if you can clojure.edn/read it, then it's got correctly balanced brackets, and well formed symbols, etc ... is that good enough?

if i'm not mistaken, read both the one in clojure.edn and also clojure.core only read the first sexpression in the stream. right?

so I think I need to loop and read until EOF.

but I don't understand the difference between those to variants of read

edn/read is safe to run and will not modify your running process, while read can.

you can repeatedly read from a reader until the end of the file with something like

(import '[java.io PushbackReader])
(require '[clojure.java.io :as io] '[clojure.edn :as edn])
(with-open [rdr (PushbackReader. (io/reader (io/file (str system-file))))]
  (loop [els []]
    (let [el (edn/read {:eof ::eof}
                       rdr)]
      (if (= el ::eof)
        els
        (recur (conj els el))))))

Actually won't that fail if there is a line containing ::eof in the file?

I'm using something like this:

(defn line-read-file
  "Read all the lines from a file and return an vector of lines which have been read"
  [filename]
  (println [:open filename])
  (with-open [r (java.io.PushbackReader.
                 (clojure.java.io/reader filename))]
    (binding [*read-eval* false]
      (let [eof-value (list :end)]
        (loop [lines []
               line (read r false eof-value)]
          (if (identical? line eof-value)
            (printf "read %d lines from %s\n" (count lines) filename)
            (recur (conj lines line)
                   (read r false eof-value))))))))

::eof is namespaced to the code that you're reading the file in ... so it seems unlikely that it'll be in the code your reading?

it is impossible for read to return a list identical? to a list I've allocated at run-time.

the way read works in CL seems to be the same conceptually as in clojure. you give read a special value and it returns exactly that value if it encounters end of file, then you can test using pointer equality whether it returned exactly that value.

no

read is explicitly different between cl and clojure

I don't think it's easy to return a value that's identical? to something already in ram from read, unless it's a keyword

that is the point. read returns exactly its 4th argument when it reaches end of file, as long as the third argument is false. it seems to me the correct way to detect end of file is to test whether read returned that exact value as documented.

am I wrong?

ah yeah ... sorry - I misunderstood what you meant ... yes, when it gets to the end of the file thing you give it

anyway, i can replace clojure.core/read with clojure.end/read, no problem to avoid run-time evaluation

but I suspect *read-eval* controls the same thing.

BTW, I don't understand what `

(java.io.PushbackReader.
                 ( filename))

clojure.lang.Var.applyTo (Var.java:705)
    clojure.main.main (main.java:40)
Caused by: java.io.FileNotFoundException: binarysearch.clj (No such file or directory)
 at java.io.FileInputStream.open0 (FileInputStream.java:-2)
    java.io.FileInputStream.open (FileInputStream.java:219)

i'm calling the script using lein exec. is that changing my CWD?

edn/read supports a subset of clojure while *read-eval* disables some features of the clojure reader

I think you need (PushbackReader. (io/reader (io/file filename)))

io/reader assumes that a string is a url and tries to fetch it

yes, lein exec changes the directory. I print the value of (-> (.File. ".") .getAbsolutePath) inside the script, and it is different than the directory lein exec is called from.

oops ... tell a lie ... it does this

(try
                          (make-input-stream (URL. x) opts)
                          (catch MalformedURLException e
                            (make-input-stream (File. x) opts)))

I took the code from here https://clojuredocs.org/clojure.core/read

is that wrong?

https://clojuredocs.org/clojure.core/read#example-542692d5c026201cdc327056

nope ... it's me that's wrong - if you pass io/reader a string it will try and turn it into a url and if it's invalid it'll assume it's a file and try to read that - you problem is probably that you've got a relative path and you're in the wrong dir

my string is exactly the string "binarysearch.clj" and I've verified it is in the directory I started lein in. But within the program I print the directory and it is no longer the same directory.

perhaps lein tries to walk up the directory to the top of the project???

it is surprising, but i can call my script with full path names to avoid this annoyance.

quite possibly ... I've not used lein in ages and can't remember what it does

by "top of the project" do you mean "where the project.clj file is"?

if so I wouldn't be surprised if lein changed dir to there before doing anything else

seems it does that. changes directory to where it finds the project.clj file

I didn't bother reading the documentation of lein exec. it is probably documented behavior.

via interop:

(let [x (Object.)]
  (.equals x x))
;; => true

(.equals (Object.) (Object.))
;; => false

why not use identical?

ha) I never use it because clojure's equivalent semantic is good enough for all my cases. IMHO, equivalency check makes sense only if you do heavy interop to java. That is why I never bothered to look for identical? function.

as I understand it is the only way to dependably use the 4th argument of read

read returns its 4th argument exactly when trying to read past end of file. to detect this you must use identical? to distinguish the improbable case that it read something = to your 4th argument

at least that how it works in common lisp and it looks like clojure read was modeled off that.

