Added conversion functions; wrote documentation.

This commit is contained in:
Simon Brooke 2019-06-24 11:58:23 +01:00
parent 9c60d925d6
commit b3f6591e4b
4 changed files with 214 additions and 10 deletions

109
README.md
View file

@ -1,9 +1,13 @@
# sparse-array
A Clojure library designed to manipulate sparse *arrays* - multi-dimensional spaces accessed by indices, but containing arbitrary values rather than just numbers. For sparse spaces which contain numbers only, you're better to use a *sparce matrix* library, for example [clojure.core.matrix](https://mikera.github.io/core.matrix/).
A Clojure library designed to manipulate sparse *arrays* - multi-dimensional spaces accessed by indices, but containing arbitrary values rather than just numbers. For sparse spaces which contain numbers only, you're better to use a *sparse matrix* library, for example [clojure.core.matrix](https://mikera.github.io/core.matrix/).
Arbitrary numbers of dimensions are supported, up to limits imposed by the JVM stack.
## Conventions:
### Sparse arrays
For the purposes of this library, a sparse array shall be implemented as a map, such that all keys are non-negative members of the set of integers, except for the following keyword keys, all of which are expected to be present:
1. `:dimensions` The number of dimensions in this array, counting the present one (value expected to be a real number);
@ -30,9 +34,110 @@ Thus an array with a single value 'hello' at coordinates x = 3, y = 4, z = 5 wou
At the present stage of development, where the expectations of an operation are violated, `nil` is returned and no exception is thrown. However, it's probable that later there will be at least the option of thowing specific exceptions, as otherwise debugging could be tricky.
### Dense arrays
For the purposes of conversion, a **dense array** is assumed to be a vector; a two dimensional dense array a vector of vectors; a three dimensional dense array a vector of vectors of vectors, and so on. For any depth `N`, all vectors at depth `N` must have the same arity. If these conventions are not respected conversion may fail.
## Usage
FIXME
### make-sparse-array
`sparse-array.core/make-sparse-array ([& dimensions])`
Make a sparse array with these `dimensions`. Every member of `dimensions` must be a keyword; otherwise, `nil` will be returned.
e.g.
```clojure
(make-sparse-array :x :y :z)
=> {:dimensions 3, :coord :x, :content (:y :z)}
```
### sparse-array?
`sparse-array.core/sparse-array? ([x])`
`true` if `x` is a sparse array conforming to the conventions established by this library, else `false`.
### put
`sparse-array.core/put ([array value & coordinates])`
Return a sparse array like this `array` but with this `value` at these `coordinates`. Returns `nil` if any coordinate is invalid.
e.g.
```clojure
(put (put (make-sparse-array :x :y) "hello" 3 4) "goodbye" 4 3)
=> {:dimensions 2,
:coord :x,
:content (:y),
3 {:dimensions 1, :coord :y, :content :data, 4 "hello"},
4 {:dimensions 1, :coord :y, :content :data, 3 "goodbye"}}
```
### get
`sparse-array.core/get ([array & coordinates])`
Return the value in this sparse `array` at these `coordinates`.
### merge-sparse-arrays
`sparse-array.core/merge-sparse-arrays ([a1 a2])`
Return a sparse array taking values from sparse arrays `a1` and `a2`, but preferring values from `a2` where there is a conflict. `a1` and `a2` must have the **same** dimensions in the **same** order, or `nil` will be returned.
e.g.
```clojure
(merge-sparse-arrays
(put (make-sparse-array :x) "hello" 3)
(put (make-sparse-array :x) "goodbye" 4)))
=> {:dimensions 1, :coord :x, :content :data, 3 "hello", 4 "goodbye"}
```
### dense-to-sparse
`sparse-array.core/dense-to-sparse ([x] [x coordinates])`
Return a sparse array representing the content of the dense array `x`, assuming these `coordinates` if specified. *NOTE THAT* if insufficient values of `coordinates` are specified, the resulting sparse array will be malformed.
e.g.
```clojure
(dense-to-sparse [nil nil nil "hello" nil "goodbye"])
=> {:dimensions 1, :coord :i0, :content :data, 3 "hello", 5 "goodbye"}
```
### sparse-to-dense
`sparse-array.core/sparse-to-dense ([x] [x arity])`
Return a dense array representing the content of the sparse array `x`.
**NOTE THAT** this has the potential to consume very large amounts of memory.
e.g.
```clojure
(sparse-to-dense
(put
(put
(make-sparse-array :x :y)
"hello" 3 4)
"goodbye" 4 3))
=> [[nil nil nil nil nil]
[nil nil nil nil nil]
[nil nil nil nil nil]
[nil nil nil nil "hello"]
[nil nil nil "goodbye" nil]]
```
## License

