hello, good morning! do you know if there is a simple way to use a function without arguments as a custom generator for a spec? something like:

(spec/def ::repository (spec/with-gen #(satisfies? AProtocol) (gen-from-function my-aprotocol-instantiator)))

the only way that i found is to use a combination of gen/return and gen/fmap: return to “start” the generation and fmap to call the function ignoring the value generated by return; this is a very ugly hack. 🙂

@andy.fingerhut hmm yeah, I want a way to test this behavior on CI

Hi all, I have the following function:

(defn ->attribute-value [attribute-spec value]
  (cond
    (as/view? attribute-spec :user)
    (->byte-buffer value)

    (as/name? attribute-spec
      #{:creation-time
        :last-access-time
        :last-modified-time})
    (->file-time value)

    (and
      (as/view? attribute-spec :posix)
      (as/name? attribute-spec :permissions))
    (->posix-file-permissions value)

    :default value))

I was thinking of using a protocol but I'd need some way of registering implementations if I took that approach

looking at your code, I don't really understand what exactly is extensible

what do you want to dispatch on?

So my library calls this function internally and I want consumers of the library to be able to register new conversions for future views and names

views? names?

"view" means there is a key :view in input map?

The attribute-spec looks like {:view <some-view> :name <some-name>}.

Yes

So in some cases, I want to apply a conversion regardless of name, based only on view

In others based only on name, regardless of view

(defmulti ->attribute-value (juxt :view :name))?

And others, more specifically based on a view and a name

In that case how would I match any name for specific view?

> I want to apply a conversion regardless of name, based only on view (defmethod ->attribute-value [:my-custom-view :default] ...)

> In others based only on name, regardless of view (defmethod ->attribute-value [:default :my-custom-name] ...)

In the view only case, the map coming in would look like {:view :user :name :some-name} so (juxt :view :name) would yield [:user :some-name].

Which would look for (defmethod ->attribute-value [:user :some-name] ...) and wouldn't match a default.

I think I'd need pattern matching or predicate dispatch to be able to achieve that. However in that case, it would be hard for users to register new matches or predicates.

ohh, :default does not work in vector...

that's strange

:default is a bit of magic

this works though:

(defmulti ->attribute-value (juxt :view :name))

(defmethod ->attribute-value [::my-custom-view ::default] [_] "my custom view")

(defmethod ->attribute-value [::default ::my-custom-name] [_] "my custom name")

(derive ::blep ::default)

(->attribute-value {:view ::blep :name ::my-custom-name}) 
;; => "my custom name"

it's not really a magic, it's configurable and a part of a hierarchy

my mistake, :default is a part of a multimethod, not part of a hierarchy, that's why it does not work in isa? check

Yeah, what I mean by a bit of magic is it's hard coded into the definition of a multimethod

If I use hierarchies, the user has to derive from :default for every possible name.

Really all I need is a :* entry that matches anything, so that I can say (defmethod ->attribute-value [:user :*] ...) or (defmethod ->attribute-value [:* :creation-time] ...)

I think it's possible to do with custom hierarchy

you can supply custom hierarchy to multimethod

It looks like for a (defmulti ...) you can override the hierarchy

Haha, yeah

yes, and!

I'd need a way to say :* is a descendent of any other keyword

Not sure how to achieve that

if you look at isa? source to see how it uses hierarchy, you'll see this: (contains? ((:ancestors h) child) parent)

your hierarchy can be a map with :ancestors key, that instead of whatever is there by default is actually a function that makes this code return true

Ah yes, nice

it will screw up derive for this particular multimethod, but you probably don't want that anyway?

Yeah that would be fine

Ok, I'll give it a go, thanks

@toby924 behold the horrors of abusing implementation details!

(defmulti ->attribute-value (juxt :view :name)
  :hierarchy (atom {:ancestors (constantly #{:default})}))

(defmethod ->attribute-value [:my-custom-view :default] [_] "my custom view")

(defmethod ->attribute-value [:default :my-custom-name] [_] "my custom name")

(->attribute-value {:view :blep :name :my-custom-name})
;; => "my custom name"

Nice

Does seem a little hacky though 🙂

The docs are insistent that you use make-hierarchy for any custom hierarchy

yep, it abuses implementation details

Hey Guys, I am trying to use http://clojurequartz.info/ for scheduling periodic jobs for one of my Clojure projects. I am wondering, what tools/libraries I can use to monitor (web UI) to monitor these jobs? Any sort of pointer would be very helpful. Thanks in advance. :spock-hand:

I think there are quartz dashboards, I remember one being mentioned in the documentation

https://github.com/royrusso/jwatch Google turned this up and also quartzdesk

I have a "what's the most beautiful way" kinda question. I have a function get-user that takes a user id and gets the rest from a database. I also have a looong ole list of users (currently in a vect, but I can change that if needs be). Whats the best/most beautiful way to parallelize the calls to get those?

pmap if you want parallelize only this

The only downside I know of there is that if one of them dies you end up chugging through the rest of the threads before you can move on, right?

https://clojuredocs.org/clojure.core.async/pipeline-blocking might give you a little more control over the rate at which you hit the service? Depending on the length of time an average query takes, it may be worth it to "batch" your requests within pmap - to avoid spinning up a new thread for each user: i.e.

(apply concat (pmap (partial map get-user) (partition-all 100 user-ids)))
;; ^ will fetch 100 users in a single thread, instead of 1 user per thread. This greatly reduces the amount of time spent managing thread versus: 
(pmap get-user user-ids)

Ohhh, that's pretty. 👏

@borkdude In school the way we tested stuff like that was by first factoring out the input and output into Readable and Writable "things" (I forget the exact java interfaces we used)

then for unit tests we would pump some input into the code and inspect its output

in the simplest case by just passing a StringBuilder for the output

so we had stuff like (pseudo-clojure)

(we didn't actually write clojure, but i think you get the idea

@emccue that's a good strategy. one issue is that the PR reads from (io/reader "/dev/tty") and I have no clue how that behaves over different operating systems, that's why I want a test for that in CI

but maybe replacing /dev/tty by an explicit reader thing and testing that is sufficient

Well there are really two components

One, given a stream of characters does your thing do the right thing

Two, is each specific source of input a valid source of input

right

That second one is the harder one to validate

But there is value in separating the first one even if you can't think of a way to do that validation

yeah

Does anyone happen to have any experience using @seancorfield’s clj-new.generate feature and has a working CLI and associated generate function?

ahh nvm found the source code 😊

i’m finding during profiling that the thing seemingly taking up the most time is the recursion call. i’m not quite sure what to make of that, any insights?

what do you mean by "the recursion call" and how are you profiling?