1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2008, 2009, 2010, 2011, 2012 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 "language/stats/aggregate.h"
23 #include "data/any-writer.h"
24 #include "data/case.h"
25 #include "data/casegrouper.h"
26 #include "data/casereader.h"
27 #include "data/casewriter.h"
28 #include "data/dataset.h"
29 #include "data/dictionary.h"
30 #include "data/file-handle-def.h"
31 #include "data/format.h"
32 #include "data/settings.h"
33 #include "data/subcase.h"
34 #include "data/sys-file-writer.h"
35 #include "data/variable.h"
36 #include "language/command.h"
37 #include "language/data-io/file-handle.h"
38 #include "language/lexer/lexer.h"
39 #include "language/lexer/variable-parser.h"
40 #include "language/stats/sort-criteria.h"
41 #include "libpspp/assertion.h"
42 #include "libpspp/i18n.h"
43 #include "libpspp/message.h"
44 #include "libpspp/misc.h"
45 #include "libpspp/pool.h"
46 #include "libpspp/str.h"
47 #include "math/moments.h"
48 #include "math/percentiles.h"
49 #include "math/sort.h"
50 #include "math/statistic.h"
52 #include "gl/c-strcase.h"
53 #include "gl/minmax.h"
54 #include "gl/xalloc.h"
57 #define _(msgid) gettext (msgid)
58 #define N_(msgid) msgid
60 /* Argument for AGGREGATE function. */
63 double f; /* Numeric. */
64 char *c; /* Short or long string. */
67 /* Specifies how to make an aggregate variable. */
70 struct agr_var *next; /* Next in list. */
72 /* Collected during parsing. */
73 const struct variable *src; /* Source variable. */
74 struct variable *dest; /* Target variable. */
75 int function; /* Function. */
76 enum mv_class exclude; /* Classes of missing values to exclude. */
77 union agr_argument arg[2]; /* Arguments. */
79 /* Accumulated during AGGREGATE execution. */
84 struct moments1 *moments;
87 struct variable *subject;
88 struct variable *weight;
89 struct casewriter *writer;
93 /* Attributes of aggregation functions. */
94 const struct agr_func agr_func_tab[] =
96 {"SUM", N_("Sum of values"), AGR_SV_YES, 0, -1, {FMT_F, 8, 2}},
97 {"MEAN", N_("Mean average"), AGR_SV_YES, 0, -1, {FMT_F, 8, 2}},
98 {"MEDIAN", N_("Median average"), AGR_SV_YES, 0, -1, {FMT_F, 8, 2}},
99 {"SD", N_("Standard deviation"), AGR_SV_YES, 0, -1, {FMT_F, 8, 2}},
100 {"MAX", N_("Maximum value"), AGR_SV_YES, 0, VAL_STRING, {-1, -1, -1}},
101 {"MIN", N_("Minimum value"), AGR_SV_YES, 0, VAL_STRING, {-1, -1, -1}},
102 {"PGT", N_("Percentage greater than"), AGR_SV_YES, 1, VAL_NUMERIC, {FMT_F, 5, 1}},
103 {"PLT", N_("Percentage less than"), AGR_SV_YES, 1, VAL_NUMERIC, {FMT_F, 5, 1}},
104 {"PIN", N_("Percentage included in range"), AGR_SV_YES, 2, VAL_NUMERIC, {FMT_F, 5, 1}},
105 {"POUT", N_("Percentage excluded from range"), AGR_SV_YES, 2, VAL_NUMERIC, {FMT_F, 5, 1}},
106 {"FGT", N_("Fraction greater than"), AGR_SV_YES, 1, VAL_NUMERIC, {FMT_F, 5, 3}},
107 {"FLT", N_("Fraction less than"), AGR_SV_YES, 1, VAL_NUMERIC, {FMT_F, 5, 3}},
108 {"FIN", N_("Fraction included in range"), AGR_SV_YES, 2, VAL_NUMERIC, {FMT_F, 5, 3}},
109 {"FOUT", N_("Fraction excluded from range"), AGR_SV_YES, 2, VAL_NUMERIC, {FMT_F, 5, 3}},
110 {"N", N_("Number of cases"), AGR_SV_NO, 0, VAL_NUMERIC, {FMT_F, 7, 0}},
111 {"NU", N_("Number of cases (unweighted)"), AGR_SV_OPT, 0, VAL_NUMERIC, {FMT_F, 7, 0}},
