我想实现一个函数,它是interleave
和interpose
的"混合",称为indent
。
的确,当你写
(interleave [1 2 3] ["_" "*"])
返回(1 "_" 2 "*")
但是我想要
(1 "_" 2 "*" 3)
所以我写了一个函数来实现这个目标:
(defn indent
"Mix of clojure.core interleave and interpose functions.
Indents the second collection in the first one."
[coll1 coll2]
(let [n1 (count coll1)
n2 (count coll2)
vcoll1 (vec coll1)
vcoll2 (vec coll2)
stop-idx (min (- n1 2) (- n2 1))]
(-> (loop [init '()
i 0]
(cond (> i stop-idx)
init
:else
(recur (concat init [(vcoll1 i) (vcoll2 i)]) (inc i))))
(concat [(vcoll1 (inc stop-idx))]))))
问题是表演太差了:
(time (dotimes [_ 10000000] (doall f [1 2 3] ["_" "*"])))
对于f = interleave: 2秒For f = indent: 7s
我试图模仿interleave
impl,但最后我有相同的高成本操作(计数和vec)。
我唯一能想到的快速计算是写Java代码…
你知道吗?谢谢!
编辑:更快的java方式
这不是Clojure的解决方案,但它减少了计算时间
package java_utils;
public class Clojure {
public static Object[] indent (Object[] coll1 , Object [] coll2) {
int len1 = coll1.length;
int len2 = coll2.length;
int stop_index = Math.min(len1 - 2, len2 - 1);
Object[] result = new Object[2*(stop_index+1) + 1];
for (int i = 0 ; i <= stop_index ; i++) {
result[2*i] = coll1[i];
result[2*i+1] = coll2[i];
}
result[2*stop_index+2] = coll1[stop_index+1];
return result;
}
}
(defn indent
[coll1 coll2]
(seq (Clojure/indent (to-array coll1) (to-array coll2))))
对于10M的迭代,它的计算速度很快,为1,7秒。
为什么不直接基于interleave
呢?这样的:
(defn indent2 [coll1 coll2]
(when (seq coll1)
(cons (first coll1)
(interleave coll2 (rest coll1)))))
它的性能应该和interleave差不多。
在Clojure中,这类问题通常最好通过创建seq来解决。通常,seqs需要较少的逻辑来实现,而且它们也很适合Clojure库。
(defn interdent
([seq1 seq2]
(if (seq seq1)
(lazy-seq
(cons (first seq1) (interdent seq2 (rest seq1)))))))
还有,我不太确定你的时间代码。不确定f在你的例子中是什么。用所讨论的函数替换f会产生错误。例如
user=> (time (dotimes [_ 10000000] (doall interpose [1 2 3] ["_" "*"])))
ArityException Wrong number of args (3) passed to: core/doall clojure.lang.AFn.throwArity (AFn.java:429)
我使用下面的代码片段对代码计时。
user=> (time (dotimes [_ 10000000] (doall (interpose [1 2 3] ["_" "*"]))))
。
user=> (interdent [1 2 3] ["_" "*"])
(1 "_" 2 "*" 3)
user=> (interdent [1 2 3] ["_" "*" ":("])
(1 "_" 2 "*" 3 ":(")
user=> (interdent [1 2 3] ["_" "*" ":)" ":("])
(1 "_" 2 "*" 3 ":)")