Like a marriage between repeatedly and delay

There's some relationship between them. In Joy of Clojure, they used it build a variant of lazy seqs: https://github.com/jumarko/clojure-experiments/blob/master/src/clojure_experiments/books/joy_of_clojure/ch06.clj#L291-L329

Ya kind of, at a high level.

a big difference is you can print a delay without forcing it.

What version of clojure are you using? I think that syntax was added in 1.11.

You use the following syntax with any version any version 1.2.0 or later:

(mk-pool :cpu-count 2)

ooo so (mk-pool {:cpu-count 5}) works in clojure 1.11? Im on 1.10.3

thank you!

is there some place i can read about this?

https://clojure.org/news/2021/03/18/apis-serving-people-and-programs

#(my-fn arg1 arg2) as a possible alternative, but personally I wouldn't care

Ah, thanks! And of course! I tend to avoid #() so it was out of my mental reach. 😃 I think I prefer #() over partial here, b/c the reasons in OP.

If you do it more than once, it warrants making a macro out of it. I like using $ for it:

(defmacro $ [& body] `(fn ~'thunk [] ~@body))

(($ (str 'foo 'bar)))
"foobar"

($ (str 'foo 'bar))
#function[user/eval63202/thunk--63203]

($ (str 'foo 'bar))
#function[user/eval63202/str-thunk--63203]

Not as simple, but sure:

(defmacro $ [& body]
  (let [name (or (and (sequential? (first body))
                      (symbol? (ffirst body))
                      (-> body ffirst name (str "-thunk") symbol))
                 'thunk)]
    `(fn ~name [] ~@body)))

($ (str "foo" "bar"))

out of curiosity, what are you thunking (about)?

I'm using a third party library and don't want to unit test it, so am using a ”command”-ish approach where I have different implementations for :test and for :prod (I'm using multimethods for it). For tests I don't want to write a lot of different implementations so have made just one that expects a thunk.

I ran lein deps :tree-data and there are a lot of suggestions on exclusions. It is a pretty large dependency tree so maybe that is expected.

What do these exclusions do really?

Generally, exclusions tell the dependency resolver to not include a particular dependency of an artifact (because a different but hopefully compatible version will be included in a different artifact).

Could be a set

Sets are functions, if you pass an element it contains you get the element back, otherwise nil

Don’t mind me, missed the other comments 😅

Isn't it more idiomatic to just use set as function?

(#{:butterfinger :snickers :zagnut} :zagnut)
=> :zagnut
(#{:butterfinger :snickers :zagnut} :nut)
=> nil

You can, but predicates should return true/false. You could also use (boolean (#{:b :s :z} x)) but I personally like the clarity of using contains? YMMV

Did you try (set! *math-context* <some-desired-value>) ?

Yes I did, it works on the repl but not on the built app. Can't change/establish root binding of: *math-context* with set

> Currently, it is an error to attempt to set the root binding of a var using set!, i.e. var assignments are thread-local. In all cases the value of expr is returned. https://clojure.org/reference/vars

If you don't start your app using clojure.main, the bindings won't be setup. Are you starting the app using a gen-class AOT?

In that case, you need to bind them in your main function:

(defn -main [& args]
  (binding [*math-context* *math-context*]
    ;; Now using (set! *math-context* ...) will work
    ...))

You can also just use this: https://clojuredocs.org/clojure.main/with-bindings So wrap your -main body in that macro

(ns foo
  (:require [clojure.main :as m]))

(defn -main [& args]
  (m/with-bindings
    ;; Now using (set! *math-context* ...) will work
    ...))

That said, normally those bindings are compile time options. I'm not sure about *math-context*, but my guess is it is as well. That means, if you want to set those globally, you can do it in the compilation itself. For example if you use tools.build: https://clojure.github.io/tools.build/clojure.tools.build.api.html#var-compile-clj You can use :bindings on the compile-clj step.

Usually, the easiest way to start a repl and connect with cider is cider-jack-in

clj -A:dev starts a repl, but cider-connect expects an nrepl

If you'd like to start the nrepl yourself, https://nrepl.org/nrepl/usage/server.html

oohhh ok that's great information. Maybe i am approaching this in a wrong way then. So i would like to cider-jack-in and i would like to do (start) which is in dev/user.clj At the moment when i cider jack in i have to evaluate user namespace and then i can (start)

What does (start) do?

it's just a wrapper for mount.core/start

(defn start []
  (mount/start)) 

I'm probably running an old version of cider and emacs. I set a the variable cider-clojure-cli-parameters to "-M:dev:project -m nrepl.cmdline --middleware '%s'" so that I can add extra aliases.

