@bj: try-let is my project 😄 I'm not aware of a better ready-made solution to the problem. I wrote it to scratch my own itch a couple of years ago, but it should work fine

How can I get with-redefs to work on a private function in another namespace?

(with-redefs [my.other.ns/fn-name (fn ...)] ...) is working for you? I was able to get that working in a REPL session test

@andy.fingerhut I didn’t for me. Ill have another go and try to see what I’m doing wrong

Calling the fn directly works for me. My function I’m wanting to redefine is in a partial though

(def my-partial (partial fn-name :foo))

fn-name is private in that namespace and redefining doesn’t seem to work

My REPL test went like this, in case it helps you (starting from user namespace): (ns foo) (defn- bar [x] (inc x)) (ns user) (#'foo/bar 10) ; returned 11

(with-redefs [foo/bar (fn [x] (* x x))] (#'foo/bar 10)) ; returned 100

Yea that works for me too.

I think the with-redefs wont work for me here though

You area trying to with-redefs my-partial ?

yea, I’m not atually invoking the function Im redefining. Im invoking the return value from partial

And the partial is invoking what I pass into it.

Again, not sure if it helps, but continuing my sample REPL session from above: (def mybar (partial #'foo/bar 10)) (mybar) ; returned 11

(with-redefs [foo/bar (fn [x] (* x x))] (mybar)) ; returned 100

Is your def of my-partial in the other namespace, and marked private?

the partial is in the same ns as the private fn im wanting to redefine

Im thinking the functions visibility is irrelevant here

And because it uses (partial private-fn ...) instead of (partial #'private-fn ...) you're getting the original value compiled into the partial expression -- so there's nothing to redefine.

Per @andy.fingerhut’s example, the partial call needs to use the Var not the symbol to give you the extra level of indirection.

But if fn-name is in another namespace and is private, how do you even refer to it in the partial call, without prefixing it with #' ?

ahh thats it

yea I need to use the symbol

> the partial is in the same ns as the private fn im wanting to redefine ^ This

ah, sorry, missed that.

(def my-partial (partial fn-name :foo)) doesn’t work and (def my-partial (partial #'my.other.ns/fn-name :foo)) does

Yup.

The former has the (partial fn-name :foo) compiled at namespace load time and bound to the Var my-partial. And fn-name is evaluated to the value it has -- its underlying fn value.

@andy.fingerhut @seancorfield Thanks for the help guys! I understand too now which is awesome!

The latter is still compiled but uses a Var instead -- for the reference to fn-name -- and maintains the indirection.

so a symbol like fn-name will evaluate to its current value and later when it is called it will invoke whatever it initially evaluated as. By using the #'some.ns/fn-name, when invoked later will invoke whatever is currently defined for that symbol?

I think you've got the right idea. I believe that the (partial fn-name :foo) expression without the #' captures and uses the current value that the var fn-name at the time that partial expression is evaluated, and changing the value of the Var fn-name later does not change the value of the partial expression.

(partial #'fn-name :foo) captures and stores the value of #'fn-name, a pointer/reference to the Var fn-name, not to its value.

As a convenience, Clojure Vars implement an interface in the internal Java implementation that causes their current value to be looked up and called.

Right, it only matters when you use a function name in an expression, rather than in another function. In an expression, everything is evaluated, inside a function it's just compiled. (well, slight simplification there but...)

(consider that top-level def forms are evaluated when the namespace is loaded so your :require causes the evaluation)

Right so invoking the function (wrap in an anonymous function) will look up the value of the symbol and would work with redef and no #‘. But passing the symbol for the function straight in with #’ doesn’t? Am I on the right track?

In case the details might be of interest, the file Var.java in the Clojure implementation defines class Var to impement interface IFn, and IFn is the interface that Clojure functions implement.

is there a way to define a wrapper around clojure.test/is where it runs in cloljure modce, but is a no-op in clojurescript mode (this is . clojuretest/is in a *.cljc file)

Something like #?(:clj (is ... )) ?

is there an idiomatic naming convention for functions which return functions?

sometimes they start with an f (`fnil`) or end in -fn (`some-fn`)

@noisesmith ok, thanks, I’ll run with that. Now that I look at that I think I might prefer the leading f. Less of an afterthought, more in your face

\\

Hello. Last week I registered on http://www.4clojure.com and up to date i'm solving the bigger part of it tasks. But. I f I solved the task, I even can not to see any solutions of other members (that is possible on Codewars, for example) - I must follow each of choosen members personally. Why I write this here? Can you suggest me some (3-5) members, that I can follow to learn good Clojure-way style of problem solving and code organize? Thank you.

I haven’t visited 4clojure in a while, but as I recall, when you solve a problem there is a link in the “congratulations” text that you can click to see other people’s answers.

If I wanted to read in a bunch of parquet files and fiddle with the data in them, what's the lightest-weight way to do that?

@manutter51 unfortunately only this >You can only see solutions of users whom you follow. Click on any name from the users listing page to see their profile, and click follow from there.

chased last one for 20 minutes -_-

(merge-with (fnil - 0 0) {:a 0} {:a 2})
=> {:a -2}
(merge-with (fnil - 0 0) {:a nil} {:a 2})
=> {:a -2}
(merge-with (fnil - 0 0) {} {:a 2})
=> {:a 2}

right, if there's no matching key the merging function is never called

it is so obvious it this example, but was not that obvious in

(merge-with
  (fnil - 0 0)
  {:gold 478 :wood 20}
  {:gold 450 :wood 20 :stone 20})
=> {:gold 28, :wood 0, :stone 20}

actually, most of the time debugging was spent on getting to the merge-with call, because this call is from days ago, and was considered correct

if you had used merge-with + and negative numbers, you wouldn't have this issue - but I still think that merge-with is a poor match for this algorithm

replaced it with reduce-kv

and it was faster to use merge-with - than to update 600 lines of edn with - in prices : )

does #?(:cljs ... #clj ... ) mess up line numbers? the line numbers seem to not match up with my eidtor line numbers

is there something that provides "core.async channels" over websockets? I don't need the full power of core.async, what I want is a channel taht: 1. provides FIFO 2. at most once delivery 3. back pressure 4. wocks cljs <-> clj over websockets