(let [t (random-uuid)]
  (loop [acc [] rdr rdr]
    (let [x (read {:eof t} rdr)]
      (if (= x t)
        acc
        (recur (conj acc x) rdr)))))

hmm.

my sense was if I'm reading edn data I should use end/read and if i'm reading clojure code I should use normal read but just make sure *read-eval* is bound to false.

what is the reasoning I'm missing?

what is random-uuid ?

it seems like you're just making it very unlikely to read t from the file. it seems better to me to depend on the documented behavior of read and just detect whether it returned the think I said to return on eof.

no need to generate a randomized uuid

BTW, my original question was also intended to ask whether there is already a function which will slurp in the entire file, or whether the programmer is required to explicitly call read (or cousin) repeatedly

read this https://clojuredocs.org/clojure.core/read#example-542692d5c026201cdc327056

> what is random-uuid ? function to generate a random UUID. it doesn't matter what function to use here. In most place where I have to use clojure.edn/read I simply use ::eof as a token. And yes, it might not consume input completely when by some reason there is a single top-level token matching :some.namespace/eof .

> it seems better to me to depend on the documented behavior of read What makes you think I'm suggesting you something different?

because if you use any token for the eof marker, if read reads that object at the top level, read will return in and your equivalence check will match and the read loop will terminate before the file is completely exhausted. it seems to me the only way to detect eof is to detect whether read returned exactly and identically your given eof marker, and to distinguish that from the case when read returned something different but not identical to that eof marker.

wrt the clojure doc, I did read that. it advises that you should not read untrusted data with read. that's not my case. I'm reading data which I just created. However, if there's some other reason than trust, perhaps I didn't really understand what I read.

Then why you read data you just created?

You can use a freshly created Java Object instance as the sentinel value that read returns if it gets an error, and then use identical? between that freshly created object and the return value. I don't know any other shorter way to guarantee whether the return value is from the file vs. not.

I believe for a freshly created Object instance like that, that Clojure = will be handled by reference equality, too.

> Then why you read data you just created?

because I didn't create it in the same process. I created in this case from a python program.

:local/root must be either a path to jar or path to clojure project with manifest (e.g. deps.edn).

Ahh sh*t, so I must specify them one by one?

yes

Okay, thanks man!! 🙂

you can also write a small babashka script to prepopulate your project's deps.edn based on a content of some folder )

hehe ,good idea... I may give it a try, or just type them out 😉

This is a total hack but Java accepts limited form of wildcarding so it probably works to add “dir/*” to your :paths or :extra-paths. Haven’t tried it

Guys I am really struggeling. I have org.babashka/http-client {:mvn/version "0.4.15"} in deps.edn and in my (ns my-ns (:require [babashka.http-client :as http])) but I get :

; Syntax error compiling at (babashka/http_client/interceptors.clj:1:1).
; namespace 'babashka.http-client.internal.multipart' not found

What am I doing possibly wrong?

off-topic, please use threads for multiple messages. It is asked in channel's description

thanks for the reminder

@U010VP3UY9X I've deleted several of your top-level channel messages from this morning. In future, please stay on topic for the channel and use threads -- we have plenty of other channels where some of that would have been on-topic /admin

👋

i'm doing a last pass through the homework problems fixing all the docstrings I see as being problematic

or perhaps it does but is not documented as such.

something like the following ought to work

(every? == '(1 2 3) '(1.0 2.0 3.0))

it does not, you'll need to (every? mytest (map vector col1 col2 col3))

and what if

(every? == '(1 2 3) '(1.0 2.0 3.0 4.0))

float comparisons are dangerous things

are you asking how should it be implemented? I'd say do it like common lisp does it. but i understand that someone else might say not to use CL as the standard.

