except those pre/post conditions often come with a performance hit

@bravilogy You need (-> (client/get uri {:headers {"apisign" sig} :as :json}) :body :result) -- You are trying to get the :result element out of the :body of the client/get response.

The conversion to text happens when Ring is rendering the result.

Random brain firing - is there a way to capture the current function's actual object instance? There are some recursion patterns which would be a lot easier to implement if you could just recursively pass the current function with its closure and all.

Or is the "right way" to do this some partial/y-combinator trickery

@arrdem I'm probably misunderstanding, but does this help?

+user=> (defn foo [] foo)
#'user/foo
+user=> (= (foo) foo)
true

does it need to be a way to automatically capture the current fn where you can't know the name (or even work in a case where a name is impossible like comp or partial) ?

oh iiiiinteresting

user=> (def a (fn [x] (fn foo [] (println x) foo)))
#'user/a
user=> ((a 1))
1
#object[user$a$foo__135 0x542e560f "[email protected]"]
user=> (((a 1)))
1
1
#object[user$a$foo__135 0x1d730606 "[email protected]"]
user=> 

named lambdas can refer to themselves by instance using their name.

yup!

guess I didn't need to patch the compiler to add this as an implicit local 😛

oh no I have become a tool of evil

Oh no this is great

totally happy to not have to run some goofy patchset to get what I want

(it was a nice patchset tho)

Hi, can anyone suggest me some small learning project which would give me some idea about handling concurrency in clojure. Something which can be finished in a week.

I'm writing a macro in a *.cljc file. Is there a way to make it expand to different things depending on whether it is being expanded in CLJ or CLJS code ?

@qqq You may find this helpful: http://blog.fikesfarm.com/posts/2015-06-19-portable-macro-musing.html

Long story short is its a bit of a pain

actually, how is it painful?

(defmacro str->int [s]
  #?(:clj  (Integer/parseInt s)
     :cljs (js/parseInt s)))

@qqq I think you example macro is not right, it’s evaluating the expression at compile-time.

You should return the code to do the work instead.

e.g. this fails:

(let [s "1"] (str->int s))

also, you need to know what you are emitting from the macro: clj or cljs. There are hacks to get this info. And a awesome lib hiding the details: https://github.com/cgrand/macrovich

(defn select-rename [m ks]
  (into {}
        (for [[old-k new-k] ks]
          [new-k (get m old-k)])))

(select-rename
 {:a 20
  :b 30
  :c 40} [[:a :foo] [:c :bar]])

@qqq There're two functions - rename-keys and select-keys.

there's clojure.set/rename-keys and select-keys

Also note that you can implement this without an intermedia seq by using a transducer: (into {} (map (fn [[old new]] ...)) ks)

@the-kenny: I agree that unnecessary intermedaite seqs are bad. Can you clarify where my code is generating it ?

(= 2 (+ 1 1))
(comment
 true)

(= (int-array [2])
   (int-array [(+ 1 1)]))
(comment
 false)

What is the best way to run all unit tests (assuming that they are in different namespaces) in a project, without requiring them first, from the repl?

@stuartsierra any idea ? ☝️

@ido why don't you want to require them?

You can't run a test, without require the namespace that test is in.

@dominicm say I have 10 test namespaces.

@ido tools.namespace will require all namespaces it can find.

This should solve your problem if you perform a refresh before running your tests.

Thanks @dominicm!

Count result of a transduction, which one is preferable?

;; 1
(count
   (sequence (map identity)
             [1 2 3]))
  ;; vs
;; 2
  (transduce (map (constantly 1))
             +
             [1 2 3])

@borkdude the first one makes more sense to me. I think you could use eduction in place of sequence if this is the exact usage, as it will be faster.

I get count is not supported on eduction

Really? I figured count worked on any reducible.

@borkdude What are you optimizing for here? Performance or readability?

@dominicm Both 🙂

In that case, use sequence if it's not fast enough, benchmark the transduce (but I wouldn't expect that to be faster, it's potentially slower!)

thanks

last but not least, use eduction with a fold to parallelise the count, if your resulting collection will be large enough. Also benchmark this approach.

