projects
/
pspp
/ blobdiff
commit
grep
author
committer
pickaxe
?
search:
re
summary
|
shortlog
|
log
|
commit
|
commitdiff
|
tree
raw
|
inline
| side by side
linreg.c: (pspp_linreg_mse) New accessor function
[pspp]
/
src
/
language
/
stats
/
aggregate.c
diff --git
a/src/language/stats/aggregate.c
b/src/language/stats/aggregate.c
index f58b97cf8994c4eebcef4509c5cc1b8267f7f871..c217310910271f2639830eb09d44b2fcdffd69db 100644
(file)
--- a/
src/language/stats/aggregate.c
+++ b/
src/language/stats/aggregate.c
@@
-1,5
+1,5
@@
/* PSPP - a program for statistical analysis.
/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006
, 2008, 2009
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
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
@@
-19,7
+19,6
@@
#include <stdlib.h>
#include <data/any-writer.h>
#include <stdlib.h>
#include <data/any-writer.h>
-#include <data/case-ordering.h>
#include <data/case.h>
#include <data/casegrouper.h>
#include <data/casereader.h>
#include <data/case.h>
#include <data/casegrouper.h>
#include <data/casereader.h>
@@
-29,6
+28,7
@@
#include <data/format.h>
#include <data/procedure.h>
#include <data/settings.h>
#include <data/format.h>
#include <data/procedure.h>
#include <data/settings.h>
+#include <data/subcase.h>
#include <data/sys-file-writer.h>
#include <data/variable.h>
#include <language/command.h>
#include <data/sys-file-writer.h>
#include <data/variable.h>
#include <language/command.h>
@@
-43,6
+43,8
@@
#include <libpspp/str.h>
#include <math/moments.h>
#include <math/sort.h>
#include <libpspp/str.h>
#include <math/moments.h>
#include <math/sort.h>
+#include <math/statistic.h>
+#include <math/percentiles.h>
#include "minmax.h"
#include "xalloc.h"
#include "minmax.h"
#include "xalloc.h"
@@
-75,12
+77,17
@@
struct agr_var
char *string;
bool saw_missing;
struct moments1 *moments;
char *string;
bool saw_missing;
struct moments1 *moments;
+ double cc;
+
+ struct variable *subject;
+ struct variable *weight;
+ struct casewriter *writer;
};
/* Aggregation functions. */
enum
{
};
/* Aggregation functions. */
enum
{
- NONE, SUM, MEAN, SD, MAX, MIN, PGT, PLT, PIN, POUT, FGT, FLT, FIN,
+ NONE, SUM, MEAN,
MEDIAN,
SD, MAX, MIN, PGT, PLT, PIN, POUT, FGT, FLT, FIN,
FOUT, N, NU, NMISS, NUMISS, FIRST, LAST,
N_AGR_FUNCS, N_NO_VARS, NU_NO_VARS,
FUNC = 0x1f, /* Function mask. */
FOUT, N, NU, NMISS, NUMISS, FIRST, LAST,
N_AGR_FUNCS, N_NO_VARS, NU_NO_VARS,
FUNC = 0x1f, /* Function mask. */
@@
-102,6
+109,7
@@
static const struct agr_func agr_func_tab[] =
{"<NONE>", 0, -1, {0, 0, 0}},
{"SUM", 0, -1, {FMT_F, 8, 2}},
{"MEAN", 0, -1, {FMT_F, 8, 2}},
{"<NONE>", 0, -1, {0, 0, 0}},
{"SUM", 0, -1, {FMT_F, 8, 2}},
{"MEAN", 0, -1, {FMT_F, 8, 2}},
+ {"MEDIAN", 0, -1, {FMT_F, 8, 2}},
{"SD", 0, -1, {FMT_F, 8, 2}},
{"MAX", 0, VAL_STRING, {-1, -1, -1}},
{"MIN", 0, VAL_STRING, {-1, -1, -1}},
{"SD", 0, -1, {FMT_F, 8, 2}},
{"MAX", 0, VAL_STRING, {-1, -1, -1}},
{"MIN", 0, VAL_STRING, {-1, -1, -1}},
@@
-135,10
+143,10
@@
enum missing_treatment
struct agr_proc
{
/* Break variables. */
struct agr_proc
{
/* Break variables. */
- struct
case_ordering *sort; /* Sort criteria
. */
+ struct
subcase sort; /* Sort criteria (break variables)
. */
const struct variable **break_vars; /* Break variables. */
