A follow-up to the discussion in #off-topic : Is there any practical difference between bound-fn and binding-conveyor-fn?

I’m trying to nail down exactly what a “symbol” can be in clojure.

I found the symbol regex in the EDN reader: https://github.com/clojure/clojure/blob/c6756a8bab137128c8119add29a25b0a88509900/src/jvm/clojure/lang/EdnReader.java#L27

But either I don’t understand Java regex syntax, or it doesn’t seem to match my understanding.

symbols seem to be officially described here: https://github.com/edn-format/edn

That also only defines what things the reader can read as a symbol

You can construct symbols from arbitrary strings using the symbol function

based on the official description, I came up with this mental model (which is a bit of short-hand):

- a (alpha): [a-zA-Z] - n (num): [0-9] - s special: [\.\*+!\-_?$%&=<>] - / slash: \/ a symbol can be: - single-char name: [as/] - multi-char name: [as][ans]+ - single-char prefix, single-char name: [as]/[as/] - multi-char prefix, single-char name: [as][ans]+/[as/] - multi-char prefix, multi-char name: [as][ans]+/[as][ans]+

so, substitute in the charset wherever you see a, n, s

i.e. the first rule is read as “a symbol can be a single-character name: [a-zA-Z\.\*+!\-_?$%&=<>\/]”

Because the set of symbols the reader can construct, and the set of all symbols is not the same, you cannot use the reader to determine what a valid symbol is

oh hmm

for (the set of all symbols) - (symbols the reader can construct), can you help me imagine what those are? or where they come from?

The symbol function will take arbitrary strings and make them symbols

For example the reader won't read 1 as a symbol, but (symbol "1") will construct such a symbol

oh, wow

user=> (symbol "///")
///
user=> (symbol "0")
0

Similar with keywords

user=> (def (symbol "1") "foo")
Syntax error compiling def at (REPL:1:1).
First argument to def must be a Symbol

I think there is some special handling of the / character when calling the symbol function for creating namespaced symbols

hmm, ok, so since def is a special form, the second arg is never evaluated.

You can use intern

so you can create invalid symbols at run-time, but you can’t actually def them to anything?

You could also programmatically construct a form containing the 1 symbol and pass it to eval, bypassing the reader

`(def ~(symbol "0") "aaaahhhh")
 (def 0 "aaaahhhh")

user=> (intern 'user 'x 1)
#'user/x
user=> x
1
user=> (intern 'user (symbol "0") "foo")
#'user/0
user=> 0
0
user=> user/0
Syntax error reading source at (REPL:7:0).
Invalid token: user/0

How would I actually refer to user/0?

You just can't use the reader

When you type anything at the reply it goes through the reader first, and the reader doesn't accept the 1 symbol

So interesting! Thanks for indulging me

if you didn't have anything else in that ns

(deref (second (first (ns-publics 'user))))
         "foo"

does anyone follow any naming convention for functions without "active" side effects, but potentially returning different values at different time (e.g. deref)?

for refs specifically I use * prefix: (let [*id (atom 0)] ...)

stolen from tonsky: https://tonsky.me/blog/readable-clojure/#use--as-prefix-for-references

that is name of a ref. I am curious about names of "getters" which are pure, but the thing they get from - is not.

> which are pure, but the thing they get from - is not does not look pure to me..

anyway, I don't have any naming convention for this 🙂

pure in a sense "does not modify things"

I feel like I'm missing something obvious, but has anyone set GOOGLE_APPLICATION_CREDENTIALS or any environment var for their Clojure code to use via (System/getenv) while using cider jack in?

@allandaviesza i always start a repl from cmdline and use cider-connect ... you could also use some configuration mechanism like juxt/aero

yea I've done that in the mean time

would be nice if I had a method to create an env var that required no work from other devs working on the project

like writing it in to profiles.clj (if it was to be committed)

@allandaviesza there is https://github.com/athos/lein-with-env-vars if it can help

@vlaaad looks like it's working! For some reason I glossed over this earlier when finding it via Google, many thanks for bringing it to my attention! 😄

is it (and if yes, is it practical, and if yes, how) possible to do timeouts when using pmap?

As a feature of pmap, no

pmap is really most ideal for chunky computational tasks

Then whats the best way to do some network calls and such in parallel without huge amounts of added complexity?

future may be of use

Something to that effect.

How would i implement something like pmap with Futures? The whole Futures Thing is somehow really confusing for me..

a call to future returns immediately then executes the body on another thread

so you'll set up a bunch of requests that will all start at the same time

then the second map will block execution while waiting for the first future to return

(I'm referring to the code in the snippet above)

the best thing to do is to try it out using sleep

So the first map creates a list of Futures (as i understand, a future is comparable to a promise in Javascript, correct?) And the second map then awaits their completion one after Another using deref. Wouldnt in your example the Futures get created only when attempting to deref them because of the lazyess of map?

future also returns what the body returns, the body being the expression within the call to future

yea thats a good point, so you can use mapv which is not lazy

you're correct about them being comparable to a promise

I think future is a lot more simple than a promise

I think i got it, thanks!

😄

Is it possible to put metadata on protocol methods?

@lkowarschick check out https://github.com/TheClimateCorporation/claypoole