Clojurians Log v2

what's up with clj-kondo in vscode, it seems like it can't see dependencies and fill up files with unresolved symbols in the editor

@clypto Probably a better question for #clj-kondo and/or #vscode / #calva ?

(although it sounds like you haven't run the project initialization for clj-kondo per the README?)

maybe, but I'm not sure where said readme is

Er, the clj-kondo readme?

https://github.com/borkdude/clj-kondo#project-setup

ah, I was looking at the VS Code extension page

which is quite sparse

I think I had grand ideas on what the extension automatically did 😄

Yeah, I had to blow away the .clj-kondo/.cache folder in one of my projects today and re-run the parallel lint initialization thing again because it got stuck thinking a couple of functions had different arities after some refactoring...

Of course, but I don't want to have it lying around for more time than built jars are lying around

...because for the build tooling, pom.xml is a mutable place that different tools write to or read from

If you refactor using the same editor that should not happen, since it lints files when editing and the last refactored file has the last known valid arities

Hey folks, I'm struggling a bit with variadic functions this morning– I have a function defined like:

(defn create-node [node-type props & children]
  ...blah)

Why is that the wrong number of arguments? I want that call to be valid and for children, in that case, to be an empty list

I don’t remember the answer to your question about arity, but the “create-node” name reminds me of a bug I ran into where I used a variadic argument to pass in some child nodes, and it blew up as soon as I tried to create a node with a large number of children. There’s a limit on the number of arguments a Java method can accept, so just a caveat in passing.

ok, thanks, that's definitely helpful

I just tried taking the list of children manually instead of via a variadic parameter, and I'm hitting the same "wrong number of argugments" still

create-node can call itself recursively... would that have anything to do with it?

Could be, depending on how it’s recursing

Your original defn looks correct to me, fwiw

What’s the exact text of the error? It should tell you how many args it was expecting vs how many it got.

WARNING: Wrong number of args (3) passed to cljs.core/create-node at line 12

It works for me ... Is create node recursive? Is there something in ...blah that's actually causing the error?

% clj --main cljs.main --repl
 ClojureScript 1.10.758
 cljs.user=> (defn create-node [node-type props & children] :blah)
 #'cljs.user/create-node
 cljs.user=> (create-node "x" {})
 :blah
 cljs.user=>

and this is my new fn def:

(defn create-node [node-type props children]

Ok, that’s sounding like your REPL has stale code stuck in it somehow, can you restart?

@nickt It says cljs.core/create-node which seems wrong

oh, derp, yeah, I missed that, good catch

That means need to refer clojure exclude create-node

or name your function something different

ohhhh man, does clojure have its own create-node?

it seems cljs has

Apparently cljs at least does.

lol

Try (doc cljs.core/create-node)

oh mannnn

wow, well thanks y'all. That would have taken me a long time to figure out 🙂

(clojurescript newb)

so how would I refer exclude it?

also, side question, is there a good way to accept a variadic children or an explicit list of children?

it seems my best bet would be to make that definition be [node-type props & children] but then check if the first element in children is itself a list. If it is, use that as my children array?

I would recommend not doing that and just picking one or the other

It’s just easier to understand and maintain

@alexmiller thanks, definitely agree, but I would like to enable syntax like (create-node "x" 1 (create-node "y" 2))

while also enabling syntax like (create-node "x" 1 (map (fn [child] (createNode...)) (range 10)) yknow what I mean?

Maybe there's some macro sugar I could use to enable this while preserving a singular notion of create-node's interface

I've only ever regretted doing this

or make two fns

@nickt In that case I would make create-node accept always a list of children. [(create-node ...)] is not much more typing than (create-node ...). or indeed create another fn

that's probably what I would do as well (list)

hmmm ok, yea, fair

btw I think props like {"x" 1} is also better than passing those separately

why's that?

you could have something like INode where you represent multiple children as a NodeSeq which implements INode so passing multiple or one is valid in terms of your model about passing props: easier parsing, passing, etc

Take a look at hiccup, reagent, etc

iiinteresting

oh man, I should have just started here

https://github.com/reagent-project/reagent/blob/master/src/reagent/core.cljs#L17

hmm ok, lot for me to think about. Thanks @borkdude and @alexmiller

Is there a better way to use a different default printer than just starting a subrepl? for example if I want my repl to always use https://github.com/brandonbloom/fipp to print, I have the following evaluated when I start a repl:

...
(:require [clojure.main :as main]
          [fipp.edn :refer [pprint] :rename {pprint fipp}]
          [fipp.repl :refer [pst]])
...
(main/repl
    :caught pst
    :print fipp
    )

Does anybody know any good way to parse nested query strings in Pedestal? It is currently able to automatically turn qs in maps (`?q=1&b=2` => {:q 1 :b 2} ), but I would need something a little bit more elaborate, such as ?foo[bar]=baz => { :foo {:bar "baz" } }

do you control the query params generated by client?

Yes I do @vlaaad; I can reshape it in case there's a format that Pedestal will deserialise automagically

IIRC query string is by definition is just a string

e.g. k1=v1&k2=v2 is pure convention

just dump your edn there 😄

Ah, that ain't a bad idea after all

Does Pedestal auto evaluates EDN strings to be a real map?

that might be a dangerous idea 🙂

Well yeah, if that calls functions and stuff

(I would be assuming Pedestal would be doing some sanitisation?)

