1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2009, 2010 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.h"
26 #include "data/casewriter-provider.h"
27 #include "data/settings.h"
28 #include "data/subcase.h"
29 #include "libpspp/assertion.h"
30 #include "libpspp/bt.h"
31 #include "math/merge.h"
33 #include "gl/xalloc.h"
36 #define _(msgid) gettext (msgid)
38 /* These should only be changed for testing purposes. */
40 int max_buffers = INT_MAX;
44 struct caseproto *proto;
45 struct subcase ordering;
47 struct pqueue *pqueue;
49 sort_distinct_combine_func *combine;
50 sort_distinct_destroy_func *destroy;
53 struct casewriter *run;
55 struct ccase *run_end;
58 static struct casewriter_class sort_casewriter_class;
60 static struct pqueue *pqueue_create (const struct subcase *,
61 const struct caseproto *,
62 sort_distinct_combine_func *, void *aux);
63 static void pqueue_destroy (struct pqueue *);
64 static bool pqueue_is_full (const struct pqueue *);
65 static bool pqueue_is_empty (const struct pqueue *);
66 static void pqueue_push (struct pqueue *, struct ccase *, casenumber);
67 static struct ccase *pqueue_pop (struct pqueue *, casenumber *);
69 static void output_record (struct sort_writer *);
72 sort_create_writer (const struct subcase *ordering,
73 const struct caseproto *proto)
75 return sort_distinct_create_writer (ordering, proto, NULL, NULL, NULL);
79 sort_distinct_create_writer (const struct subcase *ordering,
80 const struct caseproto *proto,
81 sort_distinct_combine_func *combine,
82 sort_distinct_destroy_func *destroy,
85 struct sort_writer *sort;
87 sort = xmalloc (sizeof *sort);
88 sort->proto = caseproto_ref (proto);
89 subcase_clone (&sort->ordering, ordering);
90 sort->merge = merge_create (ordering, proto, combine, aux);
91 sort->pqueue = pqueue_create (ordering, proto, combine, aux);
93 sort->combine = combine;
94 sort->destroy = destroy;
101 return casewriter_create (proto, &sort_casewriter_class, sort);
105 sort_casewriter_write (struct casewriter *writer UNUSED, void *sort_,
108 struct sort_writer *sort = sort_;
111 if (pqueue_is_full (sort->pqueue))
112 output_record (sort);
114 next_run = (sort->run_end == NULL
115 || subcase_compare_3way (&sort->ordering, c,
116 &sort->ordering, sort->run_end) < 0);
117 pqueue_push (sort->pqueue, c, sort->run_id + (next_run ? 1 : 0));
121 sort_casewriter_destroy (struct casewriter *writer UNUSED, void *sort_)
123 struct sort_writer *sort = sort_;
125 if (sort->destroy != NULL)
126 sort->destroy (sort->aux);
128 subcase_destroy (&sort->ordering);
129 merge_destroy (sort->merge);
130 pqueue_destroy (sort->pqueue);
131 casewriter_destroy (sort->run);
132 case_unref (sort->run_end);
133 caseproto_unref (sort->proto);
137 static struct casereader *
138 sort_casewriter_convert_to_reader (struct casewriter *writer, void *sort_)
140 struct sort_writer *sort = sort_;
141 struct casereader *output;
143 if (sort->run == NULL && sort->run_id == 0)
146 sort->run = mem_writer_create (sort->proto);
149 while (!pqueue_is_empty (sort->pqueue))
150 output_record (sort);
152 merge_append (sort->merge, casewriter_make_reader (sort->run));
155 output = merge_make_reader (sort->merge);
156 sort_casewriter_destroy (writer, sort);
161 output_record (struct sort_writer *sort)
163 struct ccase *min_case;
164 casenumber min_run_id;
166 min_case = pqueue_pop (sort->pqueue, &min_run_id);
168 printf ("\toutput: %f to run %d\n", case_num_idx (min_case, 0), min_run_id);
171 if (sort->run_id != min_run_id && sort->run != NULL)
173 merge_append (sort->merge, casewriter_make_reader (sort->run));
