http://dev.clojure.org/jira This file is an XML representation of an issue en-us 4.4 649 25-07-2011 http://dev.clojure.org/jira/browse/CLJ-1052 Clojure <p><a href="https://groups.google.com/forum/?fromgroups=#!topic/clojure/k2R4OdPUCzg">https://groups.google.com/forum/?fromgroups=#!topic/clojure/k2R4OdPUCzg</a></p> <p>Suggested by Ambrose Bonnaire-Sergeant:</p> <p>I think assoc should throw an error when applied with uneven arguments.</p> <p>Currently, the "missing" value is just replaced with nil.</p> <p>(assoc {} :a 1 :b)<br/> ;=&gt; {:a 1, :b nil}</p> CLJ-1052

assoc should throw exception if missing val for last key

Defect Minor Closed Completed Unassigned Andy Fingerhut Wed, 29 Aug 2012 16:40:24 -0500 Sat, 20 Oct 2012 09:53:42 -0500 Sat, 20 Oct 2012 09:53:42 -0500 Release 1.2 Release 1.3 Release 1.4 Release 1.5 8 4 <p>Patch clj-1052-assoc-should-throw-exc-if-missing-val-patch1.txt dated Aug 29 2012 makes assoc throw an IllegalArgumentException if more than one key is given, but the last key has no value. It includes a few simple test cases with correct and incorrect arguments to assoc.</p> <p>IMO if the error message included something like "assoc expects even number of arguments after target, found odd number", or some mention of the number of arguments the error would be clearer to me.</p> <p>clj-1052-assoc-should-throw-exc-if-missing-val-patch2.txt is identical to the now-removed -patch1.txt, except for the text of the exception thrown, updated as per Ambrose's suggestion.</p> <p>The patch appears correct. It does introduce a single extra (next) per iteration into the success path, although that seems unlikely to dominate the work. Wouldn't hurt to add as assessment showing this is no slower for correct programs.</p> <p>Test platform: Mac OS X 10.6.8 + Oracle/Apple Java 1.6.0_35 Java HotSpot(TM) 64-Bit Server VM</p> <p>With latest Clojure master as of Sep 19 2012:</p> <p>Clojure 1.5.0-master-SNAPSHOT<br/> user=&gt; (set! <b>unchecked-math</b> true)<br/> true<br/> (defn count-maps <span class="error">&#91;n&#93;</span><br/> (let <span class="error">&#91;base {:a 1}&#93;</span><br/> (loop [i n<br/> sum 0]<br/> (if (zero? i)<br/> sum<br/> (let <span class="error">&#91;m1 (assoc base :b 2 :c 3 :d 4 :e 5 :f 6 :g 7 :h 8 :i 9 :j 10)&#93;</span><br/> (recur (dec i) (+ sum (count m1))))))))<br/> #'user/count-maps<br/> user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 48784.077 msecs"<br/> 100000000<br/> user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 49028.52 msecs"<br/> 100000000<br/> user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 50314.729 msecs"<br/> 100000000</p> <p>Same Clojure version, plus the patch that was screened:</p> <p>user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 49576.191 msecs"<br/> 100000000<br/> user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 49957.866 msecs"<br/> 100000000<br/> user=&gt; (time (count-maps 10000000))<br/> "Elapsed time: 52149.998 msecs"<br/> 100000000</p> <p>(average of 3 times after patch) / (average of 3 times before patch) = 1.0240</p> <p>So 2.4% slowdown on average for that test case. I should add that I'm not a statistician, but note that this 2.4% difference is less than the variation in run time from one run to the next of the same experiment. Likely any real statistician would recommend collecting a lot more data before asserting there is a change in performance.</p> <p>clj-1052-assoc-should-throw-exc-if-missing-val-patch3.txt dated Oct 5 2012 is the same as the previous patch clj-1052-assoc-should-throw-exc-if-missing-val-patch2.txt dated Aug 29 2012 except some context lines have been updated so that it applies cleanly using git. The older version will be removed in a minute.</p> Approval Ok Global Rank Patch Code and Test