Given a list like:
(quote (do (println "Hello") (println "Goodbye")))
: how can I turn this into a nicely formatted string which will output something like:
(do
(println "Hello")
(println "Goodbye")
)
Use clojure.pprint/write with clojure.pprint/code-dispatch:
(clojure.pprint/write '(do (println "Hello") (println "Goodbye"))
:dispatch clojure.pprint/code-dispatch)
For that small amount of code you won't get new-lines. But a call like this
(clojure.pprint/write '(do (println "Hello") (println "Goodbye") (println "Hey, you left me out!"))
:dispatch clojure.pprint/code-dispatch)
Returns
=> (do
(println "Hello")
(println "Goodbye")
(println "Hey, you left me out!"))
To catch that as a string wrap the call in (with-out-str ...)
As a significantly more performant alternative to clojure.pprint, you might want to try Brandon Bloom's fipp:
;; add [fipp "0.4.0"] to :dependencies
(require '[fipp.edn :as fipp])
(fipp/pprint '(do (println :foo) (println :bar) (println :quux))
{:width 30}) ; force wrapping despite small size of data
;; printed output:
(do
(println :foo)
(println :bar)
(println :quux))
Unfortunately, Fipp does not yet support code-dispatch.
Related
I'm having a hard time getting the multimethods in Clojure to work as I would expect. A distillation of my code is as follows.
(defn commandType [_ command] (:command-type command))
(defmulti testMulti commandType)
(defmethod testMulti :one [game command] (str "blah"))
(defmethod testMulti :default [& args] "Cannot understand")
(testMulti "something" {:command-type :one})
(commandType "something" {:command-type :one})
Now I would expect here to have the method commandType called on the arguments which would of course return :one which should send it to the first defmethod but instead I get a null pointer exception. Even the simplest invocation of a multimethod I could come up with gives me a null pointer:
(defmulti simpleMulti :key)
(defmethod simpleMulti "basic" [params] "basic value")
(simpleMulti {:key "basic"})
And yet the example in the clojure docs located here works fine. Is there something fundamental I'm doing wrong?
So far as I can see, it works.
Given
(defmulti testMulti (fn [_ command] (:command-type command)))
(defmethod testMulti :one [game command] (str "blah"))
(defmethod testMulti :default [& args] "Cannot understand")
then
(testMulti "something" {:command-type :one})
; "blah"
(testMulti "something" {:command-type :two})
; "Cannot understand"
(testMulti "something" 5)
; "Cannot understand"
as expected.
I reset the REPL before running the above afresh.
And the simple example works too. Given
(defmulti simpleMulti :key)
(defmethod simpleMulti "basic" [params] "basic value")
then
(simpleMulti {:key "basic"})
; "basic value"
I have a situation where I am creating and destroying objects in one clojure namespace, and want another namespace to co-ordinate. However I do not want the first namespace to have to call the second explicitly on object destruction.
In Java, I could use a listener. Unfortunately the underlying java libraries do not signal events on object destruction. If I were in Emacs-Lisp, then I'd use hooks which do the trick.
Now, in clojure I am not so sure. I have found the Robert Hooke library https://github.com/technomancy/robert-hooke. But this is more like defadvice in elisp terms -- I am composing functions. More over the documentation says:
"Hooks are meant to extend functions you don't control; if you own the target function there are obviously better ways to change its behaviour."
Sadly, I am not finding it so obvious.
Another possibility would be to use add-watch, but this is marked as alpha.
Am I missing another obvious solution?
Example Added:
So First namespace....
(ns scratch-clj.first
(:require [scratch-clj.another]))
(def listf (ref ()))
(defn add-object []
(dosync
(ref-set listf (conj
#listf (Object.))))