112 {"NMISS", N_("Number of missing values"), AGR_SV_YES, 0, VAL_NUMERIC, {FMT_F, 7, 0}},
113 {"NUMISS", N_("Number of missing values (unweighted)"), AGR_SV_YES, 0, VAL_NUMERIC, {FMT_F, 7, 0}},
114 {"FIRST", N_("First non-missing value"), AGR_SV_YES, 0, VAL_STRING, {-1, -1, -1}},
115 {"LAST", N_("Last non-missing value"), AGR_SV_YES, 0, VAL_STRING, {-1, -1, -1}},
116 {NULL, NULL, AGR_SV_NO, 0, -1, {-1, -1, -1}},
119 /* Missing value types. */
120 enum missing_treatment
122 ITEMWISE, /* Missing values item by item. */
123 COLUMNWISE /* Missing values column by column. */
126 /* An entire AGGREGATE procedure. */
129 /* Break variables. */
130 struct subcase sort; /* Sort criteria (break variables). */
131 const struct variable **break_vars; /* Break variables. */
132 size_t break_var_cnt; /* Number of break variables. */
134 enum missing_treatment missing; /* How to treat missing values. */
135 struct agr_var *agr_vars; /* First aggregate variable. */
136 struct dictionary *dict; /* Aggregate dictionary. */
137 const struct dictionary *src_dict; /* Dict of the source */
138 int case_cnt; /* Counts aggregated cases. */
140 bool add_variables; /* True iff the aggregated variables should
141 be appended to the existing dictionary */
144 static void initialize_aggregate_info (struct agr_proc *);
146 static void accumulate_aggregate_info (struct agr_proc *,
147 const struct ccase *);
149 static bool parse_aggregate_functions (struct lexer *, const struct dictionary *,
151 static void agr_destroy (struct agr_proc *);
152 static void dump_aggregate_info (const struct agr_proc *agr,
153 struct casewriter *output,
154 const struct ccase *break_case);
158 /* Parses and executes the AGGREGATE procedure. */
160 cmd_aggregate (struct lexer *lexer, struct dataset *ds)
162 struct dictionary *dict = dataset_dict (ds);
164 struct file_handle *out_file = NULL;
165 struct casereader *input = NULL, *group;
166 struct casegrouper *grouper;
167 struct casewriter *output = NULL;
169 bool copy_documents = false;
170 bool presorted = false;
174 memset(&agr, 0 , sizeof (agr));
175 agr.missing = ITEMWISE;
177 subcase_init_empty (&agr.sort);
179 /* OUTFILE subcommand must be first. */
180 lex_match (lexer, T_SLASH);
181 if (!lex_force_match_id (lexer, "OUTFILE"))
183 lex_match (lexer, T_EQUALS);
184 if (!lex_match (lexer, T_ASTERISK))
186 out_file = fh_parse (lexer, FH_REF_FILE, dataset_session (ds));
187 if (out_file == NULL)
191 if (out_file == NULL && lex_match_id (lexer, "MODE"))
193 lex_match (lexer, T_EQUALS);
194 if (lex_match_id (lexer, "ADDVARIABLES"))
196 agr.add_variables = true;
198 /* presorted is assumed in ADDVARIABLES mode */
201 else if (lex_match_id (lexer, "REPLACE"))
203 agr.add_variables = false;
209 if ( agr.add_variables )
210 agr.dict = dict_clone (dict);
212 agr.dict = dict_create (dict_get_encoding (dict));
214 dict_set_label (agr.dict, dict_get_label (dict));
215 dict_set_documents (agr.dict, dict_get_documents (dict));
217 /* Read most of the subcommands. */
220 lex_match (lexer, T_SLASH);
222 if (lex_match_id (lexer, "MISSING"))
224 lex_match (lexer, T_EQUALS);
225 if (!lex_match_id (lexer, "COLUMNWISE"))
227 lex_error_expecting (lexer, "COLUMNWISE", NULL);
230 agr.missing = COLUMNWISE;
232 else if (lex_match_id (lexer, "DOCUMENT"))
