1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
23 #include <data/case-ordering.h>
24 #include <data/case.h>
25 #include <data/casereader.h>
26 #include <data/casewriter.h>
27 #include <data/casewriter-provider.h>
28 #include <data/settings.h>
29 #include <libpspp/alloc.h>
30 #include <libpspp/array.h>
31 #include <libpspp/assertion.h>
32 #include <math/merge.h>
35 #define _(msgid) gettext (msgid)
37 /* These should only be changed for testing purposes. */
39 int max_buffers = INT_MAX;
43 struct case_ordering *ordering;
45 struct pqueue *pqueue;
47 struct casewriter *run;
52 static struct casewriter_class sort_casewriter_class;
54 static struct pqueue *pqueue_create (const struct case_ordering *);
55 static void pqueue_destroy (struct pqueue *);
56 static bool pqueue_is_full (const struct pqueue *);
57 static bool pqueue_is_empty (const struct pqueue *);
58 static void pqueue_push (struct pqueue *, struct ccase *, casenumber);
59 static void pqueue_pop (struct pqueue *, struct ccase *, casenumber *);
61 static void output_record (struct sort_writer *);
64 sort_create_writer (struct case_ordering *ordering)
66 size_t value_cnt = case_ordering_get_value_cnt (ordering);
67 struct sort_writer *sort;
69 sort = xmalloc (sizeof *sort);
70 sort->ordering = case_ordering_clone (ordering);
71 sort->merge = merge_create (ordering);
72 sort->pqueue = pqueue_create (ordering);
75 case_nullify (&sort->run_end);
77 case_ordering_destroy (ordering);
79 return casewriter_create (value_cnt, &sort_casewriter_class, sort);
83 sort_casewriter_write (struct casewriter *writer UNUSED, void *sort_,
86 struct sort_writer *sort = sort_;
89 if (pqueue_is_full (sort->pqueue))
92 next_run = (case_is_null (&sort->run_end)
93 || case_ordering_compare_cases (c, &sort->run_end,
95 pqueue_push (sort->pqueue, c, sort->run_id + (next_run ? 1 : 0));
99 sort_casewriter_destroy (struct casewriter *writer UNUSED, void *sort_)
101 struct sort_writer *sort = sort_;
103 case_ordering_destroy (sort->ordering);
104 merge_destroy (sort->merge);
105 pqueue_destroy (sort->pqueue);
106 casewriter_destroy (sort->run);
107 case_destroy (&sort->run_end);
111 static struct casereader *
112 sort_casewriter_convert_to_reader (struct casewriter *writer, void *sort_)
114 struct sort_writer *sort = sort_;
115 struct casereader *output;
117 if (sort->run == NULL && sort->run_id == 0)
120 sort->run = mem_writer_create (case_ordering_get_value_cnt (
124 while (!pqueue_is_empty (sort->pqueue))
125 output_record (sort);
127 merge_append (sort->merge, casewriter_make_reader (sort->run));
130 output = merge_make_reader (sort->merge);
131 sort_casewriter_destroy (writer, sort);
136 output_record (struct sort_writer *sort)
138 struct ccase min_case;
139 casenumber min_run_id;
141 pqueue_pop (sort->pqueue, &min_case, &min_run_id);
143 printf ("\toutput: %f to run %d\n", case_num_idx (&min_case, 0), min_run_id);
146 if (sort->run_id != min_run_id && sort->run != NULL)
148 merge_append (sort->merge, casewriter_make_reader (sort->run));
151 if (sort->run == NULL)
153 sort->run = tmpfile_writer_create (case_ordering_get_value_cnt (
155 sort->run_id = min_run_id;
158 case_destroy (&sort->run_end);
159 case_clone (&sort->run_end, &min_case);
161 casewriter_write (sort->run, &min_case);
164 static struct casewriter_class sort_casewriter_class =
166 sort_casewriter_write,
167 sort_casewriter_destroy,
168 sort_casewriter_convert_to_reader,
171 /* Reads all the cases from INPUT. Sorts the cases according to
172 ORDERING. Returns the sorted cases in a new casereader, or a
173 null pointer if an I/O error occurs. Both INPUT and ORDERING
174 are destroyed upon return, regardless of success. */
176 sort_execute (struct casereader *input, struct case_ordering *ordering)
178 struct casewriter *output = sort_create_writer (ordering);
179 casereader_transfer (input, output);
180 return casewriter_make_reader (output);
185 struct case_ordering *ordering;
186 struct pqueue_record *records;
199 static int compare_pqueue_records_minheap (const void *a, const void *b,
202 static struct pqueue *
203 pqueue_create (const struct case_ordering *ordering)
207 pq = xmalloc (sizeof *pq);
208 pq->ordering = case_ordering_clone (ordering);
210 = get_workspace_cases (case_ordering_get_value_cnt (ordering));
211 if (pq->record_cap > max_buffers)
212 pq->record_cap = max_buffers;
213 else if (pq->record_cap < min_buffers)
214 pq->record_cap = min_buffers;
216 pq->records = xnmalloc (pq->record_cap, sizeof *pq->records);
223 pqueue_destroy (struct pqueue *pq)
227 while (!pqueue_is_empty (pq))
231 pqueue_pop (pq, &c, &id);
234 case_ordering_destroy (pq->ordering);
241 pqueue_is_full (const struct pqueue *pq)
243 return pq->record_cnt >= pq->record_cap;
247 pqueue_is_empty (const struct pqueue *pq)
249 return pq->record_cnt == 0;
253 pqueue_push (struct pqueue *pq, struct ccase *c, casenumber id)
255 struct pqueue_record *r;
257 assert (!pqueue_is_full (pq));
259 r = &pq->records[pq->record_cnt++];
261 case_move (&r->c, c);
264 push_heap (pq->records, pq->record_cnt, sizeof *pq->records,
265 compare_pqueue_records_minheap, pq);
269 pqueue_pop (struct pqueue *pq, struct ccase *c, casenumber *id)
271 struct pqueue_record *r;
273 assert (!pqueue_is_empty (pq));
275 pop_heap (pq->records, pq->record_cnt--, sizeof *pq->records,
276 compare_pqueue_records_minheap, pq);
278 r = &pq->records[pq->record_cnt];
280 case_move (c, &r->c);
283 /* Compares record-run tuples A and B on id, then on case data,
284 then on insertion order, in descending order. */
286 compare_pqueue_records_minheap (const void *a_, const void *b_,
289 const struct pqueue_record *a = a_;
290 const struct pqueue_record *b = b_;
291 const struct pqueue *pq = pq_;
292 int result = a->id < b->id ? -1 : a->id > b->id;
294 result = case_ordering_compare_cases (&a->c, &b->c, pq->ordering);
296 result = a->idx < b->idx ? -1 : a->idx > b->idx;