs. I think that's what the cljs-electron template is for. Probably better discussed in #clojurescript

@bja I'll definitely explore more now that I have a better understanding of what Electron is. That was my bad. What I'll probably end up doing is using Clojure to build what I need for right now and then as an exploratory exercise try to re-build it later with Electron and ClojureScript. I'm far less familiar with web related technologies so I think that's why I got tripped up with this.

Building it with cljs-electron would require me to learn more about Electron and Node/NPM and any and all other associated libs/tools/etc whereas if I stay in the JVM I have less new cognitive overhead for myself.

@dragoncube r.e. cljs & Lacinia: this is not on our radar, it is something that a group with a more client-side focus can take on. Part of the point of GraphQL is a clean separation and contract between client and server. Further discussion on #graphql please.

I have a SQL table that needs to be grouped according to any combination of ~10 columns. Am I correct that these are my best options, or can you see others? * YeSQL and do the grouping in memory (in clojure) * HugSQL and I can do the grouping in the db now using snippets

noob with js background here. what is an idiomatic way in clojure to represent a function with optional static properties attached to it?

did datomic punt the difficulty of persistence / acid by using other dbs as the backend

http://basho.com/posts/technical/using-riak-as-a-storage-backend-for-datomic/

#datomic may be a better place for datomic questions

(I would say "no" to the original question)

my understanding is that what datomic requires of storage is a key value store and some kind of compare and set operation for keys. you could argue that A and D come from the storage's compare and set and durability of values for keys, but C and I are the result of the database following clojure's epochal time model.

riak is maybe the most extreme example of a non-acid datomic backend, but even dynamodb(the original backend as far as I know) doesn't provide the guarantees of acid

the postgres backend may be the only backend that has acid guarantees out of the box

I am trying to implement a Producer-consumer type system with core.async using chans and go-loops. Basically building a web-spider. Two channels - one which stores the URLs encountered till now and another which stores the content from the crawled URLs - are used. I am writing and reading from them inside independent go-loops.

But when I execute it, an instance of clojure.core.async.impl.channels.ManyToManyChannel is returned.

Hey fam. I'm a bit confused as to why I can't get any concurrency from my core.async code. I basically have a buffer of URLs that I'm feeding to n webcrawlers. Every time there's a new URL, a new crawler service is started. But I'm only getting 1 crawler at a time. Also, my print statement doesn't print. Any ideas?

(def in-chan (async/chan 2))

(defn start-async-consumers
  "Start num-consumers threads that will consume work
  from the in-chan and put the results into the out-chan."
  [num-consumers  chrome save proxyhost proxyport proxypass driver geo]
  (dotimes [n num-consumers]
    (async/thread
      (doall (print (str "New thread" n))
        (while true
          (let [obj (async/<!! in-chan)]
            (.run (core.ServiceDriver. (core.CrawlService. chrome save proxyhost proxyport proxypass driver geo) obj ""))))))))

(defn run [concurrency driver chrome save proxyhost proxyport proxypass geo addy]
  (do
    (start-async-consumers concurrency chrome save proxyhost (Integer. proxyport) proxypass driver geo)
    (while true
      (async/>!! in-chan (core.CrawlQueue/consumeAndReplace addy)))))

LOL

@pawandubey ROFL

What are the chances

I just cut like 200 lines of Java to these 15 lines. Lolz.

Clojure is fabulous. But I have not been able to grok the async and lazy bits completely yet. And they are burning me out completely on this problem.

My crawler code is here, in case anybody is feeling charitable enough to go through it: https://github.com/pawandubey/crawljer/blob/master/src/crawljer/core.clj

@bradford Your doall wraps two expressions… I think that’s why your print doesn’t.

doall‘s docstring says it takes a coll or n and a coll (but doesn’t explain the n). You’re not passing it a collection tho’...

@seancorfield Ah, thanks. Removing that does not seem to fix the concurrency issue (I wasn't expecting it to)

You also don’t need the do in your run function — function bodies can have multiple forms already.

Oh. Cool.

What does (core.CrawlQueue/consumeAndReplace addy) do? Will it produce a new value each time it is called?

Yes, it's a blocking call to a queue on AWS

@seancorfield hahaha you're the answer on this. Small world. https://www.quora.com/Why-and-when-should-I-use-the-do-expression-in-Clojure-instead-of-just-writing-multiple-expressions-sequentially

I tried your example with a somewhat cut down example (that put random integers into the channel in run and pulled them out and printed them in start-…) and ran it with five consumers and it looked like I was getting all five consumer threads running, taking turns processing the values.

If you can’t figure it out, maybe ask in #core-async where you should find folks who use core.async more heavily.

Cool, I'll poke at it some more. .run after that let is blocking, but that should be OK because it's in a thread. Hrmm.