233 copy_documents = true;
234 else if (lex_match_id (lexer, "PRESORTED"))
236 else if (lex_force_match_id (lexer, "BREAK"))
240 lex_match (lexer, T_EQUALS);
241 if (!parse_sort_criteria (lexer, dict, &agr.sort, &agr.break_vars,
244 agr.break_var_cnt = subcase_get_n_fields (&agr.sort);
246 if (! agr.add_variables)
247 for (i = 0; i < agr.break_var_cnt; i++)
248 dict_clone_var_assert (agr.dict, agr.break_vars[i]);
250 /* BREAK must follow the options. */
257 if (presorted && saw_direction)
258 msg (SW, _("When PRESORTED is specified, specifying sorting directions "
259 "with (A) or (D) has no effect. Output data will be sorted "
260 "the same way as the input data."));
262 /* Read in the aggregate functions. */
263 lex_match (lexer, T_SLASH);
264 if (!parse_aggregate_functions (lexer, dict, &agr))
267 /* Delete documents. */
269 dict_clear_documents (agr.dict);
271 /* Cancel SPLIT FILE. */
272 dict_set_split_vars (agr.dict, NULL, 0);
277 if (out_file == NULL)
279 /* The active dataset will be replaced by the aggregated data,
280 so TEMPORARY is moot. */
281 proc_cancel_temporary_transformations (ds);
282 proc_discard_output (ds);
283 output = autopaging_writer_create (dict_get_proto (agr.dict));
287 output = any_writer_open (out_file, agr.dict);
292 input = proc_open (ds);
293 if (!subcase_is_empty (&agr.sort) && !presorted)
295 input = sort_execute (input, &agr.sort);
296 subcase_clear (&agr.sort);
299 for (grouper = casegrouper_create_vars (input, agr.break_vars,
301 casegrouper_get_next_group (grouper, &group);
302 casereader_destroy (group))
304 struct casereader *placeholder = NULL;
305 struct ccase *c = casereader_peek (group, 0);
309 casereader_destroy (group);
313 initialize_aggregate_info (&agr);
315 if ( agr.add_variables )
316 placeholder = casereader_clone (group);
320 for (; (cg = casereader_read (group)) != NULL; case_unref (cg))
321 accumulate_aggregate_info (&agr, cg);
325 if (agr.add_variables)
328 for (; (cg = casereader_read (placeholder)) != NULL; case_unref (cg))
329 dump_aggregate_info (&agr, output, cg);
331 casereader_destroy (placeholder);
335 dump_aggregate_info (&agr, output, c);
339 if (!casegrouper_destroy (grouper))
342 if (!proc_commit (ds))
349 if (out_file == NULL)
351 struct casereader *next_input = casewriter_make_reader (output);
352 if (next_input == NULL)
355 dataset_set_dict (ds, agr.dict);
356 dataset_set_source (ds, next_input);
361 ok = casewriter_destroy (output);
374 casewriter_destroy (output);
377 return CMD_CASCADING_FAILURE;
380 /* Parse all the aggregate functions. */
382 parse_aggregate_functions (struct lexer *lexer, const struct dictionary *dict,
383 struct agr_proc *agr)
385 struct agr_var *tail; /* Tail of linked list starting at agr->vars. */
387 /* Parse everything. */
394 struct string function_name;
396 enum mv_class exclude;
397 const struct agr_func *function;
400 union agr_argument arg[2];
402 const struct variable **src;
415 ds_init_empty (&function_name);
417 /* Parse the list of target variables. */
418 while (!lex_match (lexer, T_EQUALS))
420 size_t n_dest_prev = n_dest;
422 if (!parse_DATA_LIST_vars (lexer, dict, &dest, &n_dest,
423 (PV_APPEND | PV_SINGLE | PV_NO_SCRATCH
427 /* Assign empty labels. */
431 dest_label = xnrealloc (dest_label, n_dest, sizeof *dest_label);
