Is there something like defn but for a memoized function?

I wrote

(defmacro defmemo [name args & body]
  `(def ~name
     (memoize (fn [~@args] ~@body))))

Is it possible to do a string replace using a symbol as the pattern to match?

@grounded_sage you mean where the name of the symbol is the pattern itself? i don't quite see the value, but yes you can convert a symbol to a string, and a string to a re-pattern:

(clojure.string/replace "123abc789" (-> abc quote str re-pattern) "456")
=> "123456789"

If i'm implementing a tree with Node and Leaf deftypes, how can I write methods that access .-keys for each type of node without reflection? Should I write getter functions for these fields and use those?

oracle jdk SE download page provides 2 version: 8u111 and 8u112. which I should download?

what is oracle jdk even / odd numbered version means? is there any convention?

Is there a core function that partitions a collection in two parts based on a predicate?

I know group-by, but I don’t need a map

@mavbozo try #beginners channel

@shooodooken thank you. I will ask there

@borkdude split-with

well it's not exactly "partition"

partition-by tho 🙂

guess it depends on the kind of check you need to apply, might not be always suitable

I already looked at both but they don’t seem to do what I want. Group-by comes closest

Also group-by is less ambiguous and more flexible of course

E.g. what comes first, all results for which the pred is false or true? With group-by that’s clear without looking at the docs

Anyway, happy blue monday!

🎆

danielstockton: could probably use a protocol… it sound like you’re wanting some of the features of defrecord though- so might want to consider using that; otherwise you’ll need to implement those bits you need on your deftype

Let’s say I have some (def x (reify ....)) in a namespace — what would the "fully qualified class name” of x be? Tried $x but that din’t work...

@martinklepsch reify creates an anonymous instance

the classname is unspecified

(as in, it's some munged random name you can't rely on)

Oh, ok 😄 would I use deftype then?

if you need a named class, yes

I use a protocol @rickmoynihan which avoids reflection on methods but not fields apparently, which is why i'm wondering if i should just wrap the field in a lookup method.

Sometimes I think I’m already quite proficient interacting with Java stuff but then things like this come along 😄

defrecord is another option, i want it to be as performant as possible though

(where you have to pass the fully qualified name of a class as a string to something)

@danielstockton there's no perf difference between deftypes and defrecords -- a defrecord is just a deftype that implements a bunch of default interfaces to behave like a map

if you type-hint, (.-some-field ^ClassName the-type-or-record-instance) it will not use any reflection

I know @bronsa but i don't know the class at compile time, just trying to remove the reflection warning

im worried i may want different implementations of some of those default interfaces that defrecord already implements

how do you not know the class at compile time if you're doing field access? do you have multiple deftypes with the same field names? in that case a protocol function to access them should be the right abstraction. if using kw access using defrecords is out of the question

yeah, multiple deftypes with same fields...protocol function it is then

so with a (deftype MyThing []) in com.my-abc would the fqn be com.my_abc$MyThing?

no, com.my_abc.MyThing

ha, should have just tried that. Remember something about inner classes and thought I need the `… thanks @bronsa

(and sorry for asking really boring questions haha)

Another Java related question: How would I rewrite this to use reify — and should I at all?

(proxy [KeyedPooledObjectFactory] []
                         (makeObject [k] (DefaultPooledObject. (make k)))
                         (destroyObject [k ^PooledObject o] (destroy k (.getObject o)))
                         (validateObject [k ^PooledObject o] (validate k (.getObject o)))
                         (activateObject [k ^PooledObject o] (activate k (.getObject o)))
                         (passivateObject [k ^PooledObject o] (passivate k (.getObject o))))

I tried something like this but it complains about wrong return types for destroyObject (should be void is Object)

(reify KeyedPooledObjectFactory
  (makeObject [_ k] (DefaultPooledObject. (mk-session k)))
  (destroyObject [_ k ^PooledObject o] nil)
  (validateObject [_ k ^PooledObject o] nil)
  (activateObject [_ k ^PooledObject o] nil)
  (passivateObject [_ k ^PooledObject o] nil))

(reify KeyedPooledObjectFactory (makeObject [this k] (DefaultPooledObject. (make k))) ..)

type hint the argvec

^void [_ k ..]

hm, same result for this

(destroyObject ^void [_ k ^PooledObject o] (destroy k (.getObject o)))

gah, try (^void destroyObject [..] ..)

that did the trick!

@bronsa Thanks!

🙇

I am trying to integrate jmx metrics into my Clojure app

Using https://github.com/clojure/java.jmx

Is it possible to set the type of a custom mbean

something like java.lang:type=Memory

um. do we have a rails analog presence function in clojure?

(presence nil) ; nil
(presence ""); nil
(presence "hi"); "hi"

(presence []); nil etc

or in env var config (or (presence config-var "default"))

(defn presence [x] (when (seq x) x))
=> #'user/presence
(presence nil)
=> nil
(presence "")
=> nil
(presence "hi")
=> "hi"
(presence [])
=> nil

it satisfies the requirements, but probably you want it to work for Number too 😛

How can I compose a regex out of variables?

re-pattern creates a regex from a string that represents the pattern

Does subvec keep a reference to whole original vector or only to the relevant part?

That is, can I subvec a needed small piece of data out of a huge vector and don't be worried about retaining the whole huge vector in memory?

Is there something I can call to check CPU utilization on a clojure function?

Answering myself after some REPL experimentation: subvec does retain whole original vector in memory.

interesting @dottedmag, i guess it has to as there is no easy clue which parts can be GCed

You think this is something specific to subvec? I'd expect it to be the same with conj for example.

Yes, it's specific, because conj does not create a smaller structure.

