begin comment This example doesn't work with marst 2.1 because of incorrect passing a formal parameter called by name to other procedure (but it does work with marst 2.2). This example program was reported by Paulo Barreto <paulo.barreto@terra.com.br>; comment ALGORITHM 232 HEAPSORT J. W. J. Williams (Received 1 Oct. 1963 and revised 15 Feb. 1964) Elliott Bros. (London) Ltd., Borehamwood, Herts, England; comment The following procedures are related to TREESORT [R. W. Floyd, Alg. 113, Comm. ACM 5 (Aug. 1962), 434, and A. F. Kaupe, Jr., Alg. 143 and and 144, Comm. ACM 5 (Dec. 1962), 604] but avoid the use of pointers and so preserve storage space. All the procedures operate on single word items, stored as elements 1 to n of the array A. The elements are normally so arranged that A[i] <= A[j] for 2 <= j <= n, i = j % 2. Such an arrangement will be called a heap. A[1] is always the least element of the heap. The procedure SETHEAP arranges n elements as a heap, INHEAP adds a new element to an existing heap, OUTHEAP extracts the least element from a heap, and SWOPHEAP is effectively the result of INHEAP followed by OUTHEAP. In all cases the array A contains elements arranged as a heap on exit. SWOPHEAP is essentially the same as the tournament sort described by K. E. Iverson - A Programming Language, 1962, pp. 223--226 - which is a top to bottom method, but it uses an improved storage allocation and initialisation. INHEAP resembles TREESORT in being a bottom to top method. HEAPSORT can thus be considered as a marriage of these two methods. The procedures may be used for replacement-selection sorting, for sorting the elements of an array, or for choosing the current minimum of any set of items to which new items are added from time to time. The procedures are the more useful because the active elements of the array are maintained densely packed, as elements A[1] to A[n]; procedure SWOPHEAP(A, n, in, out); value in, n; integer n; real in, out; real array A; comment SWOPHEAP is given an array A, elements A[1] to A[n] forming a heap, n >= 0. SWOPHEAP effectively adds the element in to the heap, extracts ans assigns to out the value of the least member of the resulting set, and leaves the remaining elements in a heap of the original size. In this process elements 1 to (n + 1) of the array A may be disturbed. The maximum number of repetitions of the cycle labeled scan is log_2 n; begin integer i, j; real temp, temp 1; if in <= A[1] then out := 1 else begin i := 1; A[n + 1] := in; comment this last statement is only necessary in case j = n at some stage, or n = 0; out := A[1]; scan: j := i + i; if j <= n then begin temp := A[j]; temp 1 := A[j + 1]; if temp 1 < temp then begin temp := temp 1; j := j + 1 end; if temp < in then begin A[i] := temp; i := j; go to scan end end; A[i] := in end end SWOPHEAP; procedure INHEAP(A, n, in); value in; integer n; real in; real array A; comment INHEAP is given an array A, elements A[1] to A[n] forming a heap and n >= 0. INHEAP adds the element in to the heap and adjusts n accordingly. The cycle labeled scan may be repeated log_2 n times, but on average is repeated twice only; begin integer i, j; i := n := n + 1; scan: if i > 1 then begin j := i % 2; if in < A[j] then begin A[i] := A[j]; i := j; go to scan end end; A[i] := in end INHEAP; procedure OUTHEAP(A, n, out); integer n; real out; real array A; comment given array A, elements 1 to n of which form a heap, n >= 1, OUTHEAP assigns to out the value of A[1], the least member of the heap, and rearranges the remaining members as elements 1 to n - 1 of A. Also, n is adjusted accordingly; begin SWOPHEAP(A, n - 1, A[n], out); n := n - 1 end OUTHEAP; procedure SETHEAP(A, n); value n; integer n; real array A; comment SETHEAP rearranges the elements A[1] to A[n] to form a heap; begin integer j; j := 1; L: INHEAP(A, j, A[j + 1]); if j < n then go to L end SETHEAP; procedure HEAPSORT TEST(n); value n; integer n; comment HEAPSORT TEST tests the implementation on a random array of n elements. This procedure is not part of the original algorithm definition, and is only provided for completeness; begin integer i; real out; real array A[1:n]; comment create a test array of n elements; for i := 1 step 1 until n do begin A[i] := random end; SETHEAP(A, n); comment CAVEAT - the following loop does not work; i := n; loop: OUTHEAP(A, i, out); A[i + 1] := out; comment remember that i is decremented by OUTHEAP; if i > 1 then go to loop; comment HOWEVER, the following code does work; for i := n step -1 until 2 do begin SWOPHEAP(A, i - 1, A[i], out); A[i] := out end; comment now check if A is sorted in descending order; for i := 2 step 1 until n do begin if A[i] > A[i - 1] then begin outstring(1, "HEAPSORT error!\n"); go to exit end end; outstring(1, "HEAPSORT implementation is correct!\n"); exit: end HEAPSORT TEST; real procedure random; begin inline("my_dsa.retval.u.real _val = rand();") end random; comment main program; HEAPSORT TEST(10) end