432 for (j = n_dest_prev; j < n_dest; j++)
433 dest_label[j] = NULL;
438 if (lex_is_string (lexer))
440 dest_label[n_dest - 1] = xstrdup (lex_tokcstr (lexer));
445 /* Get the name of the aggregation function. */
446 if (lex_token (lexer) != T_ID)
448 lex_error (lexer, _("expecting aggregation function"));
452 ds_assign_substring (&function_name, lex_tokss (lexer));
453 exclude = ds_chomp_byte (&function_name, '.') ? MV_SYSTEM : MV_ANY;
455 for (function = agr_func_tab; function->name; function++)
456 if (!c_strcasecmp (function->name, ds_cstr (&function_name)))
458 if (NULL == function->name)
460 msg (SE, _("Unknown aggregation function %s."),
461 ds_cstr (&function_name));
464 ds_destroy (&function_name);
465 func_index = function - agr_func_tab;
468 /* Check for leading lparen. */
469 if (!lex_match (lexer, T_LPAREN))
471 if (function->src_vars == AGR_SV_YES)
473 lex_force_match (lexer, T_LPAREN);
479 /* Parse list of source variables. */
481 int pv_opts = PV_NO_SCRATCH;
483 if (func_index == SUM || func_index == MEAN || func_index == SD)
484 pv_opts |= PV_NUMERIC;
485 else if (function->n_args)
486 pv_opts |= PV_SAME_TYPE;
488 if (!parse_variables_const (lexer, dict, &src, &n_src, pv_opts))
492 /* Parse function arguments, for those functions that
493 require arguments. */
494 if (function->n_args != 0)
495 for (i = 0; i < function->n_args; i++)
499 lex_match (lexer, T_COMMA);
500 if (lex_is_string (lexer))
502 arg[i].c = recode_string (dict_get_encoding (agr->dict),
503 "UTF-8", lex_tokcstr (lexer),
507 else if (lex_is_number (lexer))
509 arg[i].f = lex_tokval (lexer);
514 msg (SE, _("Missing argument %zu to %s."),
515 i + 1, function->name);
521 if (type != var_get_type (src[0]))
523 msg (SE, _("Arguments to %s must be of same type as "
524 "source variables."),
530 /* Trailing rparen. */
531 if (!lex_force_match (lexer, T_RPAREN))
534 /* Now check that the number of source variables match
535 the number of target variables. If we check earlier
536 than this, the user can get very misleading error
537 message, i.e. `AGGREGATE x=SUM(y t).' will get this
538 error message when a proper message would be more
539 like `unknown variable t'. */
542 msg (SE, _("Number of source variables (%zu) does not match "
543 "number of target variables (%zu)."),
548 if ((func_index == PIN || func_index == POUT
549 || func_index == FIN || func_index == FOUT)
550 && (var_is_numeric (src[0])
551 ? arg[0].f > arg[1].f
552 : str_compare_rpad (arg[0].c, arg[1].c) > 0))
554 union agr_argument t = arg[0];
558 msg (SW, _("The value arguments passed to the %s function "
559 "are out-of-order. They will be treated as if "
560 "they had been specified in the correct order."),
565 /* Finally add these to the linked list of aggregation
567 for (i = 0; i < n_dest; i++)
569 struct agr_var *v = xzalloc (sizeof *v);
571 /* Add variable to chain. */
572 if (agr->agr_vars != NULL)
580 /* Create the target variable in the aggregate
583 struct variable *destvar;
585 v->function = func_index;
591 if (var_is_alpha (src[i]))
593 v->function |= FSTRING;
594 v->string = xmalloc (var_get_width (src[i]));
597 if (function->alpha_type == VAL_STRING)
598 destvar = dict_clone_var_as (agr->dict, v->src, dest[i]);
601 assert (var_is_numeric (v->src)