176 if (sort->run == NULL)
178 sort->run = tmpfile_writer_create (sort->proto);
179 sort->run_id = min_run_id;
182 case_unref (sort->run_end);
183 sort->run_end = case_ref (min_case);
184 casewriter_write (sort->run, min_case);
187 static struct casewriter_class sort_casewriter_class =
189 sort_casewriter_write,
190 sort_casewriter_destroy,
191 sort_casewriter_convert_to_reader,
194 /* Reads all the cases from INPUT. Sorts the cases according to
195 ORDERING. Returns the sorted cases in a new casereader. */
197 sort_execute (struct casereader *input, const struct subcase *ordering)
199 struct casewriter *output =
200 sort_create_writer (ordering, casereader_get_proto (input));
201 casereader_transfer (input, output);
202 return casewriter_make_reader (output);
205 /* Reads all the cases from INPUT. Sorts the cases in ascending
206 order according to VARIABLE. Returns the sorted cases in a
209 sort_execute_1var (struct casereader *input, const struct variable *var)
212 struct casereader *reader;
214 subcase_init_var (&sc, var, SC_ASCEND);
215 reader = sort_execute (input, &sc);
216 subcase_destroy (&sc);
222 struct subcase ordering;
227 sort_distinct_combine_func *combine;
233 struct bt_node bt_node;
239 static int compare_pqueue_records (const struct bt_node *a,
240 const struct bt_node *b,
241 const void *ordering);
243 static struct pqueue *
244 pqueue_create (const struct subcase *ordering, const struct caseproto *proto,
245 sort_distinct_combine_func *combine, void *aux)
249 pq = xmalloc (sizeof *pq);
250 subcase_clone (&pq->ordering, ordering);
251 pq->record_cap = settings_get_workspace_cases (proto);
252 if (pq->record_cap > max_buffers)
253 pq->record_cap = max_buffers;
254 else if (pq->record_cap < min_buffers)
255 pq->record_cap = min_buffers;
256 bt_init (&pq->bt, compare_pqueue_records, &pq->ordering);
259 pq->combine = combine;
266 pqueue_destroy (struct pqueue *pq)
270 while (!pqueue_is_empty (pq))
273 struct ccase *c = pqueue_pop (pq, &id);
276 subcase_destroy (&pq->ordering);
282 pqueue_is_full (const struct pqueue *pq)
284 return bt_count (&pq->bt) >= pq->record_cap;
288 pqueue_is_empty (const struct pqueue *pq)
290 return bt_is_empty (&pq->bt);
294 pqueue_push (struct pqueue *pq, struct ccase *c, casenumber id)
296 struct pqueue_record *r;
298 assert (!pqueue_is_full (pq));
300 r = xmalloc (sizeof *r);
304 bt_insert (&pq->bt, &r->bt_node);
306 if (pq->combine != NULL)
308 struct bt_node *q_ = bt_prev (&pq->bt, &r->bt_node);
311 struct pqueue_record *q = bt_data (q_, struct pqueue_record,
313 if (q->id == r->id && subcase_equal (&pq->ordering, q->c,
314 &pq->ordering, r->c))
316 bt_delete (&pq->bt, &r->bt_node);
318 q->c = pq->combine (q->c, r->c, pq->aux);
324 static struct ccase *
325 pqueue_pop (struct pqueue *pq, casenumber *id)
327 struct pqueue_record *r;
330 assert (!pqueue_is_empty (pq));
332 r = bt_data (bt_first (&pq->bt), struct pqueue_record, bt_node);
333 bt_delete (&pq->bt, &r->bt_node);
340 /* Compares record-run tuples A and B on id, then on case data,
341 then on insertion order. */
343 compare_pqueue_records (const struct bt_node *a_, const struct bt_node *b_,
344 const void *ordering_)
346 const struct pqueue_record *a = bt_data (a_, struct pqueue_record, bt_node);
347 const struct pqueue_record *b = bt_data (b_, struct pqueue_record, bt_node);
348 const struct subcase *ordering = ordering_;
349 int result = a->id < b->id ? -1 : a->id > b->id;
351 result = subcase_compare_3way (ordering, a->c, ordering, b->c);
353 result = a->idx < b->idx ? -1 : a->idx > b->idx;