@paulocuneo only use one of those (not both), and you have to use a protocol with the deftype as all the mutable fields are private

and as the docs say, only use :unsynchronized-mutable if you really, really, know what you're doing.

@tbaldridge Thanks! I see now, the reflection error was because of private access field. The field is intended to become read only after i set it so guess is a fair use of mutable.

is there a builtin for #(clojure.string/join "\n" %) ?

@qqq so instead of (join "\n" coll) you want something like (make-lines coll)?

does core java provide anything for doing "ttf font" -> "list of bezier curves" ?

I'm trying to redefine the logging function a 3rd party library uses during initialisation (for fun mostly). My understanding is that something like

(with-redefs [(var-get (ns-resolve 'third.party.library.core 'log/info)) (fn [& args] (println "foo"))] (library/start))

(defn add-five [a] (+ 5 a))

(with-redefs [add-five (fn [a] (+ 10))] (add-five 5)) ;; 15

(with-redefs [(var-get (ns-resolve *ns* 'add-five)) (fn [a] (+ 10))] (add-five 5)) ;; Error

add-five ;; #object[....]
(var-get (ns-resolve *ns* 'add-five)) ;; #object[..]

I suspect this might have something to do with the argument list being a vector

@kauko with-redefs already wraps the binding in a var, so use with-redefs-fn instead

(with-redefs-fn {(ns-resolve *ns* 'add-five) (fn [a] (+ 10)} #(add-five 5)) ;; 15

Thanks @rauh 🙂 I need to read a bit more about symbols and vars

@kauko The problem is you can't go from value -> var.

But instead of getting the value from a symbol, you can get the var (with var).

Yeah I think I get it

(with-redefs-fn 
{(ns-resolve 'library 'log/info) 
#(println "foobar")} 

(fn [] (library/start)))

I can run this in the REPL with no errors, but it doesn't seem like the log/info function is getting redefined

The library requires clojure.tools.logging :as log

guys, a schemaless dbms can help the development process to decresing the time ?

@kauko What are you actually trying to do? log/info is a macro...

ahh right, that complicates things

I was trying to see if I could make the library use a different method of logging

mostly for fun, as I mentioned

@kauko Java logging is extremely configurable. But it's a lot of stuff to just explain in chat

So basically ct.logging will use some backends (that can be configured with some xml files, java style) and they'll receive the clj-namespaces as an argument

You can configure which namspace can log which level (warn, info etc) and I believe even redirect just that ns to some other file

@rauh @kauko You could as well use a properties file, if xml scares you

Yeah, actually I already have an .xml file for configuring the logging. 🙂

I was just trying to see if I could use with-derefs (or something similar) in this case

for learning purposes

There's a lib called unilog that makes all this stuff more clojure friendly

How does Clojure STM deal with Reentrancy?

@abhiroop inner transactions become part of the outer transaction

So for every yield it will spawn a new transaction?

And treat it as nested?

is there anything like python setup.py develop in Clojure? To be more clear, some way to try out a lein plugin from my local checked out version?

@abhiroop not sure what you mean by yield

@andrea.crotti you can do lein install and that will put the project in your local .m2 folder

ah cool that should it do it thanks @tbaldridge

@tbaldridge I guess I was not very clear. So what I mean by reentrancy is when a thread actually yields its control flow without the scheduler preempting. And reenter the original control flow. So Thread1 --yields--> Thread2 --reenter--> Thread1. I am not sure if STM supports this kind of interaction, which is common in normal lock based concurrency. Thats why the question.

I am not sure if yield is the correct word here. I might use interrupt instead

yes, that is supported.

refs are slots where the mutable state being modified is stored. When a transaction commits, it locks all the refs it will operate on and performs the updates. Other transactions can barge the locks, and not all modifications of locks cause rollbacks of other transactions, but that's the general gist.