size_t break_var_cnt; /* Number of break variables. */
const struct variable **break_vars; /* Break variables. */
size_t break_var_cnt; /* Number of break variables. */
- struct ccase
break_case;
/* Last values of break variables. */
+ struct ccase
*break_case;
/* Last values of break variables. */
enum missing_treatment missing; /* How to treat missing values. */
struct agr_var *agr_vars; /* First aggregate variable. */
enum missing_treatment missing; /* How to treat missing values. */
struct agr_var *agr_vars; /* First aggregate variable. */
@@
-149,6
+157,7
@@
struct agr_proc
static void initialize_aggregate_info (struct agr_proc *,
const struct ccase *);
static void initialize_aggregate_info (struct agr_proc *,
const struct ccase *);
+
static void accumulate_aggregate_info (struct agr_proc *,
const struct ccase *);
/* Prototypes. */
static void accumulate_aggregate_info (struct agr_proc *,
const struct ccase *);
/* Prototypes. */
@@
-178,10
+187,11
@@
cmd_aggregate (struct lexer *lexer, struct dataset *ds)
memset(&agr, 0 , sizeof (agr));
agr.missing = ITEMWISE;
memset(&agr, 0 , sizeof (agr));
agr.missing = ITEMWISE;
-
case_nullify (&agr.break_case)
;
+
agr.break_case = NULL
;
agr.dict = dict_create ();
agr.src_dict = dict;
agr.dict = dict_create ();
agr.src_dict = dict;
+ subcase_init_empty (&agr.sort);
dict_set_label (agr.dict, dict_get_label (dict));
dict_set_documents (agr.dict, dict_get_documents (dict));
dict_set_label (agr.dict, dict_get_label (dict));
dict_set_documents (agr.dict, dict_get_documents (dict));
@@
-220,13
+230,10
@@
cmd_aggregate (struct lexer *lexer, struct dataset *ds)
int i;
lex_match (lexer, '=');
int i;
lex_match (lexer, '=');
- agr.sort = parse_case_ordering (lexer, dict,
-
- &saw_direction);
- if (agr.sort == NULL)
+ if (!parse_sort_criteria (lexer, dict, &agr.sort, &agr.break_vars,
+ &saw_direction))
goto error;
goto error;
- case_ordering_get_vars (agr.sort,
- &agr.break_vars, &agr.break_var_cnt);
+ agr.break_var_cnt = subcase_get_n_fields (&agr.sort);
for (i = 0; i < agr.break_var_cnt; i++)
dict_clone_var_assert (agr.dict, agr.break_vars[i],
for (i = 0; i < agr.break_var_cnt; i++)
dict_clone_var_assert (agr.dict, agr.break_vars[i],
@@
-277,10
+284,10
@@
cmd_aggregate (struct lexer *lexer, struct dataset *ds)
}
input = proc_open (ds);
}
input = proc_open (ds);
- if (
agr.sort != NULL
&& !presorted)
+ if (
!subcase_is_empty (&agr.sort)
&& !presorted)
{
{
- input = sort_execute (input, agr.sort);
-
agr.sort = NULL
;
+ input = sort_execute (input,
&
agr.sort);
+
subcase_clear (&agr.sort)
;
}
for (grouper = casegrouper_create_vars (input, agr.break_vars,
}
for (grouper = casegrouper_create_vars (input, agr.break_vars,
@@
-288,18
+295,17
@@
cmd_aggregate (struct lexer *lexer, struct dataset *ds)
casegrouper_get_next_group (grouper, &group);
casereader_destroy (group))
{
casegrouper_get_next_group (grouper, &group);
casereader_destroy (group))
{
- struct ccase c;
-
- if (!casereader_peek (group, 0, &c))
+ struct ccase *c = casereader_peek (group, 0);
+ if (c == NULL)
{
casereader_destroy (group);
continue;
}
{
casereader_destroy (group);
continue;
}
- initialize_aggregate_info (&agr,
&
c);
- case_
destroy (&
c);
+ initialize_aggregate_info (&agr, c);
+ case_
unref (
c);
- for (;
