1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2009-12, 2014 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/>. */
19 #include "math/sort.h"
23 #include "data/case.h"
24 #include "data/casereader.h"
25 #include "data/casewriter-provider.h"
26 #include "data/casewriter.h"
27 #include "data/settings.h"
28 #include "data/subcase.h"
29 #include "libpspp/array.h"
30 #include "libpspp/assertion.h"
31 #include "libpspp/bt.h"
32 #include "math/merge.h"
34 #include "gl/xalloc.h"
37 #define _(msgid) gettext (msgid)
39 /* These should only be changed for testing purposes. */
41 int max_buffers = INT_MAX;
45 struct caseproto *proto;
46 struct subcase ordering;
48 struct pqueue *pqueue;
50 struct casewriter *run;
52 struct ccase *run_end;
55 static struct casewriter_class sort_casewriter_class;
57 static struct pqueue *pqueue_create (const struct subcase *,
58 const struct caseproto *);
59 static void pqueue_destroy (struct pqueue *);
60 static bool pqueue_is_full (const struct pqueue *);
61 static bool pqueue_is_empty (const struct pqueue *);
62 static void pqueue_push (struct pqueue *, struct ccase *, casenumber);
63 static struct ccase *pqueue_pop (struct pqueue *, casenumber *);
65 static void output_record (struct sort_writer *);
68 sort_create_writer (const struct subcase *ordering,
69 const struct caseproto *proto)
71 struct sort_writer *sort;
73 sort = xmalloc (sizeof *sort);
74 sort->proto = caseproto_ref (proto);
75 subcase_clone (&sort->ordering, ordering);
76 sort->merge = merge_create (ordering, proto);
77 sort->pqueue = pqueue_create (ordering, proto);
82 return casewriter_create (proto, &sort_casewriter_class, sort);
86 sort_casewriter_write (struct casewriter *writer UNUSED, void *sort_,
89 struct sort_writer *sort = sort_;
92 if (pqueue_is_full (sort->pqueue))
95 next_run = (sort->run_end == NULL
96 || subcase_compare_3way (&sort->ordering, c,
97 &sort->ordering, sort->run_end) < 0);
98 pqueue_push (sort->pqueue, c, sort->run_id + (next_run ? 1 : 0));
102 sort_casewriter_destroy (struct casewriter *writer UNUSED, void *sort_)
104 struct sort_writer *sort = sort_;
106 subcase_destroy (&sort->ordering);
107 merge_destroy (sort->merge);
108 pqueue_destroy (sort->pqueue);
109 casewriter_destroy (sort->run);
110 case_unref (sort->run_end);
111 caseproto_unref (sort->proto);
115 static struct casereader *
116 sort_casewriter_convert_to_reader (struct casewriter *writer, void *sort_)
118 struct sort_writer *sort = sort_;
119 struct casereader *output;
121 if (sort->run == NULL && sort->run_id == 0)
124 sort->run = mem_writer_create (sort->proto);
127 while (!pqueue_is_empty (sort->pqueue))
128 output_record (sort);
130 merge_append (sort->merge, casewriter_make_reader (sort->run));
133 output = merge_make_reader (sort->merge);
134 sort_casewriter_destroy (writer, sort);
139 output_record (struct sort_writer *sort)
141 struct ccase *min_case;
142 casenumber min_run_id;
144 min_case = pqueue_pop (sort->pqueue, &min_run_id);
146 printf ("\toutput: %f to run %d\n", case_num_idx (min_case, 0), min_run_id);
149 if (sort->run_id != min_run_id && sort->run != NULL)
151 merge_append (sort->merge, casewriter_make_reader (sort->run));
154 if (sort->run == NULL)
156 sort->run = tmpfile_writer_create (sort->proto);
157 sort->run_id = min_run_id;
160 case_unref (sort->run_end);
161 sort->run_end = case_ref (min_case);
162 casewriter_write (sort->run, min_case);
165 static struct casewriter_class sort_casewriter_class =
167 sort_casewriter_write,
168 sort_casewriter_destroy,
169 sort_casewriter_convert_to_reader,
172 /* Reads all the cases from INPUT. Sorts the cases according to
173 ORDERING. Returns the sorted cases in a new casereader.
174 INPUT is destroyed by this function.
177 sort_execute (struct casereader *input, const struct subcase *ordering)
179 struct casewriter *output =
180 sort_create_writer (ordering, casereader_get_proto (input));
181 casereader_transfer (input, output);
182 return casewriter_make_reader (output);
185 /* Reads all the cases from INPUT. Sorts the cases in ascending
186 order according to VARIABLE. Returns the sorted cases in a
187 new casereader. INPUT is destroyed by this function. */
189 sort_execute_1var (struct casereader *input, const struct variable *var)
192 struct casereader *reader;
194 subcase_init_var (&sc, var, SC_ASCEND);
195 reader = sort_execute (input, &sc);
196 subcase_destroy (&sc);
202 struct subcase ordering;
210 struct bt_node bt_node;
216 static int compare_pqueue_records (const struct bt_node *a,
217 const struct bt_node *b,
218 const void *ordering);
220 static struct pqueue *
221 pqueue_create (const struct subcase *ordering, const struct caseproto *proto)
225 pq = xmalloc (sizeof *pq);
226 subcase_clone (&pq->ordering, ordering);
227 pq->record_max = settings_get_workspace_cases (proto);
228 if (pq->record_max > max_buffers)
229 pq->record_max = max_buffers;
230 else if (pq->record_max < min_buffers)
231 pq->record_max = min_buffers;
232 bt_init (&pq->bt, compare_pqueue_records, &pq->ordering);
239 pqueue_destroy (struct pqueue *pq)
243 while (!pqueue_is_empty (pq))
246 struct ccase *c = pqueue_pop (pq, &id);
249 subcase_destroy (&pq->ordering);
255 pqueue_is_full (const struct pqueue *pq)
257 return bt_count (&pq->bt) >= pq->record_max;
261 pqueue_is_empty (const struct pqueue *pq)
263 return bt_is_empty (&pq->bt);
267 pqueue_push (struct pqueue *pq, struct ccase *c, casenumber id)
269 struct pqueue_record *r;
271 assert (!pqueue_is_full (pq));
273 r = xmalloc (sizeof *r);
277 bt_insert (&pq->bt, &r->bt_node);
280 static struct ccase *
281 pqueue_pop (struct pqueue *pq, casenumber *id)
283 struct pqueue_record *r;
286 assert (!pqueue_is_empty (pq));
288 r = bt_data (bt_first (&pq->bt), struct pqueue_record, bt_node);
289 bt_delete (&pq->bt, &r->bt_node);
296 /* Compares record-run tuples A and B on id, then on case data,
297 then on insertion order. */
299 compare_pqueue_records (const struct bt_node *a_, const struct bt_node *b_,
300 const void *ordering_)
302 const struct pqueue_record *a = bt_data (a_, struct pqueue_record, bt_node);
303 const struct pqueue_record *b = bt_data (b_, struct pqueue_record, bt_node);
304 const struct subcase *ordering = ordering_;
305 int result = a->id < b->id ? -1 : a->id > b->id;
307 result = subcase_compare_3way (ordering, a->c, ordering, b->c);
309 result = a->idx < b->idx ? -1 : a->idx > b->idx;