Almost working quickselect that will pave the way for median and

order statistics.

examples/select.j

@@ -0,0 +1,33 @@
+# http://en.wikipedia.org/wiki/Selection_algorithm#Partition-based_general_selection_algorithm
+
+function _jl_quickselect(a::AbstractVector, k::Integer, lo::Integer, hi::Integer)
+    while true
+        #partition(a, lo, hi, m)
+        i, j = lo, hi
+        pivot = a[div(lo+hi, 2)]
+        while i <= j
+            while a[i] < pivot; i += 1; end
+            while pivot < a[j]; j -= 1; end
+            if i <= j
+                a[i], a[j] = a[j], a[i]
+                i += 1
+                j -= 1
+            end
+        end
+
+        pivot_new = j + 1
+        pivot_dist = pivot_new - lo + 1
+        if pivot_dist == k
+            return a[pivot_new]
+        elseif k < pivot_dist
+            hi = pivot_new - 1
+        else
+            k = k - pivot_dist
+            lo = pivot_new + 1
+        end
+
+    end
+end
+
+select(a::AbstractVector, k::Integer) = _jl_quickselect(copy(a), k, 1, length(a))
+select!(a::AbstractVector, k::Integer) = _jl_quickselect(a, k, 1, length(a))