casereader_read (group, &c); case_destroy (&
c))
- accumulate_aggregate_info (&agr,
&
c);
+ for (;
(c = casereader_read (group)) != NULL; case_unref (
c))
+ accumulate_aggregate_info (&agr, c);
dump_aggregate_info (&agr, output);
}
if (!casegrouper_destroy (grouper))
dump_aggregate_info (&agr, output);
}
if (!casegrouper_destroy (grouper))
@@
-344,7
+350,8
@@
error:
/* Parse all the aggregate functions. */
static bool
/* Parse all the aggregate functions. */
static bool
-parse_aggregate_functions (struct lexer *lexer, const struct dictionary *dict, struct agr_proc *agr)
+parse_aggregate_functions (struct lexer *lexer, const struct dictionary *dict,
+ struct agr_proc *agr)
{
struct agr_var *tail; /* Tail of linked list starting at agr->vars. */
{
struct agr_var *tail; /* Tail of linked list starting at agr->vars. */
@@
-545,7
+552,7
@@
parse_aggregate_functions (struct lexer *lexer, const struct dictionary *dict, s
variables. */
for (i = 0; i < n_dest; i++)
{
variables. */
for (i = 0; i < n_dest; i++)
{
- struct agr_var *v = x
m
alloc (sizeof *v);
+ struct agr_var *v = x
z
alloc (sizeof *v);
/* Add variable to chain. */
if (agr->agr_vars != NULL)
/* Add variable to chain. */
if (agr->agr_vars != NULL)
@@
-684,9
+691,9
@@
agr_destroy (struct agr_proc *agr)
{
struct agr_var *iter, *next;
{
struct agr_var *iter, *next;
-
case_ordering_destroy (
agr->sort);
+
subcase_destroy (&
agr->sort);
free (agr->break_vars);
free (agr->break_vars);
- case_
destroy (&
agr->break_case);
+ case_
unref (
agr->break_case);
for (iter = agr->agr_vars; iter; iter = next)
{
next = iter->next;
for (iter = agr->agr_vars; iter; iter = next)
{
next = iter->next;
@@
-703,6
+710,10
@@
agr_destroy (struct agr_proc *agr)
}
else if (iter->function == SD)
moments1_destroy (iter->moments);
}
else if (iter->function == SD)
moments1_destroy (iter->moments);
+
+ var_destroy (iter->subject);
+ var_destroy (iter->weight);
+
free (iter);
}
if (agr->dict != NULL)
free (iter);
}
if (agr->dict != NULL)
@@
-755,6
+766,23
@@
accumulate_aggregate_info (struct agr_proc *agr, const struct ccase *input)
iter->dbl[0] += v->f * weight;
iter->dbl[1] += weight;
break;
iter->dbl[0] += v->f * weight;
iter->dbl[1] += weight;
break;
+ case MEDIAN:
+ {
+ double wv ;
+ struct ccase *cout = case_create (2);
+
+ case_data_rw (cout, iter->subject)->f
+ = case_data (input, iter->src)->f;
+
+ wv = dict_get_case_weight (agr->src_dict, input, NULL);
+
+ case_data_rw (cout, iter->weight)->f = wv;
+
+ iter->cc += wv;
+
+ casewriter_write (iter->writer, cout);
+ }
+ break;
case SD:
moments1_add (iter->moments, v->f, weight);
break;
case SD:
moments1_add (iter->moments, v->f, weight);
break;
@@
-885,9
+913,7
@@
accumulate_aggregate_info (struct agr_proc *agr, const struct ccase *input)
static void
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
{
static void
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
{
- struct ccase c;
-
- case_create (&c, dict_get_next_value_idx (agr->dict));
+ struct ccase *c = case_create (dict_get_next_value_idx (agr->dict));
{
int value_idx = 0;
{
int value_idx = 0;
@@
-897,8
+923,8
@@
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
{
const struct variable *v = agr->break_vars[i];
size_t value_cnt = var_get_value_cnt (v);
{