Other transactions can barge the locks -> Really? Could you elaborate a little bit on this?

yeah, if it comes time for one transaction to commit, but it can't because another transaction has the refs locked, there's a way for the two threads to communicate. The slow transaction will rollback and the faster one will commit.

the code is here, a quick glance over it may help https://github.com/clojure/clojure/blob/master/src/jvm/clojure/lang/LockingTransaction.java

entry point is here: https://github.com/clojure/clojure/blob/clojure-1.9.0-alpha14/src/clj/clojure/core.clj#L4980

@tbaldridge I was actually looking at this right now https://github.com/clojure/clojure/blob/f572a60262852af68cdb561784a517143a5847cf/src/jvm/clojure/lang/LockingTransaction.java#L223

Yes found the barge https://github.com/clojure/clojure/blob/f572a60262852af68cdb561784a517143a5847cf/src/jvm/clojure/lang/LockingTransaction.java#L192

So here's the thing that makes Clojure STM different from more mutable languages. The only things that are tracked during a transaction are refs (and agents, but skip that for now).

Oh I see it now you have a CountDownLatch which countsdown for the other thread till 10 seconds

And then barges in

I was looking for a bit more than this communication. For instance, (I am not sure if this is possible) but if 2 core.async channels exchange messages with each other via a channel rather than just talk about whether the other is holding a lock for too long.

I was wondering if 2 threads in the middle of a transaction can do that.

Though I think that would break the ACI guarantee of STM

Hi guys. I have just started learning clojurescript and started with a leningen template. The build was working initially , but now the figwheel repl is stuck and I see Please run lein figwheel to start the ClojureScript compiler and reload the page. message. Where should I look for errors.

Sorry I posted in the wrong channel.

Let’s say I have the following edn:

{:db/id #db/id[:db.part/user -1000007], :thing/name “thing1”}

(def tx {:db/id #db/id[:db.part/user -1000007], :thing/name “thing1”}  
(edn/read-string (prn-str tx))

RuntimeException No reader function for tag db/id  clojure.lang.EdnReader$TaggedReader.readTagged (EdnReader.java:784)

(edn/read-string {:readers *data-readers*} (prn-str tx))
;; {:db/id #db/id[:db.part/user -1000009], :thing/name "thing1"}

Hello there, folks! I'm n00b in clojure yet, just trying the water.

@ezmiller by default I don't think read-string uses the data-readers found in data-readers. However, that data reader is installed by Datomic, so it works fine if you give it the correct information

Trying to make simple http request, use clj-http (should I try another library for that?), but got strange ssl error:

1. Unhandled javax.net.ssl.SSLException
   Certificate for <api.dribbble.com> doesn't match any of the subject
   alternative names: [, ]

@netsu are you following the tutorial and using clj-http 2.0.0?

[clj-http "2.3.0"] in project's :dependencies.

Make request like that:

(http/get (followers-url given-user)
          {:headers {:access_token api-client-access-token}})

Pretty sure that I doing dumb mistake somewhere.

Try upgrading clj-http to the latest version. 3.4.1

thanks, will try.

Well, nope, didn't help(

Did you restart your repl?

sure

@tbaldridge not sure what the data-readers are… is there a doc you could point me to?

@jmsul trace:

AbstractVerifier.java:  165  org.apache.http.conn.ssl.AbstractVerifier/verify
StrictHostnameVerifier.java:   67  org.apache.http.conn.ssl.StrictHostnameVerifier/verify
     AbstractVerifier.java:  141  org.apache.http.conn.ssl.AbstractVerifier/verify
     AbstractVerifier.java:  114  org.apache.http.conn.ssl.AbstractVerifier/verify
     SSLSocketFactory.java:  580  org.apache.http.conn.ssl.SSLSocketFactory/verifyHostname
     SSLSocketFactory.java:  554  org.apache.http.conn.ssl.SSLSocketFactory/connectSocket
     SSLSocketFactory.java:  412  org.apache.http.conn.ssl.SSLSocketFactory/connectSocket
DefaultClientConnectionOperator.java:  179  org.apache.http.impl.conn.DefaultClientConnectionOperator/openConnection
ManagedClientConnectionImpl.java:  328  org.apache.http.impl.conn.ManagedClientConnectionImpl/open
DefaultRequestDirector.java:  612  org.apache.http.impl.client.DefaultRequestDirector/tryConnect
DefaultRequestDirector.java:  447  org.apache.http.impl.client.DefaultRequestDirector/execute
   AbstractHttpClient.java:  884  org.apache.http.impl.client.AbstractHttpClient/doExecute
  CloseableHttpClient.java:   82  org.apache.http.impl.client.CloseableHttpClient/execute
  CloseableHttpClient.java:  107  org.apache.http.impl.client.CloseableHttpClient/execute
                  core.clj:  304  clj-http.core/request

may it be apache problem? (Not sure why apache may required).