No, I'm asking about what you expect

in CL it is documented that if the 2nd list (or nth) list is longer than the first, the extra values are ignored. but if the first argument is longer than any subsequent, then a run-time error is signaled

isn't that what map does?

I guess I'd like it to be consistent with map

no, map takes from args until one of them is run out

@U0P0TMEFJ just an example. actually what I wanted was (every? (almost-equal 0.0001) '(1 2 3) '(1.0 2.000000000001 3.0) )

how does (every? (partial apply ==) (map vector [1 2 3 4 5] [1.0 2.0 3.0 4.0 5.0])) seem?

good to know about map.

but anyway, every? implies it should eagerly check everything. That is why I'm asking what do you expect in a situation when there are uneven length sequences

no, i'd expect it to work by the same traversal rules as map

@U0P0TMEFJ that's ugly, but not your fault. could also do some sort of take-until-false or take-until-true concoction with (map == list-1 list-2)

(defn my-every? [pred & args]
  (if (= 1 (count args))
    (every? pred (first args))
    (every? #(apply pred %) (apply map list args))))

(my-every? == '(1 2 3) '(1.0 2.0 3.0))
;; => true

(my-every? == '(1 2 3) '(1.0 2.0001 3.0))
;; => false

take-until-false - that's what every? does ... right?

there's some function like some which asks whether something in the list is false.

I always mix up the functions like some vs some? etc

(every? identity (map == list-1 list-2))

(into (sorted-set) (subseq #{,,,} <= 3))

or clojure.set/select

(type #{1 2 3})
;; => clojure.lang.PersistentHashSet

(type (sorted-set 1 2 3))
;; => clojure.lang.PersistentTreeSet

(type
 (clojure.set/select #(<= % 3) (sorted-set 1 2 3)))
;; => clojure.lang.PersistentTreeSet

@U04V4KLKC Did you mean to use subseq instead of subset in your code snippet above?

@U0CMVHBL2 yes, thanks

i think those proposals are linear whereas the ask is for a less-than-linear approach

It should be linear in the size of the subset, not the full set you are taking the subset of, but if that is what is desired, I am not aware of anything built into Clojure that would achieve faster-than-linear subset creation like that.

🤷 the question was about returned type not performance

(and to be precise, when I say "linear" here it is really more like O(N log N) than it is O(N))

clojure.data.avl has a sorted set with some nice properties

"An implementation of persistent sorted maps and sets based on AVL
  trees which can be used as drop-in replacements for Clojure's
  built-in sorted maps and sets based on red-black trees. Apart from
  the standard sorted collection API, the provided map and set types
  support the transients API and several additional logarithmic time
  operations: rank queries via clojure.core/nth (select element by
  rank) and clojure.data.avl/rank-of (discover rank of element),
  \"nearest key\" lookups via clojure.data.avl/nearest, splits by key
  and index via clojure.data.avl/split-key and
  clojure.data.avl/split-at, respectively, and subsets/submaps using
  clojure.data.avl/subrange."

❯ clj -Sdeps '{:deps {org.clojure/data.avl {:mvn/version "RELEASE"}}}'
Downloading: org/clojure/data.avl/maven-metadata.xml from central
Clojure 1.11.1
user=> (require '[clojure.data.avl :as avl])
nil
user=> (avl/subrange (avl/sorted-set 0 1 2 3 4 5) > 1)
#{2 3 4 5}

Thanks for all answers!

I'm trying to read scripts folder that are on same folder "infra" where is scripts.clj , so I can read each script. (relative to it)

*file* is bound at compile time when the compiler loads a file, outside of that context the value it has is undefined

the way this is kind of thing is usually done is using what are called "resources" which are things that are not class files that can be loaded from classloaders

http://clojure.java.io/resource is a convenient way to get a handle on a resource

most build tools(like lein) follow the convention of putting resources in a distinct folder usually "resources/"

Got it, thanks. I shall move the files to there right now

I wanted to leave inside infra, because is part of it haha. But i'll move them to there right now

you can leave them in there

it all gets mingled together at the end

got it, thanks!

a separate resources/ folder is something build tools do just so people can feel more organized

I have not used metamorphic, but if I understand correctly, Jackdaw might be something you could use https://github.com/FundingCircle/jackdaw

Thanks! I'll check this out