(println listf))
(defn remove-object []
(scratch-clj.another/do-something-useful (first #listf))
(dosync
(ref-set listf (rest #listf)))
(println listf))
(add-object)
(remove-object)
Second namespace
(ns scratch-clj.another)
(defn do-something-useful [object]
(println "object removed is:" object))
The problem here is that scratch-clj.first has to require another and explicitly push removal events across. This is a bit clunky, but also doesn't work if I had "yet-another" namespace, which also wanted to listen.
Hence I thought of hooking the first function.
Is this solution suitable to your requirements?
scratch-clj.first:
(ns scratch-clj.first)
(def listf (atom []))
(def destroy-listeners (atom []))
(def add-listeners (atom []))
(defn add-destroy-listener [f]
(swap! destroy-listeners conj f))
(defn add-add-listener [f]
(swap! add-listeners conj f))
(defn add-object []
(let [o (Object.)]
(doseq [f #add-listeners] (f o))
(swap! listf conj o)
(println #listf)))
(defn remove-object []
(doseq [f #destroy-listeners] (f (first #listf)))
(swap! listf rest)
(println #listf))
Some listeners:
(ns scratch-clj.another
(:require [scratch-clj.first :as fst]))
(defn do-something-useful-on-remove [object]
(println "object removed is:" object))
(defn do-something-useful-on-add [object]
(println "object added is:" object))
Init binds:
(ns scratch-clj.testit
(require [scratch-clj.another :as another]
[scratch-clj.first :as fst]))
(defn add-listeners []
(fst/add-destroy-listener another/do-something-useful-on-remove)
(fst/add-add-listener another/do-something-useful-on-add))
(defn test-it []
(add-listeners)
(fst/add-object)
(fst/remove-object))
test:
(test-it)
=> object added is: #<Object java.lang.Object#c7aaef>
[#<Object java.lang.Object#c7aaef>]
object removed is: #<Object java.lang.Object#c7aaef>
()
It sounds a lot like what you're describing is callbacks.
Something like:
(defn make-object
[destructor-fn]
{:destructor destructor-fn :other-data "data"})
(defn destroy-object
[obj]
((:destructor obj) obj))
; somewhere at the calling code...
user> (defn my-callback [o] (pr [:destroying o]))
#'user/my-callback
user> (destroy-object (make-object my-callback))
[:destroying {:destructor #<user$my_callback user$my_callback#73b8cdd5>, :other-data "data"}]
nil
user>
So, here is my final solution following mobytes suggestion. A bit more work, but
I suspect that I will want this in future.
Thanks for all the help
;; hook system
(defn make-hook []
(atom []))
(defn add-hook [hook func]
(do
(when-not
(some #{func} #hook)
(swap! hook conj func))
#hook))
(defn remove-hook [hook func]
(swap! hook
(partial
remove #{func})))
(defn clear-hook [hook]
(reset! hook []))
(defn run-hook
([hook]
(doseq [func #hook] (func)))
([hook & rest]
(doseq [func #hook] (apply func rest))))
(defn phils-hook []
(println "Phils hook"))
(defn phils-hook2 []
(println "Phils hook2"))
(def test-hook (make-hook))
(add-hook test-hook phils-hook)
(add-hook test-hook phils-hook2)
(run-hook test-hook)
(remove-hook test-hook phils-hook)
(run-hook test-hook)
So, in my core.clj file I have:
(def page-buffer (BufferedReader. (InputStreamReader. (clojure.java.io/input-stream (clojure.java.io/resource "mitochondria.html")))))
(def parsed-page (atom ""))
and then later:
(defn -main [& args]
(let [port (Integer/parseInt (first args))]
(swap! parsed-page (with-open []
(.toString (reduce #(.append %1 %2)
(StringBuffer.) (line-seq page-buffer)))))
(println "Server is starting")
(println "port: " port)
(run-server port)))
This compiles and then I turn it into an uberjar. But when I run it I get the line with swap! blows up:
Exception in thread "main" java.lang.ClassCastException: java.lang.String cannot be cast to clojure.lang.IFn
at clojure.core$swap_BANG_.invoke(core.clj:2106)
at serve_pages_from_memory.core$_main.doInvoke(core.clj:29)
at clojure.lang.RestFn.applyTo(RestFn.java:137)
at serve_pages_from_memory.core.main(Unknown Source)
I tried this without using an atom and everything worked just fine (using a var defined with "def" to be a string) but eventually I want to send this var to multiple threads, so I need for this to be an atom or agent.