I also don't really use user.clj

i will ask there as well. Oh? Why not? How do you start your dev environment

I would usually just run it manually I guess

You could also try adding a defadvice to your cider-jack-in command to run it, but I'm not really sure what the best way is since that's not something I usually do

in your deps.edn dev alias you could add something like this:

{:extra-deps {cider/cider-nrepl {:mvn/version "0.28.3"}}
         :main-opts ["-m" "nrepl.cmdline" "--middleware"
                     "[cider.nrepl/cider-middleware,refactor-nrepl.middleware/wrap-refactor]"
                     "-e" "\"(mount.core/start)\""]}

that way running clj -A:dev will run an nrepl and execute mount.core/start, then you can cider-connect-clj to it.

This works, this is basically the comment cider jack in uses to initiate te repl. When i cider-connect i can use (start) right away without the need to evaluate the namespace. So what exactly prevents plain cider jack in on loading namespace ? this is my dev alias now

{:extra-paths ["dev"]
                 :extra-deps {cider/cider-nrepl {:mvn/version "0.25.1"}}
                 :main-opts ["-m" "nrepl.cmdline" "--middleware" "[cider.nrepl/cider-middleware]"]}

I think cider-jack-in might override some of that

If you do C-u M-x cider-jack-in it should show you the full command it uses (and allow you to edit it)

I think there are some variables you can customize to tweak the command it uses though

I thought i shared it...

/usr/local/bin/clojure -Sdeps '{:deps {nrepl {:mvn/version "0.7.0"} cider/cider-nrepl {:mvn/version "0.25.1"}}}' -m nrepl.cmdline --middleware '["cider.nrepl/cider-middleware"]'

It doesn't look like it's using the dev alias

true it seems that this might help https://docs.cider.mx/cider/basics/up_and_running.html#clojure-cli-options I will check it out and report back Thank you both for your time

You can do also do M-x customize-group and check all of the groups that start with cider

A .dir-locals.el file that defines the Clojure CLI aliases to include when running a REPL is a simple approach https://clojurians.slack.com/archives/C0617A8PQ/p1650665847187939?thread_ts=1650655838.638489&cid=C0617A8PQ

specs are just named with their keyword, so you can just use it from another namespace

you do need to ensure the spec has been loaded into the registry though, so you'll need to ensure the namespace has been required before use

I see, thanks

Add :join false to the hash-map that is passed to the jetty/run-jetty function to return control to the REPL prompt after jetty starts https://practical.li/clojure-web-services/app-servers/atom-based-restart.html

It wasn’t about the :join flag. I thought the jetty process was running as a REPL child process. So I tried to CTRL-C the jetty process at the REPL 🙂 now I understand that it’s running in the vs code terminal.

You need to stop your repl from the terminal it’s running in, which is not the Calva repl window in the right pane in your screenshot. In the terminal where the repl is running, hit ctrl+c.

But as was stated already, to prevent needing to do that again you’ll need to use the join flag.

So your running server isn’t coupled to your running repl, essentially.

now I got it running 🙂 Thank you.

No problem!

according to https://stackoverflow.com/questions/12389303/clojure-cons-vs-conj-with-lazy-seq it sounds like cons is actually lazy even though the docs don't mention it.

I would suggest fixing the docs for cons then 🙂 kind of confusing

No, cons is not lazy

A call to the lazy-seq macro suspends everything inside it until the seq is being realized

Yes, but cons also doesn't evaluate the elements

It does

Cons is itself a list type which just kind of cats lists together

But lazy seq doesn't

It doesn't care what they are made of

It doesn't evaluate them

It does

cons builds non lazy seqs

No it doesn't :) otherwise you would get infinite loops with even trivial uses

Nope

Usually calls to cons are wrapped in calls to lazy-seq which is what makes a lazy seq

No

Yep

I think you are wrong

I am not

Which is why the docs for cons don't say it is lazy, it isn't

check that stackoverflow where that guy has basically the same question as me

the answer is, it doesn't evaluate and return a new list

it is its own structure which lazy cats things to the beginning of another list

because seq isn't evaluated, its only evaluated when a seq gets to it

I feel pretty confident I know more about clojure internals than the majority of people answering clojure stackoverflow questions

thus, its lazy

You are misreading the answer

lazy-seq doesn't modify the functionality of other unrelated functions

which is what you are suggesting

The answer says when you cons a value on to another seq S, it does not force S

Which is correct

But does not mean cons is not lazy

lazy seq is a lazy seq, list is a list, you can't make a real value and pre-pend it to a lazy seq

You can

cons does exactly that

but if you read how its implemented, or at least his description of it, that is not what it does?