@dominicm This is just #adventofcode related, so I guess readability is slightly more important. I’ll go with sequence.

sorry for the crosspost - has anyone seen this before? https://clojurians.slack.com/archives/C075TNSSC/p1512387863000190

@tristram Try https://confluence.atlassian.com/jira/connecting-to-ssl-services-117455.html (see Resolution section)

it tells how to get certificate for your url and to import into Java cert store to trust it

thanks, but I'm just trying to use boot or lein

which seems like it should work? it's not my servers

@tristram those instructions are to be done at client computer

i. e. at your work machine

it's not required for it to be a server

to successfully run Portecle app, to get http://github.com certificate by it and to import that certificate to local cert store

well, I fixed it by uninstalling and reinstalling openjdk-8-jdkconstantly throughout the day

¯\(ツ)/¯

FWIW, the last line I ran, from having no jdk installed, was

sudo apt install openjdk-8-jdk

@dominicm using net.cgrand.xforms: (x/count xform input)

@borkdude Yep, xforms is always the correct answer in these cases.

I'l have to remember that xforms has a count

(thanks @thegeez)

Any rumors about 1.9's stability and its final release? 🙂

1.9.0-RC2 is pretty much what 1.9.0 should be in a short while

Wow

I expect 1.9.0-RC2 is exactly what 1.9.0 will be :)

(let [a (if editing? edit-a (-> data :foo :bar :a))
      b (if editing? edit-b (-> data :quux :b format-b-for-display))
      c (if editing? edit-c (-> data :otherkey :c))]
  [1 2 a b c 3 4])

(Without repeating the 1 2 3 4.)

I want to keep the readable [1 2 a b c 3 4] at the end.

@matan Re: stability -- we've had all the 1.9 Alpha and Beta builds in production for ages. We have RC1 in production right now with RC2 going to production at 2pm Pacific today.

whats the main difference between float? and double?

i want to validate input from a string and check if its content is a valid double

@seancorfield sounds good. but did you have a specific agenda in chasing the bleeding edge like that? I mean 1.8 was good and stable. care to clarify the motivations for this chase? I just can't help but wonder as most people won't go that far (or fast :) I guess..

@matan there's really not a reason not to. Clojure alpha/beta versions very rarely break existing features. Most of the time I upgrade, run the test suite, and if those tests pass we're good.

Last time I remember Clojure "breaking" something critical was when they changed the hashing algorithm back in 1.6. But even that wasn't Clojure's fault since order of items in sets is undefined (hence why they are unordered sets). So that was a bit painful to fix, but it was considered a code cleanup since the problem was in the code that used Clojure.

But I can't even think of a time I've had my correct code break from a Clojure upgrade.

weren't there some issues with 1.2 -> 1.3

i wasn't around back then so i don't know but i thought i remember hearing about that

well yeah, I'm not going that far back. and I joined Clojure right when 1.3 was released 😉

That was back in 2011, lol

well there's a good answer then. there haven't been breaking changes in an upgrade for 6 years or so

There are times where there were perhaps legitimate questions whether a change in a release was breaking or not, such as the hash change in 1.6 also leading to hash inconsistency between Java sets and Clojure sets that were clojure.core/= before and after, but no longer had the same hash any more.

Which then causes sets or maps containing those things not to be clojure.core/= any more.

Such questions usually lead to refinements in the definition of what was promised, and what wasn't.

adding validation to the ns macro which subsequently stopped (ns foo (use bar)) being acceptable is another one of those things

no one said it would work, but it did, and now it doesn't. given the low number of people affected and how easy it was to fix I'm glad they went with locking it down though

Hmmm,

> (println '(and "foo"))
(and foo)

prn

Indeed, thanks!

ya, println tries to be human friendly. prn is for serializing data

ISTR 1.4 "broke" some stuff around numeric overflow / representation? (didn't affect us) and 1.5.0 had a memory leak (fixed in 1.5.1 -- we were spared that as we were on a pre-release of 1.5 in production at the time, before the leak was introduced, and we went straight to 1.5.1 because of our release schedules at the time). And the hash code change in 1.6 didn't bite us either (I'm still not clear on how widespread that breakage was?).