@ezmiller77 https://clojure.org/reference/reader#_tagged_literals

Can you make a request to https://api.dribbble.com?

oh, nope, still same error

request to works just fine 🙂

I get an OK from https://api.dribble.com using clj-http 3.3.0

can't believe that dribbble use invalid certificate. Maybe.. Is there a way to force use that one certificate as correct?

let me try 3.3.0..

> SunCertPathBuilderException unable to find valid certification path to requested target sun.security.provider.certpath.SunCertPathBuilder.build (SunCertPathBuilder.java:141)

I had this same problem with Let's Encrypt certificates with clj-http 2.0.0; updating fixed it. I would get a Java SSL error but not for other cert authorities

1. Unhandled sun.security.provider.certpath.SunCertPathBuilderException
   unable to find valid certification path to requested target

@jmsul thank you so much, upgrading to new version and lein install saved my day.

If I have this,

weights 
=> #{{:name betsy, :weight [12 32]} {:name jake, :weight [1 2 3]} {:name shyq, :weight [342 4 6]}}

(get by-weight {:weight #(some #{12} %)})

@tmtwd you mean the vector of the weight contains number or the key :weight is keyed to a number?

yes I think so

so for example, if I input 12, I would get {:name betsy, :weight [12 32]} or something like that

(map :name (filter #(number? (first (:weight %)))
                   #{{:name 'betsy, :weight [12 32]} {:name 'jake, :weight [1 2 3]} {:name 'shyq, :weight [342 4 6]}}))

oh ok

(map :name (filter #(some #{6} (:weight %)) #{{:name 'betsy, :weight [12 32]} {:name 'jake, :weight [1 2 3]} {:name 'shyq, :weight [342 4 6]}}))

this works

Hi! I’m trying to make my data structure compatible with map/filter fns, etc. Was trying to extend ISeq but it seemed to still pickup default things for structure. Here’s my latest code https://gist.github.com/sudodoki/7141ce03f934a47fe13ff41df51387f2 and I get java.lang.ClassFormatError: Duplicate method name&signature in class file. Can someone point out how I make this work?

defrecord already extends seqable

if you want custom seqable behaviour you'll have to use deftype instead of defrecord

also looks like you end there with (concat 1 nil nil)

@hlolli it would be (concat nil (list 5) nil) -> (5), so this is expected behavior

@bronsa thanks. No way to have prefer-method or similar mechanics for this?

that's right @sudodoki very recursive stuff there.

@sudodoki no, there's no way to override default methods in defrecord

Guys, do transducer achieve the same effect as Elixir's stream module? For example, if I had a range from 0 to a million and I wanted to apply maps and filters to it but without having to pass the result of one map to the next map or filter and so on until the pipeline is done.

I think Elixir calls it lazy enum operations.

This is correct. Effectively, transducers will look like a (map (f (f' (f'' x ))) xs), for example

or whatever

point being, it is very efficient and doesn't create intermediate sequences, so doesn't unnecessarily copy

After switching from defrecord to deftype, to avoid Don't know how to create ISeq from: TreeNode I updated code https://gist.github.com/sudodoki/7141ce03f934a47fe13ff41df51387f2 but now I’m getting IllegalArgumentException interface clojure.lang.ISeq is not a protocol clojure.core/extend (core_deftype.clj:769) anything I’m still missing? 🙇

ISeq is an interface, not a protocol

you can't extend it

you have to inline the def on deftype

@bronsa sorry, just got to these parts, and can’t find a good example of deftype with inline method definition w/o protocol, can you give me a hint?