const struct variable *v = agr->break_vars[i];
size_t value_cnt = var_get_value_cnt (v);
- memcpy (case_data_rw_idx (
&
c, value_idx),
- case_data (
&
agr->break_case, v),
+ memcpy (case_data_rw_idx (c, value_idx),
+ case_data (agr->break_case, v),
sizeof (union value) * value_cnt);
value_idx += value_cnt;
}
sizeof (union value) * value_cnt);
value_idx += value_cnt;
}
@@
-909,7
+935,8
@@
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
for (i = agr->agr_vars; i; i = i->next)
{
for (i = agr->agr_vars; i; i = i->next)
{
- union value *v = case_data_rw (&c, i->dest);
+ union value *v = case_data_rw (c, i->dest);
+
if (agr->missing == COLUMNWISE && i->saw_missing
&& (i->function & FUNC) != N && (i->function & FUNC) != NU
if (agr->missing == COLUMNWISE && i->saw_missing
&& (i->function & FUNC) != N && (i->function & FUNC) != NU
@@
-919,6
+946,9
@@
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
memset (v->s, ' ', var_get_width (i->dest));
else
v->f = SYSMIS;
memset (v->s, ' ', var_get_width (i->dest));
else
v->f = SYSMIS;
+
+ casewriter_destroy (i->writer);
+
continue;
}
continue;
}
@@
-930,6
+960,25
@@
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
case MEAN:
v->f = i->dbl[1] != 0.0 ? i->dbl[0] / i->dbl[1] : SYSMIS;
break;
case MEAN:
v->f = i->dbl[1] != 0.0 ? i->dbl[0] / i->dbl[1] : SYSMIS;
break;
+ case MEDIAN:
+ {
+ struct casereader *sorted_reader;
+ struct order_stats *median = percentile_create (0.5, i->cc);
+
+ sorted_reader = casewriter_make_reader (i->writer);
+
+ order_stats_accumulate (&median, 1,
+ sorted_reader,
+ i->weight,
+ i->subject,
+ i->exclude);
+
+ v->f = percentile_calculate ((struct percentile *) median,
+ PC_HAVERAGE);
+
+ statistic_destroy ((struct statistic *) median);
+ }
+ break;
case SD:
{
double variance;
case SD:
{
double variance;
@@
-1013,7
+1062,7
@@
dump_aggregate_info (struct agr_proc *agr, struct casewriter *output)
}
}
}
}
- casewriter_write (output,
&
c);
+ casewriter_write (output, c);
}
/* Resets the state for all the aggregate functions. */
}
/* Resets the state for all the aggregate functions. */
@@
-1022,8
+1071,8
@@
initialize_aggregate_info (struct agr_proc *agr, const struct ccase *input)
{
struct agr_var *iter;
{
struct agr_var *iter;
- case_
destroy (&
agr->break_case);
-
case_clone (&agr->break_case,
input);
+ case_
unref (
agr->break_case);
+
agr->break_case = case_ref (
input);
for (iter = agr->agr_vars; iter; iter = iter->next)
{
for (iter = agr->agr_vars; iter; iter = iter->next)
{
@@
-1044,6
+1093,23
@@
initialize_aggregate_info (struct agr_proc *agr, const struct ccase *input)
case MAX | FSTRING:
memset (iter->string, 0, var_get_width (iter->src));
break;
case MAX | FSTRING:
memset (iter->string, 0, var_get_width (iter->src));
break;
+ case MEDIAN:
+ {
+ struct subcase ordering;
+
+ if ( ! iter->subject)
+ iter->subject = var_create_internal (0);
+
+ if ( ! iter->weight)
+ iter->weight = var_create_internal (1);
+
+ subcase_init_var (&ordering, iter->subject, SC_ASCEND);
+ iter->writer = sort_create_writer (&ordering, 2);
+ subcase_destroy (&ordering);
+
+ iter->cc = 0;
+ }
+ break;
case SD:
if (iter->moments == NULL)
iter->moments = moments1_create (MOMENT_VARIANCE);
case SD:
if (iter->moments == NULL)
iter->moments = moments1_create (MOMENT_VARIANCE);