我想映射大量的元组。我的地图看起来像:
{[1 2] :thing}
除了可能有几百万。我有一种感觉,树图可能是一个好东西,所以我试图让它工作。我似乎无法正确获得比较功能。
(defn compare
[[x y] [xx yy]]
(cond
(and (= x xx) (= y yy)) 0
(and (<= x xx) (<= y yy)) -1
(and (<= x xx) (> y yy)) -1
(and (> x xx) (<= y yy)) 1
(and (> x xx) (> y yy)) 1))
一些微不足道的输入似乎有效
user=> (compare [1 1] [1 1])
0
user=> (compare [1 1] [2 2])
-1
user=> (compare [1 2] [2 1])
-1
user=> (compare [2 1] [1 2])
1
但是如果我创建涵盖所有组合的输入,那么地图应该认为它们都是不同的。
(def inputs
"All tuples of [0-4, 5-10]."
(clojure.math.combinatorics/cartesian-product
(range 0 4)
(range 5 10)))
(def input-pairs
"All possible pairs of tuples"
(clojure.math.combinatorics/cartesian-product inputs inputs))
如果我测试比较函数,当两个向量在结构上相同时,它仅返回零 。
user=> (doseq [[a b] input-pairs]
#_=> (when (zero? (compare a b)) (prn a b)))
(0 5) (0 5)
(0 6) (0 6)
(0 7) (0 7)
(0 8) (0 8)
(0 9) (0 9)
(1 5) (1 5)
etc
所以我认为我的比较功能是正确的。但是,在树形图中使用它会产生一些奇怪的结果:
(def inputs-kvs
"Inputs in the format that the hash-map and sorted-map constructor understand"
(mapcat #(vector % (apply str %))
(clojure.math.combinatorics/cartesian-product
(range 0 4)
(range 5 10))))
将这些放在散列图中会给出正确的答案
(count (apply assoc (hash-map) inputs-kvs))
=> 20
但是使用给定的比较将它们放在树形图中:
(def structure (sorted-map-by compare))
(count (apply assoc structure inputs-kvs))
=> 4
(apply assoc structure inputs-kvs)
=> {(0 5) "25", (1 6) "36", (2 7) "37", (3 5) "39"}
“25”已存储在(0 5)
广告位中。但比较函数并未说明(0 5)
和(2 5)
是相同的:
=> (compare [0 5] [2 5])
-1
我做错了什么?我可以做这个吗?将二维空间投影到一维空间上是否有意义?
(关于你可能有的问题,是的,我尝试过二维结构,例如(sorted-map 1 (sorted-map 2 :value))
,但我正在努力寻找性能更佳的替代方案)
答案 0 :(得分:2)
Clojure已经附带了它自己的compare
:
user=> (doc compare)
-------------------------
clojure.core/compare
([x y])
Comparator. Returns a negative number, zero, or a positive number
when x is logically 'less than', 'equal to', or 'greater than'
y. Same as Java x.compareTo(y) except it also works for nil, and
compares numbers and collections in a type-independent manner. x
must implement Comparable
其行为与OP自身的功能相同,但最有可能更有效:
user=> (compare [1 1] [1 1])
0
user=> (compare [1 1] [2 2])
-1
user=> (compare [2 1] [1 2])
1
行为记录在Section about Vectors (IPersistentVector) in the Data Structures docs:
中首先按照长度比较矢量,然后按顺序比较每个元素。
因此,您可以使用核心中的sorted-map-by compare
,或者因为您的数据结构只是sorted-map
的默认值:
user=> (def m (into {} (let [r #(- (rand-int 10) (rand-int 10))] (for [a (range -1 2) b (range -1 2)] [[(r) (r)] (str a b)]))))
#'user/m
user=> (>pprint m)
{[-7 -4] "10",
[-3 5] "01",
[-5 -7] "00",
[5 2] "11",
[-3 1] "-10",
[7 -4] "-11",
[0 -6] "0-1",
[3 1] "-1-1",
[-8 -1] "1-1"}
nil
user=> (>pprint (into (sorted-map-by compare) m))
{[-8 -1] "1-1",
[-7 -4] "10",
[-5 -7] "00",
[-3 1] "-10",
[-3 5] "01",
[0 -6] "0-1",
[3 1] "-1-1",
[5 2] "11",
[7 -4] "-11"}
nil
user=> (>pprint (into (sorted-map) m))
{[-8 -1] "1-1",
[-7 -4] "10",
[-5 -7] "00",
[-3 1] "-10",
[-3 5] "01",
[0 -6] "0-1",
[3 1] "-1-1",
[5 2] "11",
[7 -4] "-11"}
nil
user=> (assert (= (into (sorted-map-by compare) m) (into (sorted-map) m)))
nil
答案 1 :(得分:1)
我刚刚添加(vec %)
来保留元组向量 - 不应该改变任何东西。
你可以看到它在这里工作。
可能是你有一些旧的REPL东西 - 尤其是你别名clojure.core/compare
?
; using your compare function
(def inp (mapcat #(vector (vec %) (apply str %))
(clojure.math.combinatorics/cartesian-product (range 0 4) (range 5 10))))
; => ([0 5] "05" [0 6] "06" [0 7] "07" [0 8] "08" ...
(count inp)
; => 40
(apply assoc structure inp)
; => {[0 9] "09", [0 8] "08", [0 7] "07", [0 6] "06", ....
(count (apply assoc structure inp))
; => 20