Clojurians Log v2

If you have main-opts in any of those aliases, then they are winding up in the wrong order

Better in that case to pass the jvm system property version of that flag using -J

-J-Dclojure.main.report=stderror

I want to intercept all incoming calls in my application, for incoming calls I’m using ring and for making outgoing calls I’m using clj-http, what is the preferred way to do it?

For ring the standard pattern is middleware. For outbound my first solution would be to make a namespace that exposes vars that do whatever else needs to be done around calls to clj-http

What is the goal? For intercepting incoming calls you could use a ring middleware. For outgoing calls you could wrap the clj-http client with your logic and use a https://github.com/dakrone/clj-http#custom-middleware there too.

Hey, I would like to start a clojure analysis tool. I need to parse clojure code and metadata. Would you recommend writing a specific parser myself, or is there an already existing generic parser that I can build onto?

How can I make Clojure print seqs as vecs (`("hi") -> ["hi"]`) so that I can easily copy-paste them into the REPL?

Hi, anyone know how to enforce required options (ie. not missing) with tools.cli?

Or must they be required as positional arguments

could be a combination of :default and :validate {:default ::not-present :validate [#(not= % ::not-present) "Required option!"]}

it might be easier to instead just ' the data structure that you copy

I've tried :validate and :validate-fn but they don't seem to be invoked for the value specified by the :default key

I guess options are really optional 😅 I'll have to do my validation without resorting to the :errors key

Is there an issue tracker for https://github.com/clojure/tools.cli?

@nicolas.estrada938 just report on http://ask.clojure.org or in #tools-deps channel. If required, Alex should open a ticket.

Is there any conceivable reason that a reductions like fn can’t have an equivalent transducer version?

they can but it has a strong dependency on the usecase for example:

(defn xf-reductions
  "Stateful transducer that mimics clojure.core/reductions.

  DIFFERENCE:
    1. There is no one-arity form.
    2. init value will not be added to the reductions list as already known."
  [f init]
  (fn [xf]
    (let [acc (volatile! init)]
      (fn
        ([] (xf))
        ([result] (xf result))
        ([result input]
         (let [res (f @acc input)]
           (vreset! acc res)
           (xf result res)))))))

@delaguardo - right. Exactly as I was thinking about it.

Every stateful transducer then can be written in these terms right? As in, aren’t all stateful transducers, reductions essentially? Or do I have this wrong?

I can’t say “every” but most of them can be written like that. And I think the only problem with reductions not having transducer equivalent is only because of it’s interface.

Right. Thanks.

Is there some EDN reading library that has similar performance profile to JSON libraries? Or should I try to use Transit for storing basic Clojure datastructures to text?

What would be a good way of getting elements 0 1 3 from each element of this list?

(def ls  '([73 62 15 3] [10 94 91 66] [44 24 4 23] [30 34 24 80] [46 58 60 56])))

(for [l ls] (vals (select-keys l [0 1 3])))

(defrecord Square [l])

(->Square 10)
(Square. 10)

Is there any meaningful difference between the two ways of constructing a record? Which one should I privilege? I know that the latter is Java interop (and we should stay away from it as much as possible) but other than that, is there a difference under the hood?

Is there a reason this trailing metadata syntax should be supported in newer clojure dialects, other than compatibility?

(defn foo
  ([])
  ([_ _])
  {:x true})

(defn foo
  {:x true}
  ([])
  ([_ _]))

clj-kondo also isn't able to deal with the first (currently) and nobody every complained about it so far, but I see it's being used in a library called omniconfig

https://github.com/grammarly/omniconf/blob/master/src/omniconf/core.clj#L272-L288 actually

https://github.com/cognitect/transit-clj/commit/b2e4bc8c95a5c14c8cae5e9edf450d61461d89dc I would go with Transit as storage.

Compatibility seems like the main concern

It’s rarely used

You should definitely prefer the constructor function over interop

They do ultimately do the same thing but the constructor will be more portable and may have features the interop does not at some future point

@alexmiller Are you referring to anything specific regarding missing features? (This is out of curiosity)

juxt is a good thing to consider for “get multiple things from the same data”. It’s certainly not going to be shorter than what you have here though

edn is generally going to be much slower to parse than transit. If you want data for people to read/edit edn is better. If your concern is sending data or reading/writing data between programs, transit is better/faster

No

But if I imagine where I’d put features as a language implementor, that’s one place

Thanks 🙂 But numbers cannot be used with juxt, right?

((juxt 0 1 3) "abcdefg") ;; error

Each index should be wrapped in #(nth % index).

Just out of curiosity, did you search for this by using a regex with http://grep.app?

I did not

Mind sharing how you found that the library uses metadata that way? 🙂 Just want to hone my searching skills too

someone tried to use it with #babashka and bb didn't support that syntax yet

If you want to search code using clojure.spec regexes, you can use: https://github.com/borkdude/grasp

Sweet, I do know about this lib from one of the clojure podcasts ✌️:skin-tone-4:

I say print compiler warnings for next release, then kill it... 😅

Never

There is no reason to break people for no reason

Like I said, it’s not shorter :)

what would be the bestway to handle maps(google maps/ mapbox) in clojure -> integrate it in js or handle it with clojure

Can someone share article / thoughts about reasons of consistent google trends Clojure dropping? It can be subjective, but I also see less traffic on slack, than a few years ago. What is happening? Do (Don’t) developers have fun using Clojure or only developers which should stay with Clojure? What changed in September 2020 Clojure trends drop so much? Pandemic? How?

This is just a guess but let's break it down by region:

Let's narrow down on China:

Very sharp drop off. Can it possibly account for the largest part of it?

Interest in Clojure in Finland seems pretty consistent (go metosin!)

good point

Israel Most of the traffic is probably from Appsflyer and during the pandemic our management took a very very conservative fiscal strategy and almost froze hiring. New hires do lots of googling

I think that explains most of it

Curious to know what changed in China

Do you know website with statistic how many developers use Clojure each year?

Makes sense. Thank you!

thank you

No, but you could refer to the annual Clojure survey

https://clojure.org/news/2021/01/14/clojure-2021-survey

I think people are not generally aware what a high percentage of traffic tracking is being blocked by browsers these days which makes me look skeptically at these numbers

We recently found as much as 90% of traffic to Datomic sites was being blocked from tracking so things like Google analytics were fairly inaccurate

good point too

That's catchier than Linus Torvalds's quote about this subject

My take is that Clojure has entered the “Plateau of productivity” phase of the hype cycle. I mean it’s been around for 14 years now so I think most of the novelty factor has worn off. (Although this doesn’t account for the sudden drop in Sept. 2020) (Also worth remembering is that if Google Trends is a valid quality metric we should all be coding in Java / JavaScript 😉)

Another thing is that mainstream adoption takes time; after the novelty factor has worn off you have to scale a way larger hill, which is to become a battle-tested language that’s regarded as “production-ready”. Erlang lingered in obscurity for about two decades before the rise of instant messaging created a demand for the same kind of throughput and robustness that was a solved problem in telecom. TBH I don’t think the world “at large” is ready for Clojure; Python has begun to overtake Java / .NET in just the last few years, so I think it’ll be a while before “the industry” is ready for functional programming

That said, I personally believe that Clojure is uniquely well-suited for what’s going to be standard application development in a few years, which is to move yet more towards data processing and enabling systems to talk to each other. Most of my work nowadays is just building APIs that string together other APIs. Strong type systems and OOP are not much help here, but “interactive” (e.g. REPL-driven) development, persistent data structures, easy serialization and schemas / data-driven systems sure are.

Well you can put quotes around that and attribute it to rich