Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

ClassCastException: clojure.lang.PersistentVector$Node cannot be cast to [I

Description

I'm seeing some strange behaviour which I've been unable to find a small failing case for.

(defn quicksort [v] (if (<= (count v) 1) v (let [[x & xs] v] (rrb/catvec (quicksort (filterv #(<= % x) xs)) [x] (quicksort (filterv #(> % x) xs))))))

We can check this implementation with test.check

(defn ascending? [coll] (every? (fn [[a b]] (<= a b)) (partition 2 1 coll))) (def property (prop/for-all [v (gen/vector gen/int)] (let [s (quicksort v)] (and (= (count v) (count s)) (ascending? s))))) (tc/quick-check 10000 property) ;; => {:result true, :num-tests 10000, :seed 1440948212354}

(I've bumped the number of checks to 1e6 but test.check was not able to find a failing case)

Still, for some reason I get a "ClassCastException: clojure.lang.PersistentVector$Node cannot be cast to [I" for the attached 1193 element vector.

code
(quicksort (read-string (slurp "failing-1193.edn")))
code

Environment

core.rrb-vector version 0.0.11 and clojure version 1.7.0

Attachments

2
  • 13 Sep 2015, 07:28 AM
  • 30 Aug 2015, 04:34 PM

Activity

Show:

Andy Fingerhut September 30, 2019 at 11:23 PM

Resolving this issue as fixed, given the test results described in previous comments.

Andy Fingerhut September 19, 2019 at 4:55 PM

With the latest master version of core.rrb-vector as of today, I have run (tc/quick-check 1e6 property) 3 times with no errors or exceptions, and (quicksort (shuffle (range 1e6))) 20 times with no errors, and the return value always returns true for ascending?.

I do not know whether all of the bugs are fixed in this version, but at least these tests are now much less likely to find any remaining issues, it seems. Please try it out if you are interested, and report what you find.

Andy Fingerhut September 19, 2019 at 4:44 AM

With this commit: https://github.com/clojure/core.rrb-vector/commit/b5481a5709743e15a72fbb09cece178c8fc81d4d

I have added the attached failing test case to the automated tests of core.rrb-vector. I have recently fixed several other failing test cases, and one or more of them seems to have caused this failing test case to pass.

I will wait the hour or two for the tests to run on build.clojure.org, and in the mean time perhaps try to run the test.check version of the test reported in this issue to see if any of them fail.

Jonas Enlund September 13, 2015 at 7:28 AM

I added failing tests for the ClassCastException and the NullPointerException. For some reason, I'm only able to reproduce the ArrayIndexOutOfBoundsException at the repl

Michał Marczyk September 11, 2015 at 7:43 PM

Thanks again for the bug report and many thanks for the follow-ups! Would you be interested in wrapping any of these test cases in a patch?

As for a fix – I can see roughly what is happening with the original problem case, what remains is to track down exactly where the miscalculation happens… I'll cut a new release as soon as that's sorted.

Fixed

Details

Assignee

Reporter

Priority

Created August 30, 2015 at 4:34 PM
Updated September 30, 2019 at 11:23 PM
Resolved September 30, 2019 at 11:23 PM

Flag notifications