What have I done wrong?
UPDATE:
Jeremy Heiler, thanks. Fixed, but the result is ugly as sin:
(defn parse-buffer [& everything-else]
(with-open []
(.toString (reduce #(.append %1 %2)
(StringBuffer.) (line-seq page-buffer)))))
(defn -main [& args]
(let [port (Integer/parseInt (first args))]
(swap! parsed-page parse-buffer)
(println "Server is starting")
(println "port: " port)
(run-server port)))
I have to give an argument to parse-buffer or I get an error. But I do not use the argument, so this is ugly. I must be writing this wrong, yes?
You need to pass a function to swap!. Right now you are passing it a value.
(let [foo (atom 1)]
(swap! foo + 2)
#foo)
The above expression will return 3. The passed in function takes the current value of the atom, and the return value becomes the new value of the atom. Any extra arguments are passed to the given function.
To comment on your update: Yes, you do need to give an argument to the function used in swap!. The first argument is the current value of the the atom. However, you don't need to hack parse-buffer in order to make it work. You can just wrap it with an anonymous function.
Also, parse-buffer can be simplified greatly. Since line-seq returns a sequence of lines from the given reader, and str uses a StringBuilder internally, you can just apply it over the sequence.
(defn parse-buffer []
(with-open [buf page-buffer]
(apply str (line-seq buf))))
(defn -main [& args]
(let [port (Integer/parseInt (first args))]
(swap! parsed-page (fn [cur-val] (parse-buffer)))
(println "Server is starting")
(println "port: " port)
(run-server port)))
I'm trying to understand the -main and its namespace settings. The default namespace in a -main function seems to be "user", but function defined "above" the main function, in the same file, can be called. Are they referred? I was running this to find out:
(ns ack.doodle)
(defn fo [] "bar")
(defn -main [& args]
(println (fo)) ;; bar
(println *ns*) ;; #<Namespace user>
(println (get (ns-refers *ns*) 'sorted-map)) ;; #'clojure.core/sorted-map
(println (get (ns-refers *ns*) 'fo)) ;; nil
(println (get (ns-map *ns*) 'fo)) ;; nil
(println (get (ns-publics *ns*) 'fo)) ;; nil
(println (get (ns-interns *ns*) 'fo)) ;; nil
(println (get (ns-aliases *ns*) 'fo)) ;; nil
(println (ns-resolve *ns* 'fo))) ;; nil
The call (fo) succeeds, yet apparently the symbol 'fo is not known in the current namespace *ns*. What's going on?
This problem hit me when trying to pass the name of some function, along with some arguments, as command line arguments to -main.
steps to reproduce
paste above code in a file ./src/ack/doodle.clj
create ./project.clj with these contents:
(defproject ack "1" :main ack.doodle)
run it:
lein run
It looks like leiningen runs (-main) by calling it from user namespace like that: (ack.doodle/-main). Thus *ns* is bound to the user namespace.
Try running lein repl in your project root. Then run (-main) and see what happens. :)
Didn't I answer this for you last night in #clojure? If there's something unsatisfying about the answer, you'll have to clarify your question.
I want to send var-args of a function to a macro, still as var-args.
Here is my code:
(defmacro test-macro
[& args]
`(println (str "count=" ~(count args) "; args=" ~#args)))
(defn test-fn-calling-macro
[& args]
(test-macro args))
The output of (test-macro "a" "b" "c") is what I want: count=3; args=abc
The output of (test-fn-calling-macro "a" "b" "c") is : count=1; args=("a" "b" "c") because args is sent as a single argument to the macro. How can I expand this args in my function in order to call the macro with the 3 arguments?