602 || function->alpha_type == VAL_NUMERIC);
603 destvar = dict_create_var (agr->dict, dest[i], 0);
607 if ((func_index == N || func_index == NMISS)
608 && dict_get_weight (dict) != NULL)
609 f = fmt_for_output (FMT_F, 8, 2);
611 f = function->format;
612 var_set_both_formats (destvar, &f);
618 destvar = dict_create_var (agr->dict, dest[i], 0);
621 if ((func_index == N || func_index == NMISS)
622 && dict_get_weight (dict) != NULL)
623 f = fmt_for_output (FMT_F, 8, 2);
625 f = function->format;
626 var_set_both_formats (destvar, &f);
632 msg (SE, _("Variable name %s is not unique within the "
633 "aggregate file dictionary, which contains "
634 "the aggregate variables and the break "
642 var_set_label (destvar, dest_label[i], true);
647 v->exclude = exclude;
653 if (var_is_numeric (v->src))
654 for (j = 0; j < function->n_args; j++)
655 v->arg[j].f = arg[j].f;
657 for (j = 0; j < function->n_args; j++)
658 v->arg[j].c = xstrdup (arg[j].c);
662 if (src != NULL && var_is_alpha (src[0]))
663 for (i = 0; i < function->n_args; i++)
673 if (!lex_match (lexer, T_SLASH))
675 if (lex_token (lexer) == T_ENDCMD)
678 lex_error (lexer, "expecting end of command");
684 ds_destroy (&function_name);
685 for (i = 0; i < n_dest; i++)
688 free (dest_label[i]);
694 if (src && n_src && var_is_alpha (src[0]))
695 for (i = 0; i < function->n_args; i++)
708 agr_destroy (struct agr_proc *agr)
710 struct agr_var *iter, *next;
712 subcase_destroy (&agr->sort);
713 free (agr->break_vars);
714 for (iter = agr->agr_vars; iter; iter = next)
718 if (iter->function & FSTRING)
723 n_args = agr_func_tab[iter->function & FUNC].n_args;
724 for (i = 0; i < n_args; i++)
725 free (iter->arg[i].c);
728 else if (iter->function == SD)
729 moments1_destroy (iter->moments);
731 dict_destroy_internal_var (iter->subject);
732 dict_destroy_internal_var (iter->weight);
736 if (agr->dict != NULL)
737 dict_destroy (agr->dict);
742 /* Accumulates aggregation data from the case INPUT. */
744 accumulate_aggregate_info (struct agr_proc *agr, const struct ccase *input)
746 struct agr_var *iter;
748 bool bad_warn = true;
750 weight = dict_get_case_weight (agr->src_dict, input, &bad_warn);
752 for (iter = agr->agr_vars; iter; iter = iter->next)
755 const union value *v = case_data (input, iter->src);
756 int src_width = var_get_width (iter->src);
758 if (var_is_value_missing (iter->src, v, iter->exclude))
760 switch (iter->function)
763 case NMISS | FSTRING:
764 iter->dbl[0] += weight;
767 case NUMISS | FSTRING:
771 iter->saw_missing = true;
775 /* This is horrible. There are too many possibilities. */
776 switch (iter->function)
779 iter->dbl[0] += v->f * weight;
783 iter->dbl[0] += v->f * weight;
784 iter->dbl[1] += weight;
791 cout = case_create (casewriter_get_proto (iter->writer));
793 case_data_rw (cout, iter->subject)->f
794 = case_data (input, iter->src)->f;
796 wv = dict_get_case_weight (agr->src_dict, input, NULL);
798 case_data_rw (cout, iter->weight)->f = wv;
802 casewriter_write (iter->writer, cout);
806 moments1_add (iter->moments, v->f, weight);
809 iter->dbl[0] = MAX (iter->dbl[0], v->f);
813 /* Need to do some kind of Unicode collation thingy here */
814 if (memcmp (iter->string, value_str (v, src_width), src_width) < 0)
815 memcpy (iter->string, value_str (v, src_width), src_width);