We started with Clojure 1.3 (Alpha 7 or 8 in production) because we wanted to skip ahead of the big 1.2 -> 1.3 Contrib split and the AOT compatibility issues around that. It's also how I ended up maintaining clojure.java.jdbc -- because clojure.contrib.sql needed a maintainer to bring it across the 1.2/1.3 divide 🙂

We test our apps against Clojure's master SNAPSHOT as well as whatever version we're going to production with so we get a heads up of any breakage.

@tbaldridge @seancorfield thanks for the perspectives and explanation ― I sure learned something new

As someone who does not maintain a clojure core library, I might be a bit more conservative myself 🙂 but testing against Clojure's master snapshot sure sounds nice (especially if it doesn't break the test suite very often)

One of Clojure's strengths, in my opinion, is how stable it has been across versions. The Clojure/core team are very serious about accretion rather than breakage 🙂

Something else ― roughly one week ago I asked about writing code for stream-processing of data, which gave rise to a short discussion suggesting that although collections are sufficiently lazy (only up to 32 items may be realized under the covers at any point in time), transducers should be strictly preferred for a streaming application. In streaming I mean, mainly, processing an input file using only a constant memory footprint, rather than consuming the entire (potentially humongous) file into memory and such. Then also processing the stream forward through whatever few processing steps, in a similar way, for the same reason. Do folks here agree to that suggestion, about preferring transducers over plain lazy collections?

lazy collections and transducers aren’t mutually exclusive

right

transducers are a good option for describing and composing the individual steps

they also are useful for streaming because they don’t care about the source or the sink

in general, for streaming processing, you are going to be working with folds (a tree shaped process, the degenerate shape of which is a linear reduce)

as far as processing data with constant memory requirements, you can use lazy sequences, but there are other good options as well

once you are working folds, using transducers and transduce is pretty much a no brainer

there are also a lot of stream processing libraries to choose from that help provide a good story for how to put everything together

when you are using a fold like reduce (or transduce) you also get a pretty clear scope for managing resources (like closing db connections, or files, or whatever) which lazy-seqs don't really give you

(interleave @smith.adriane @hiredman) 🙂

XD

@matan I prefer transducers in this case simply beacause it's really hard not to hold onto the head of a seq with lazy seqs. And with transducers its pretty much impossible to shoot yourself in the foot this way.

I've built too many systems that work fine until you get about 2GB of data through them then they fall over due to some sort of lazy-seq related memory "leak"

Hello Everyone. Are there any good resources to learn Clojure quickly and efficiently?

If you don't mind paying some $, Clojure Programming is a good book: http://shop.oreilly.com/product/0636920013754.do There are other fine books to learn from, also.

@faraj.9798 I will also add Clojure for the Brave and true https://www.braveclojure.com/ If you are looking for raw time to workflow, then brave clojure has the advantage there IMHO. If you want a more in depth resource, then the O'Reilly book is always a really safe bet. It just depends on where your priorities are.

Or Living Clojure by Carin Meier, or Programming Clojure 3rd Edition... 🙂

Clojure Programming is good for the basics but a little outdated now (it targets Clojure 1.5 I think?).

Nope, 1.4 (examples work on 1.3 and 1.4). So I'd be cautious to recommend Clojure Programming at this point (although I love it as an all-round introduction to the language and some common libraries -- it's clojure.java.jdbc section is outdated, due to lots of improvements in that library... I'm not sure it's example even work with recent clojure.java.jdbc versions).

@faraj.9798 If you're new to Clojure, you'll find the folks in the #beginners channel are super helpful.

I'm interviewing at a company that uses Clojure and React. I'm comfortable with React, just need to increase skills with Clojure.

Thanks everyone for your help.