I guess I'm just missing a simple core function but I'm not able to find it. Thanks
EDIT 2 - My "real" code, shown in EDIT section below is not a valid situation to use this technique.
As pointed out by #Brian, the macro xml-to-cass can be replaced with a function like this:
(defn xml-to-cass
[zipper table key attr & path]
(doseq [v (apply zf/xml-> zipper path)] (cass/set-attr! table key attr v)))
EDIT - the following section goes beyond my original question but any insight is welcome
The code above is just the most simple I could come with to pinpoint my problem. My real code deals with clj-cassandra and zip-filter. It may also look over-engineering but it's just a toy project and I'm trying to learn the language at the same time.
I want to parse some XML found on mlb.com and insert values found into a cassandra database. Here is my code and the thinking behind it.
Step 1 - Function which works fine but contains code duplication
(ns stats.importer
(:require
[clojure.xml :as xml]
[clojure.zip :as zip]
[clojure.contrib.zip-filter.xml :as zf]
[cassandra.client :as cass]))
(def root-url "http://gd2.mlb.com/components/game/mlb/year_2010/month_05/day_01/")
(def games-table (cass/mk-cf-spec "localhost" 9160 "mlb-stats" "games"))
(defn import-game-xml-1
"Import the content of xml into cassandra"
[game-dir]
(let [url (str root-url game-dir "game.xml")
zipper (zip/xml-zip (xml/parse url))
game-id (.substring game-dir 4 (- (.length game-dir) 1))]
(doseq [v (zf/xml-> zipper (zf/attr :type))] (cass/set-attr! games-table game-id :type v))
(doseq [v (zf/xml-> zipper (zf/attr :local_game_time))] (cass/set-attr! games-table game-id :local_game_time v))
(doseq [v (zf/xml-> zipper :team [(zf/attr= :type "home")] (zf/attr :name_full))] (cass/set-attr! games-table game-id :home_team v))))
The parameter to import-game-xml-1 can be for example "gid_2010_05_01_colmlb_sfnmlb_1/". I remove the "gid_" and the trailing slash to make it the key of the ColumnFamily games in my database.
I found that the 3 doseq were a lot of duplication (and there should be more than 3 in the final version). So code templating using a macro seemed appropriate here (correct me if I'm wrong).
Step 2 - Introducing a macro for code templating (still works)
(defmacro xml-to-cass
[zipper table key attr & path]
`(doseq [v# (zf/xml-> ~zipper ~#path)] (cass/set-attr! ~table ~key ~attr v#)))
(defn import-game-xml-2
"Import the content of xml into cassandra"
[game-dir]
(let [url (str root-url game-dir "game.xml")
zipper (zip/xml-zip (xml/parse url))
game-id (.substring game-dir 4 (- (.length game-dir) 1))]
(xml-to-cass zipper games-table game-id :type (zf/attr :type))
(xml-to-cass zipper games-table game-id :local_game_time (zf/attr :local_game_time))
(xml-to-cass zipper games-table game-id :home_team :team [(zf/attr= :type "home")] (zf/attr :name_full))))
I believe that's an improvement but I still see some duplication in always reusing the same 3 parameters in my calls to xml-to-cass. That's were I introduced an intermediate function to take care of those.
Step 3 - Adding a function to call the macro (the problem is here)
(defn import-game-xml-3
"Import the content of xml into cassandra"
[game-dir]
(let [url (str root-url game-dir "game.xml")
zipper (zip/xml-zip (xml/parse url))
game-id (.substring game-dir 4 (- (.length game-dir) 1))
save-game-attr (fn[key path] (xml-to-cass zipper games-table game-id key path))]
(save-game-attr :type (zf/attr :type)) ; works well because path has only one element
(save-game-attr :local_game_time (zf/attr :local_game_time))
(save-game-attr :home :team [(zf/attr= :type "home"] (zf/attr :name_full))))) ; FIXME this final line doesn't work
Here's a some simple code which may be illuminating.