(deftype TreeNode [val l r]
    (first [this] (leftmost-in this))
    (next [this] (drop 1 (dfs this)))
    (more [this] (drop 1 (dfs this)))
    (cons [this v] (xconj this v)))
; => ClassCastException clojure.lang.PersistentList cannot be cast to clojure.lang.Symbol

you're missing the interface name

deftype has the same signature as defrecord

duh! thanks

for some reason thought those were protocols as well from docs

nah, nevermind, I guess I’m not yet ready for doing this yet. ¯\(ツ)/¯

I'm super curious. go/parking threads, is there an equivalent when using libraries like Manifold?

@dominicm there is the concept of deferred that should kind of being similar iirc..

well, it is more like an abstraction over the specific concurrency primitive you are using

@richiardiandrea it's the parking thread mechanic I'm mostly curious about (or why it doesn't apply)

How can I annotate in Clojure unused function? Is (def ^{:suppress-warnings "unused"} fn would be correct analogue for @SuppressWarnings("unused") fn() {?

What tooling is giving you a warning about it?

There’s no standard for Clojure tooling, as far as I know. For example, Eastwood will give various warnings for unused function arguments and unused locals etc — but you can’t suppress most of those.

And simple Clojure metadata does not correspond to Java annotations so if you’re talking about Java tooling, it’s more complicated than that.

This blog post talks about how to add Java-level annotations to functions in Clojure: http://seancorfield.github.io/blog/2013/05/01/instrumenting-clojure-for-new-relic-monitoring/ — bear in mind that Java annotations are actually types (classes) and the annotations need to happen at the Java level — hence that post uses definterface and deftype to expose Java-level methods that can be annotated with Java-level classes.

heh, with pipelining Rapidoid can do ~3M req/s serving JSON API (sort of hello world no other IO) looks promising https://gist.github.com/l1x/2b12971e94d6160ac70f551ab8ee40ff#file-wrk_rapidoid-10g-36core-pipeline3-out

um...wrong chat window?

well, i was looking for a Java HTTP framework that I could use in Clojure that has decent performance and easy to use

so what are you doing that needs 3M req/s ? For more realistic performance numbers there's quite a few good clojure frameworks out there

for example, this one is really fast: https://github.com/pedestal/pedestal/tree/master/samples/fast-pedestal

writing 200Byte messages to an analytical store

what do you mean by more realistic?

is this based on jetty?

So, they ran the 3 tests at the end which seem to be identical, and got 3x improvement on the last vs the first. Am I reading it right? @istvan

yeah i have increased the pipelining to 10

but the last 3 tests are identical?

on the server end yes

pipelining went 3 9 10

what's being pipelined specifically?

http requests

https://en.wikipedia.org/wiki/HTTP_pipelining

you open a connection and send N requests instead of 1

this is pretty common scenario in industrial grade load-balancers

this is how F5-s achieve several million req/s / load-balancer for example

yeah, so this is nothing "special" then it seems?

no, it is absolutely standard

just people forget about it all the time

ok, making sure i was not missing something 🙂

i have spent some of my time as systems engineer with amazon so we had to understand HTTP really well

there are a lot services with more than 1M req/s / datacenter

there are other challenges when you reach this sort of performance

i am really glad NIO brings that to the JVM finally

well nio's been around for a while 😉

true that, i wasn't familiar with NIO 🙂

most of the projects done in Java are happy with the 30K req/s that you can get with any framework you like, i did not need to look into JVM performance. i just started to discover what can we do to go beyond that sort of performance

So yeah, I guess when you said "decent performance" that made me think "normal performance".

tbaldridge :))

these are relative terms, yes

i am curious about this tool though: https://github.com/pinterest/jbender

havent had a chance to try it

are there any reasons to use bidi over compojure, other than a preference of data structures over macros?

e.g. any performance metrics or async stuff that compels usage?

@albaker I don't know about bidi, but there is with Pedestal's router. It has several advanced routing algorithms that are much faster than what you will find in other libraries. See the "routers" section here: http://pedestal.io/reference/routing-quick-reference

no idea, never used it, did you run into performance issues that you definitely need to solve? if yes look into it, if not just use what you got. i guess 🙂

@albaker bidi works in cljs too

What’s a de-facto transit ring middleware library?