/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2009 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2009, 2011, 2012 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
#include <config.h>
-#include "sort.h"
+#include "math/sort.h"
#include <stdio.h>
-#include <data/case.h>
-#include <data/casereader.h>
-#include <data/casewriter.h>
-#include <data/casewriter-provider.h>
-#include <data/settings.h>
-#include <data/subcase.h>
-#include <libpspp/array.h>
-#include <libpspp/assertion.h>
-#include <math/merge.h>
+#include "data/case.h"
+#include "data/casereader.h"
+#include "data/casewriter-provider.h"
+#include "data/casewriter.h"
+#include "data/settings.h"
+#include "data/subcase.h"
+#include "libpspp/array.h"
+#include "libpspp/assertion.h"
+#include "math/merge.h"
-#include "xalloc.h"
-
-#include "gettext.h"
-#define _(msgid) gettext (msgid)
+#include "gl/xalloc.h"
/* These should only be changed for testing purposes. */
int min_buffers = 64;
struct sort_writer
{
- size_t value_cnt;
+ struct caseproto *proto;
struct subcase ordering;
struct merge *merge;
struct pqueue *pqueue;
static struct casewriter_class sort_casewriter_class;
-static struct pqueue *pqueue_create (const struct subcase *, size_t);
+static struct pqueue *pqueue_create (const struct subcase *,
+ const struct caseproto *);
static void pqueue_destroy (struct pqueue *);
static bool pqueue_is_full (const struct pqueue *);
static bool pqueue_is_empty (const struct pqueue *);
static void output_record (struct sort_writer *);
struct casewriter *
-sort_create_writer (const struct subcase *ordering, size_t value_cnt)
+sort_create_writer (const struct subcase *ordering,
+ const struct caseproto *proto)
{
struct sort_writer *sort;
sort = xmalloc (sizeof *sort);
- sort->value_cnt = value_cnt;
+ sort->proto = caseproto_ref (proto);
subcase_clone (&sort->ordering, ordering);
- sort->merge = merge_create (ordering, value_cnt);
- sort->pqueue = pqueue_create (ordering, value_cnt);
+ sort->merge = merge_create (ordering, proto);
+ sort->pqueue = pqueue_create (ordering, proto);
sort->run = NULL;
sort->run_id = 0;
sort->run_end = NULL;
- return casewriter_create (value_cnt, &sort_casewriter_class, sort);
+ return casewriter_create (proto, &sort_casewriter_class, sort);
}
static void
pqueue_destroy (sort->pqueue);
casewriter_destroy (sort->run);
case_unref (sort->run_end);
+ caseproto_unref (sort->proto);
free (sort);
}
if (sort->run == NULL && sort->run_id == 0)
{
/* In-core sort. */
- sort->run = mem_writer_create (casewriter_get_value_cnt (writer));
+ sort->run = mem_writer_create (sort->proto);
sort->run_id = 1;
}
while (!pqueue_is_empty (sort->pqueue))
}
if (sort->run == NULL)
{
- sort->run = tmpfile_writer_create (sort->value_cnt);
+ sort->run = tmpfile_writer_create (sort->proto);
sort->run_id = min_run_id;
}
};
\f
/* Reads all the cases from INPUT. Sorts the cases according to
- ORDERING. Returns the sorted cases in a new casereader. */
+ ORDERING. Returns the sorted cases in a new casereader.
+ INPUT is destroyed by this function.
+ */
struct casereader *
sort_execute (struct casereader *input, const struct subcase *ordering)
{
struct casewriter *output =
- sort_create_writer (ordering, casereader_get_value_cnt (input));
+ sort_create_writer (ordering, casereader_get_proto (input));
casereader_transfer (input, output);
return casewriter_make_reader (output);
}
/* Reads all the cases from INPUT. Sorts the cases in ascending
order according to VARIABLE. Returns the sorted cases in a
- new casereader. */
+ new casereader. INPUT is destroyed by this function. */
struct casereader *
sort_execute_1var (struct casereader *input, const struct variable *var)
{
{
struct subcase ordering;
struct pqueue_record *records;
- size_t record_cnt;
- size_t record_cap;
+ size_t n_records; /* Current number of records. */
+ size_t allocated_records; /* Space currently allocated for records. */
+ size_t max_records; /* Max space we are willing to allocate. */
casenumber idx;
};
const void *pq_);
static struct pqueue *
-pqueue_create (const struct subcase *ordering, size_t value_cnt)
+pqueue_create (const struct subcase *ordering, const struct caseproto *proto)
{
struct pqueue *pq;
pq = xmalloc (sizeof *pq);
subcase_clone (&pq->ordering, ordering);
- pq->record_cap
- = settings_get_workspace_cases (value_cnt);
- if (pq->record_cap > max_buffers)
- pq->record_cap = max_buffers;
- else if (pq->record_cap < min_buffers)
- pq->record_cap = min_buffers;
- pq->record_cnt = 0;
- pq->records = xnmalloc (pq->record_cap, sizeof *pq->records);
+ pq->max_records = settings_get_workspace_cases (proto);
+ if (pq->max_records > max_buffers)
+ pq->max_records = max_buffers;
+ else if (pq->max_records < min_buffers)
+ pq->max_records = min_buffers;
+ pq->n_records = 0;
+ pq->allocated_records = 0;
+ pq->records = NULL;
pq->idx = 0;
return pq;
static bool
pqueue_is_full (const struct pqueue *pq)
{
- return pq->record_cnt >= pq->record_cap;
+ return pq->n_records >= pq->max_records;
}
static bool
pqueue_is_empty (const struct pqueue *pq)
{
- return pq->record_cnt == 0;
+ return pq->n_records == 0;
}
static void
assert (!pqueue_is_full (pq));
- r = &pq->records[pq->record_cnt++];
+ if (pq->n_records >= pq->allocated_records)
+ {
+ pq->allocated_records = pq->allocated_records * 2;
+ if (pq->allocated_records < 16)
+ pq->allocated_records = 16;
+ else if (pq->allocated_records > pq->max_records)
+ pq->allocated_records = pq->max_records;
+ pq->records = xrealloc (pq->records,
+ pq->allocated_records * sizeof *pq->records);
+ }
+
+ r = &pq->records[pq->n_records++];
r->id = id;
r->c = c;
r->idx = pq->idx++;
- push_heap (pq->records, pq->record_cnt, sizeof *pq->records,
+ push_heap (pq->records, pq->n_records, sizeof *pq->records,
compare_pqueue_records_minheap, pq);
}
assert (!pqueue_is_empty (pq));
- pop_heap (pq->records, pq->record_cnt--, sizeof *pq->records,
+ pop_heap (pq->records, pq->n_records--, sizeof *pq->records,
compare_pqueue_records_minheap, pq);
- r = &pq->records[pq->record_cnt];
+ r = &pq->records[pq->n_records];
*id = r->id;
return r->c;
}