Macros are about code generation. If you want that to happen at runtime, for some reason, then you have to build and evaluate the code at runtime. This can be a powerful technique.
(defmacro test-macro
[& args]
`(println (str "count=" ~(count args) "; args=" ~#args)))
(defn test-fn-calling-macro
[& args]
(test-macro args))
(defn test-fn-expanding-macro-at-runtime
[& args]
(eval (cons `test-macro args)))
(defmacro test-macro-expanding-macro-at-compile-time
[& args]
(cons `test-macro args))
;; using the splicing notation
(defmacro test-macro-expanding-macro-at-compile-time-2
[& args]
`(test-macro ~#args))
(defn test-fn-expanding-macro-at-runtime-2
[& args]
(eval `(test-macro ~#args)))
(test-macro "a" "b" "c") ;; count=3; args=abc nil
(test-fn-calling-macro "a" "b" "c") ;; count=1; args=("a" "b" "c") nil
(test-fn-expanding-macro-at-runtime "a" "b" "c") ; count=3; args=abc nil
(test-macro-expanding-macro-at-compile-time "a" "b" "c") ; count=3; args=abc nil
(test-macro-expanding-macro-at-compile-time-2 "a" "b" "c") ; count=3; args=abc nil
(test-fn-expanding-macro-at-runtime "a" "b" "c") ; count=3; args=abc nil
If contemplation of the above doesn't prove enlightening, might I suggest a couple of my own blog articles?
In this one I go through macros from scratch, and how clojure's work in particular:
http://www.learningclojure.com/2010/09/clojure-macro-tutorial-part-i-getting.html
And in this one I show why run-time code generation might be useful:
http://www.learningclojure.com/2010/09/clojure-faster-than-machine-code.html
The typical way to use a collection as individual arguments to a function is to use (apply function my-list-o-args)
(defn test-not-a-macro [& args]
(print args))
(defn calls-the-not-a-macro [& args]
(apply test-not-a-macro args))
though you wont be able to use apply because test-macro is a macro. to solve this problem you will need to wrap test macro in a function call so you can apply on it.
(defmacro test-macro [& args]
`(println ~#args))
(defn calls-test-macro [& args]
(eval (concat '(test-macro) (args)))) ;you almost never need eval.
(defn calls-calls-test-macro [& args]
(calls-test-macro args))
This is actually a really good example of one of the ways macros are hard to compose. (some would say they cant be composed cleanly, though i think thats an exageration)
Macros are not magic. They are a mechanism to convert code at compile-time to equivalent code; they are not used at run-time. The pain you are feeling is because you are trying to do something you should not be trying to do.
I don't know the library in question, but if cass/set-attr! is a function, I see no reason why the macro you defined has to be a macro; it could be a function instead. You can do what you want to do if you can rewrite your macro as a function instead.
Your requirements aren't clear. I don't see why a macro is necessary here for test-macro, unless you're trying to print the unevaluated forms supplied to your macro.
These functions provide your expected results, but that's because your sample data was self-evaluating.
(defn test-args
[& args]
(println (format "count=%d; args=%s"
(count args)
(apply str args))))
or
(defn test-args
[& args]
(print (format "count=%d; args=" (count args)))
(doseq [a args]
(pr a))
(newline))
You can imagine other variations to get to the same result.
Try calling that function with something that doesn't evaluate to itself, and note the result:
(test-args (+ 1 2) (+ 3 4))
Were you looking to see the arguments printed as "37" or "(+ 1 2)(+ 3 4)"?
If you were instead trying to learn about macros and their expansion in general, as opposed to solving this particular problem, please tune your question to probe further.