View file

@ -1,5 +1,5 @@
(defproject sparse-array "0.1.0-SNAPSHOT"
:description "FIXME: write description"
:description "A Clojure library designed to manipulate sparse *arrays* - multi-dimensional spaces accessed by indices, but containing arbitrary values rather than just numbers. For sparse spaces which contain numbers only, you're better to use a *sparse matrix* library, for example [clojure.core.matrix](https://mikera.github.io/core.matrix/)."
:url "http://example.com/FIXME"
:license {:name "Eclipse Public License"
:url "http://www.eclipse.org/legal/epl-v10.html"}

View file

@ -63,7 +63,9 @@
(defn merge-sparse-arrays
"Return a sparse array taking values from sparse arrays `a1` and `a2`,
but preferring values from `a2` where there is a conflict."
but preferring values from `a2` where there is a conflict. `a1` and `a2`
must have the **same** dimensions in the **same** order, or `nil` will
be returned."
[a1 a2]
(cond
(nil? a1) a2
@ -86,16 +88,88 @@
(keys a1)
(keys a2))))))))
(defn dense-dimensions
"How many usable dimensions (represented as vectors) does the dense array
`x` have?"
[x]
(if
(vector? x)
(if
(every? vector? x)
(inc (apply min (map dense-dimensions x)))
1)
0))
(defn dense-to-sparse
"Return a sparse array representing the content of the dense array `x`,
assuming these `coordinates` if specified."
assuming these `coordinates` if specified. *NOTE THAT* if insufficient
values of `coordinates` are specified, the resulting sparse array will
be malformed."
([x]
:TODO)
(dense-to-sparse x (map #(keyword (str "i" %)) (range))))
([x coordinates]
:TODO))
(let
[dimensions (dense-dimensions x)]
(reduce
merge
(apply make-sparse-array (take dimensions coordinates))
(map
(fn [i v] (if (nil? v) nil (hash-map i v)))
(range)
(if
(> dimensions 1)
(map #(dense-to-sparse % (rest coordinates)) x)
x))))))
(defn arity
"Return the arity of the sparse array `x`."
[x]
(inc (apply max (filter integer? (keys x)))))
(defn child-arity
"Return the largest arity among the arities of the next dimension layer of
the sparse array `x`."
[x]
(apply
max
(cons
-1 ;; if no children are sparse arrays, we should return 0ß
(map
arity
(filter sparse-array? (vals x))))))
(defn sparse-to-dense
"Return a dense array representing the content of the sparse array `x`.
If this blows up out of memory, that's strictly your problem."
[x]
:TODO)
**NOTE THAT** this has the potential to consume very large amounts of memory." ([x]
(sparse-to-dense x (arity x)))
([x arity]
(if
(map? x)
(let [a (child-arity x)]
(apply
vector
(map
#(let [v (x %)]
(if
(= :data (:content x))
v
(sparse-to-dense v a)))
(range arity))))
(apply vector (repeat arity nil)))))
(sparse-to-dense (put (make-sparse-array :x) "hello" 4))
(def x
(put
(put
(make-sparse-array :x :y)
"hello" 3 4)
"goodbye" 4 3))
(child-arity x)
(sparse-to-dense (x 1) 4)
(sparse-to-dense x)

View file

@ -65,3 +65,28 @@
sparse (dense-to-sparse dense)]
(is (= "hello" (get sparse 3)))
(is (= "goodbye" (get sparse 5))))))
(deftest dense-dimensions-tests
(testing "dense-dimensions"
(is (= 0 (dense-dimensions 1)))
(is (= 1 (dense-dimensions [1 2 3])))
(is (= 2 (dense-dimensions [[1 2 3][2 4 6][3 6 9]])))
(is (= 3
(dense-dimensions
[[[1 2 3][2 4 6][3 6 9]]
[[2 4 6][4 16 36][6 12 18]]
[[3 6 9][18 96 (* 6 36 3)][100 200 300]]])))))
(deftest sparse-to-dense-tests
(testing "Conversion from sparse to dense arrays"
(let [expected [[nil nil nil nil nil]
[nil nil nil nil nil]
[nil nil nil nil nil]
[nil nil nil nil "hello"]
[nil nil nil "goodbye" nil]]
actual (sparse-to-dense (put
(put
(make-sparse-array :x :y)
"hello" 3 4)
"goodbye" 4 3))]
(is (= actual expected)))))