I mean like what would you do, in the case of a cons being somewhere inside of a lazy-seq, selectively create another lazy-seq which then transfers to another lazy-seq

thats bonkers

(clearly I'm confused 🙂 )

I'm not arguing to argue, I'm trying to correct myself by explaining it the way I understand it

Like, you can see that cons is not lazy, if you type (do (cons (println 1) (println 2)) nil)

Into the repl, both printlns run

But there is nothing that would force them if it was lazy

lemme try, 1 sec

no thats not what I'm talking about

And you won't get an error because println returns nil, so you are effectively calling (cons nil nil) where nil is taken to mean the empty seq, so that cons call makes a seq which is not lazy, that is has a single value, nil

in that case that you describe, (println 2) is evaluated of course, but if it was a (lazy-seq) it evaluates into lazy evaluated body etc...

so

I suppose what the disconnect is, is that lazy-seq can have both lazy and real elements in it

Seq is just a Java interface right, it can have many different implementations that do whatever

so clojure calls cons on the lazy-seq member function or something

No

cons returns a clojure.lang.Cons which is an implementation of Seq, and is not lazy

lazy-seq produces an implementation of Seq that is basically a delay

So the lazy-seq seq forces its delay(more or less) when you call first or rest on it, which produces the non-lazy seq made via cons, and the lazy-seq forwards the first or rest call to the non-lazy seq

The laziness of a seq is not really a property of the "whole" seq

It is a property of each node/cell/link that makes up the seq

oh geez

Perhaps this is more compelling evidence?

dev=> (take 1 (cons (println 1) (lazy-seq (cons (println 2) (println 3)))))
1
(nil)
dev=> (take 2 (cons (println 1) (lazy-seq (cons (println 2) (println 3)))))
1
(2
3
nil nil)

And, in modified form, if you extend it:

dev=> (take 2 (cons (println 1) (lazy-seq (cons (println 2) (lazy-seq (cons (println 3) []))))))
1
(2
nil nil)
dev=> (take 3 (cons (println 1) (lazy-seq (cons (println 2) (lazy-seq (cons (println 3) []))))))
1
(2
nil 3
nil nil)

(in particular, compare the two different take 2 versions)

Seqs are really linked lists, but just lifted to an interface instead of a concrete cons cell representation as in older lisps

It can be particularly confusing when you interact with other "lazy" functions that operate on chunked sequences because you'll get blocks of 32 items at a time, even if you only ask for a dozen items...

And the interface allows you to do things like lazy seqs (delays) or custom seqs for iterating over collection types (like most clojure collections have)

(but that's just an efficiency artifact of "lazy sequences" rather than lazy-seq)

alright, lemme mull over this a bit

But it is still fundamentally a linked node kind of thing

more questions later perhaps

this was a great and helpful thread though, thank you guys so much!

And, yeah, @U0NCTKEV8 is a pretty definitive source of correctness about Clojure internals 🙂

two examples that i think can help explain: (def l (cons :a (cons :b (cons :c (cons :d (range 1e6)))))) is l completely lazy or completely realized? it wouldn't make sense to make the cons lazy when we have concrete values a, b, c, d, etc. Cons isn't lazy there

oh and I'm definitely not, I'm just trying to get a more thorough understanding of it all

and (def l (concat (range 1e6) [:a :b :c :d] (range 1e6))) a "lazy" portion with a million elements, then a clearly realized portion with 4 elements, and then a lazy million element portion again

concat I get, yes

I was thinking cons is just like concat

or would have to be

but each individual linked list node itself being a totally arbitrary way to point to the next element in a linked list cell is so very much OOP that its probably right

that seems slightly inefficient

but whatever 🙂 not my circus

It's why we use vectors in Clojure a lot. They can behave like sequences but are much more efficient in a lot of situations.

While all three of these behave like sequences of 1, 2, 3, they are different types:

dev=> (type (list 1 2 3))
clojure.lang.PersistentList
dev=> (type (cons 1 (cons 2 (cons 3 ()))))
clojure.lang.Cons
dev=> (type [1 2 3])
clojure.lang.PersistentVector
dev=> (type (range 1 4))
clojure.lang.LongRange

As in:

dev=> (range 1 4)
(1 2 3)
dev=> (seq [1 2 3])
(1 2 3)
dev=> (cons 1 (cons 2 (cons 3 ())))
(1 2 3)
dev=> (list 1 2 3)
(1 2 3)

I have no idea if boxes-for-Java-objects-in-memory and arrows-for-pointers/references pictures of Clojure lazy sequences being realized one element at a time would be at all useful to you, but there is a library I wrote called cljol with a gallery of images I produced using that library, showing such pictures, with some associated text attempting to describe what is happening: https://github.com/jafingerhut/cljol/blob/master/doc/README-gallery.md