So, I suppose it's because subvec is supposed to be very fast, so there is no overhead of creating a new persistent data structure.

So, if one wants to pick a subrange out of a large vector and discard it, it ought to be done via (vec (subvec largestuff N M)), which will build a new persistent data structure.

no it won't

the only way is not to use subvec at all

or to do like (vec (seq (subvec ov n m)))

Interesting.

I was wrong.

but at this point just do (into [] (comp (drop n) (take (- m n))) ov)

which will probably be faster (unless n is really big)

@stijn , thanks, that actually works, however it's tricky to understand that it works like that) also it needs to work only for {}, [] and strings, numbers are ok)

@bronsa Why doesn't (into [] (subvec ov m n)) work? Is it due to fast path in into which says ah, it's a vector too, let's just not copy anything?

(into []) would work

it's vec that wouldn't

well defining "work" as in "would create a new DS w/o the old elements"

vec has the fast path check

Interesting. Let me try again, (into []) kept the whole data structure in memory for me.

Yes, PEBKAC.

After (into [] (subvec a 1 100)) (ns-unmap 'user 'a) (System/gc) memory is released.

@dottedmag side question, how are you measuring the memory consumption of these things?

@danielstockton Using jmc(1)

There's probably CLI utility as well, but this one works well for experimentation.

@leov: it works for strings because a string is a sequence of characters, if it is empty, it will return nil

http://www.oracle.com/technetwork/java/javaseproducts/mission-control/java-mission-control-1998576.html ?

@danielstockton Yes. It's a GUI tool which comes with JVM and creates nifty memory/CPU/whatever graphs.

ok, thanks

@dottedmag @danielstockton I dont want to be nitpicking here, but in case someone else reads this: The java mission control comes with the oracle JDK and is bound to the oracle license whereas it is not part of the OpenJDK.

It uses JMX to collect metrics, which is in OpenJDK, so there are probably other similar clients.

Yea, most probably 🙂

VisualVM (http://visualvm.github.io) for example

There has any way re-extend some protocol methods on record?

@stijn , turns out this function is named not-empty in clojure world

in Spec, how do I say that a map must have :req keys and that 2 of their vals depend on eachother ? like, one of the map's vals must be a collection and another one must in a number in the range (count coll)

@octo221: (s/and (s/keys ...) your-predicate-on-map)

@dottedmag yep that's done it - thanks so much!

I’m trying to connect to AWS Athena over JDBC (clojure.java.jdbc). I’m getting an error saying Exception in thread "main" com.amazonaws.athena.jdbc.NotImplementedException: Method Connection.prepareStatement is not yet implemented, compiling:(/private/var/folders/h1/pt7t6l9s2bq2zblx4v2mmnnh0000gn/T/form-init4382409266245799902.clj:1:125) Is there a way to query jdbc without using prepareStatement? I’m using (query db [”SELECT…”] {:row-fn :cost})

@erikcw Looking at the source, one way might be to try and use db-do-commands, looks like it uses Statement https://github.com/clojure/java.jdbc/blob/master/src/main/clojure/clojure/java/jdbc.clj#L760

Saying that though, maybe can't use that with a select because it uses executeBatch

but the source can guide you on doing what you need to with plain Java interop?

I would assume the Athena JDBC library provides a way to create a PreparedStatement object via some subclass? If so, you can pass that into query as part of the vector, so it doesn’t need to call prepareStatement itself.

@brabster Yeah, you’re going to run foul of what you can / can’t do with the various underlying JDBC execute functions.

This might be helpful... someone already did Athena from Clojure https://github.com/mikeplavsky/athena-cmd/blob/master/src/athena_cmd/core.clj

@seancorfield good idea on looking for other ways to create a preparedstatement but I don't see anything in the jar that looks like one. I'm guessing it's just not implemented yet at all

I would mercilessly copy from mikeplavsky's example and star him up on Github 🙂

Looks like it calls .createStatement rather than .prepareStatement so that would be the path I’d suggest — but that looks like a plain Statement rather than a PreparedStatement so I’m not sure java.jdbc will recognize it.

(since it specifically tests instance? PreparedStatement)

does the rich hickey/ cognitect crew have anything to say on the subject of release management? I'm moving into the role of "release manager"/"please help us fix our release process" person at my company. I'm looking for role models for modern release management practices.

you could look into @mtnygard blog posts: http://blog.cognitect.com/blog/?author=538e1607e4b0e4caa9812f47

not sure if they're about release management per se, but in the vicinity for sure

@tjtolton If you haven’t read Jez Humble & Dave Farley’s book “Continuous Delivery” that would be a great place to start.

I'll look into that book and your blog, thanks @mtnygard, and @pesterhazy!

don’t forget mtnygards own book on the subject: https://www.amazon.com/Release-Production-Ready-Software-Pragmatic-Programmers/dp/0978739213/

@tjtolton one of the key principles is “build-once” deploy everywhere. The Google SRE book also has a few chapters you might find helpful although don’t get overwhelmed by it. Pick the things you think will fit and benefit your org most.

google sre book

seconding michael's book, it's def worth reading

What's the easiest way to pick up a build time env var and pack it up into my jar so that I can use it in a function?

I'm using leiningen to build. Trying to build a URL using my Travis build number

characters are not strings

"1" is a string, \1 is a character

a java.lang.String and a java.lang.Character

thanks… oddly the char coercion function only operates on numbers

it operates on ints

is there a general function to return the type of a value? typeof or something?

because java characters need the range

type or class

thank you thank you

It's just going to be write the build number to a file during build and read that at runtime isn't it :)

it is pretty common to generate a file containing build related information as part of the build

whatever tooling you are using may already do that in some way

@hiredman aye thanks!