I would regularly use a body, but this is a GET call and I want to go through a query string

nono

…but I do need to pass a deep object

clojure.edn/read-string is safe

…and I do have a spec for my object

this is a dangerous idea if you’ll want to make this endpoint consumable by other clients as well

e.g. public

My HTTP endpoint should receive a ::children from the Query String (somehow)

not dangerous — perhaps annoying at best 😄

it’s just an encoding

idk, I guess I can see the argument for, “EDN doesn’t have an encoder in every lang.”

that’s my argument, yes :)

Well as long as there's a JSON to EDN converter I'd be ok I guess 🙂

Is there a way to declaratively set a spec check on a payload in Pedestal or shall I manually put this boilerplate every time? https://github.com/XVincentX/clojure-playground/blob/master/src/app/core.clj#L21-L23

I would write an interceptor to do that

http://pedestal.io/guides/what-is-an-interceptor

Ok thanks @codonnell — I was thinking about the same, just wondering if Pedestal had something out of the box

There is likely a library that provides such an interceptor, but I'm not sure how worthwhile it would be to bring in a dependency for that. Shouldn't be more than a few lines of code, and it's nice IMO to have full control over the behavior.

I generally agree with others clucking their tongues over this, but for geenrating children it is definitely mad useful to be able to supply an atom or list or even nested list and then Just Flatten(tm) the input. The one constraint of course is being sure we can live with flatten ! In my fav use case I am building children for a UI/Dom node and I know flatten will stop when it gets to a node, and also I knw the parent demands a flat list of child nodes. Now I can bung together whatever I want creating children and have it all flattened before hitting the parent. hth

I will check now on what clj flatten does. (I have my own from my CL days.)

Yes, makes sense. I'll see if I can come up with some code. Thanks!

Ah, no, (flatten 42) => nil, we want 42.

Here’s my wrapper for any thing from an atom to a nested list of atoms. The *par* bit can be ignored. The doall is likewise sth specific to my usage. Note that the rest arg guarantees some list will be seen by the innards to avoid the flatten discard of atoms. Oh, and we gracefully handle some nested generator deciding not to create any nodes, a classic in reactive/dynamic UIs, and discard nils.

(defmacro the-kids [& tree]
  `(binding [*par* ~'me]
      (assert *par*)
      (doall (remove nil? (flatten (list ~@tree))))))

Hello guys! I made a reservation system. The mechanism is: 1. read from mongo->decide if there is free space 2. if free, insert reservation, if not free, don't. This function is called add-reservation . My problem is, add-reservation is behind a server. I would need the to run only one add-reservation at a time, because there is a concurrency problem when multiple run at once. I use agent and send reservations to it. Is there a better way?

Locks are commonly used for this kind of stuff, but that's beyond Clojure's scope

well, locks are fine if you need one. an agent is a good solution, assuming you have only one server

Definitely, locks only make sense if you have more than one server. Maybe I misunderstood the question :thinking_face:

Hello! Quick question: what’s the term for an aggregate (e.g. a sum) that isn’t declaratively specified, but instead is updated imperatively on every event? e.g.

total_payroll = select sum(salary) from employee; 

total_payroll = 0
on insert employee, total_payroll += salary
on delete employee, total_payroll -= salary
on update employee, total_payroll += (new_salary - previous_salary)

I'll check the locks too! thanks!

running total? rolling sum?

CIDER uses fipp for all projects if you set it as the default (setq cider-pprint-fn 'fipp) Does the editor you are using support something similar?

@paul931224 Maybe not super helpful if its set in stone, but Mongo probably isn't the best database for this

it sounds like you actually want strong consistency

and your ultimate data size won't be that huge

so...just use a sql database

then you can do that whole process in a transaction

you will run into more scaling problems trying to have what would effectively need to be a distributed lock on the reservations collection than with scaling a sql db

if you don't need to scale ever and really can just have a single server for all time then server side locks would be just fine

i know there is some degree of transaction support in mongo

but truthfully I don't know that much about it

namespace and protocol question: i have a protocol CostFns in game.core.abilities, and then (:require [game.core.abilities :refer [CostFns]]) it in game.core. I wrote a factory function to store the records that implement these in a var, so I can pass in a keyword to get the right record out, and if I print out the calls, the var is correctly filled during compilation. However, at a later point during compilation an error is thrown, saying that No implementation of method: :cost-name of protocol: #'game.core.abilities/CostFns found for class: game.core.Click

has anyone run into something like this before?

yes, you've reloaded the protocol at some point

ah, okay

the protocol, when compiled, makes a Java interface and the records implement that interface

if you reload the protocol, you make a new interface (same name), and the old records don't implement it

oh interesting, okay. thank you

hey all. Another quick question... I'd like to write a macro that takes N arguments such that if the first argument is a map literal, do nothing, otherwise inject an empty map at the front of the N arguments list. Seems straightforward enough, but how do I check the type of that first argument in a macro?

that's only possible if the map is a literal, you can't check the type of an arg otherwise, it's just a symbol, you don't see the value the symbol resolves to locally until runtime

of course the map being added to the arg list could be done at runtime, you just need to emit code that checks and modifies the arg list in that case

hmm ok, yea, makes sense

(defmacro foo-bar [& args] `(let [arg-list# (if (map? (first ~args)) ~args (cons {} ~args))] ...))

cool, thanks!

ok unrelated question: I'm aware of list? and vector?. Is there a similar predicate that is true if either list or vector?

perhaps you want sequential?

that is also true for lazy-seqs and conses, which are not lists and are false for list?

in fact, list? is so rarely useful I consider it a smell in code review

ah, awesome

oh, duh, map? won't work for me in that argslist check because all of the args will be maps heh. What's the idiomatic way to basically "type" a map? As in define a map as a type A or not

deftype?

you could have a special key or metadata, or use defrecord to have a map with a named type

how does that first map have to be different from the rest?

defrecord was the one! Thank you!

@codonnell Here it is. Likely not the best code ever, but it does the job.