I think that's Running With Scissors: https://github.com/matthiasn/talk-transcripts/blob/master/Halloway_Stuart/RunningWithScissors.md where he found a bug in the chart library?

Oh right it was a charting library! Yes this is the one, good memory 😁

I can kind of see it being a reader thing, how :: keys must be known in advance in order to be resolved, as opposed to the concrete qualified syntax that has no outside dependence.

(ns-name (get (ns-aliases *ns*) 'itunes)) should give you qualified namespace symbol which alias refer to.

user=> (require '[clojure.data.xml :as xml])
nil
user=> (xml/alias-uri 'xh "")
nil
user=> (def xh (name (ns-name (get (ns-aliases *ns*) 'xh))))
#'user/xh
user=> (keyword xh "summary")
:xmlns.http%3A%2F%2Fwww.w3.org%2F1999%2Fxhtml/summary

Excellent, thank you very much!

It's a pity that :as-alias can't be used in this case in an obvious way

What would that mean?

i guess that means using URI instead of namespace in require with as-alias

I don’t see how that would be possible while still getting the uri aliasing

I guess it would mean evaling the alias name expression. I don’t think that’s a door we want to open

you mean like this ?

clj -e "(do (require 'clojure.test) (alter-var-root #'clojure.test/*load-tests* (constantly true)))" -r

That might be a reasonable var to add to https://github.com/clojure/clojure/blob/527b330045ef35b47a968d80ed3dc4999cfa2623/src/clj/clojure/main.clj#L83-L101 so that you can just set! it Worth an ask :)

Personally I use clojure.test/*load-tests* in conjunction with a specific way of using tools.namespace

A jvm property is also good 🙃

namespaces don't exist in a hierarchy like that

there is no relationship between my.ns.foo and my.ns.bar other than the name shares a common prefix

and in clojure there is nothing called a "module"

sub-namespace then, but you've answered the question. So I guess my other option is to use in-ns to separate "modules" into multiple files and then use something like ns-publics to access their vars. I seem to remember having problems with that strategy previously when it came to hotloading changes to the namespace

Thanks for setting me straight on the namespaces.

"sub-namespace" implies some kind of hierarchy between namespaces still, and there is none. it is a flat key space (a single k/v map not a filesystem of nested trees)

I got that, thanks. I was just clarifying what I meant by "module"

https://github.com/techascent/tech.ml.dataset#parquet-support

Schema$Parser is the name of the class

and you need to import it explicitly

Parser is a nested class in Schema in java, and the way that gets compiled for the jvm is Parser generates a class named Schema$Parser entirely distinct from the Schema class

This way?

(:import
   [org.apache.avro Schema$Parser] )

you need both Schema and Schema$Parser

Why both?

Because eventually you'll https://github.com/nomnom-insights/abracad/blob/f7b399e323b7972e9fe631721ab035500623f7a1/src/clojure/abracad/avro.clj#L22-L24 - at least to avoid reflection

Ah, I thought you meant for just that line. Right.

it works! tks, guys! Can I ask another thing? How can I rewrite this piece of code?

ParquetWriter<GenericRecord> writer = 
  AvroParquetWriter.<GenericRecord>builder(path)
                   .withSchema(schema)
                   .withCompressionCodec(CompressionCodecName.SNAPPY)
                   .build()

this 'Generics' thing... I've no idea how to code this with clojure..

Write the code as if the <...> parts don't exist.

I've trid this way, but do not work:

writer (-> (.builder AvroParquetWriter path)
         (.withSchema schema)
         (.withCompressionCode CompressionCodecName/SNAPPY)
         (.build))

I got this error: "No matching method builder found taking 1 args for class java.lang.Class"

Because builder is static, it should be (-> (AvroParquetWriter/builder path) ...).

thank you very much!

Btw. there's #C053AK3F9 channel which would be better fit for these questions

Schema$Parser

no

(long answer, depends what you mean by process, but if you mean a os process, then the only way to run something that outlives a given process is to run another process)

So i thought a future would do it

(future (do-long-running-thing))

a future is another thread

but i can't tell if it is

so i lost the connection to the REPL and it was unclear if the other thread was killed too

if the process that was hosting the repl is still running it is likely the future is still running

oh interesting