819 iter->dbl[0] = MIN (iter->dbl[0], v->f);
823 if (memcmp (iter->string, value_str (v, src_width), src_width) > 0)
824 memcpy (iter->string, value_str (v, src_width), src_width);
829 if (v->f > iter->arg[0].f)
830 iter->dbl[0] += weight;
831 iter->dbl[1] += weight;
835 if (memcmp (iter->arg[0].c,
836 value_str (v, src_width), src_width) < 0)
837 iter->dbl[0] += weight;
838 iter->dbl[1] += weight;
842 if (v->f < iter->arg[0].f)
843 iter->dbl[0] += weight;
844 iter->dbl[1] += weight;
848 if (memcmp (iter->arg[0].c,
849 value_str (v, src_width), src_width) > 0)
850 iter->dbl[0] += weight;
851 iter->dbl[1] += weight;
855 if (iter->arg[0].f <= v->f && v->f <= iter->arg[1].f)
856 iter->dbl[0] += weight;
857 iter->dbl[1] += weight;
861 if (memcmp (iter->arg[0].c,
862 value_str (v, src_width), src_width) <= 0
863 && memcmp (iter->arg[1].c,
864 value_str (v, src_width), src_width) >= 0)
865 iter->dbl[0] += weight;
866 iter->dbl[1] += weight;
870 if (iter->arg[0].f > v->f || v->f > iter->arg[1].f)
871 iter->dbl[0] += weight;
872 iter->dbl[1] += weight;
876 if (memcmp (iter->arg[0].c,
877 value_str (v, src_width), src_width) > 0
878 || memcmp (iter->arg[1].c,
879 value_str (v, src_width), src_width) < 0)
880 iter->dbl[0] += weight;
881 iter->dbl[1] += weight;
885 iter->dbl[0] += weight;
898 case FIRST | FSTRING:
901 memcpy (iter->string, value_str (v, src_width), src_width);
910 memcpy (iter->string, value_str (v, src_width), src_width);
914 case NMISS | FSTRING:
916 case NUMISS | FSTRING:
917 /* Our value is not missing or it would have been
918 caught earlier. Nothing to do. */
924 switch (iter->function)
927 iter->dbl[0] += weight;
938 /* Writes an aggregated record to OUTPUT. */
940 dump_aggregate_info (const struct agr_proc *agr, struct casewriter *output, const struct ccase *break_case)
942 struct ccase *c = case_create (dict_get_proto (agr->dict));
944 if ( agr->add_variables)
946 case_copy (c, 0, break_case, 0, dict_get_var_cnt (agr->src_dict));
953 for (i = 0; i < agr->break_var_cnt; i++)
955 const struct variable *v = agr->break_vars[i];
956 value_copy (case_data_rw_idx (c, value_idx),
957 case_data (break_case, v),
966 for (i = agr->agr_vars; i; i = i->next)
968 union value *v = case_data_rw (c, i->dest);
969 int width = var_get_width (i->dest);
971 if (agr->missing == COLUMNWISE && i->saw_missing
972 && (i->function & FUNC) != N && (i->function & FUNC) != NU
973 && (i->function & FUNC) != NMISS && (i->function & FUNC) != NUMISS)
975 value_set_missing (v, width);
976 casewriter_destroy (i->writer);
983 v->f = i->int1 ? i->dbl[0] : SYSMIS;
986 v->f = i->dbl[1] != 0.0 ? i->dbl[0] / i->dbl[1] : SYSMIS;
992 struct percentile *median = percentile_create (0.5, i->cc);
993 struct order_stats *os = &median->parent;
994 struct casereader *sorted_reader = casewriter_make_reader (i->writer);
997 order_stats_accumulate (&os, 1,
1002 i->dbl[0] = percentile_calculate (median, PC_HAVERAGE);
1003 statistic_destroy (&median->parent.parent);
1012 /* FIXME: we should use two passes. */
1013 moments1_calculate (i->moments, NULL, NULL, &variance,
1015 if (variance != SYSMIS)
1016 v->f = sqrt (variance);
1023 v->f = i->int1 ? i->dbl[0] : SYSMIS;
1028 memcpy (value_str_rw (v, width), i->string, width);
1030 value_set_missing (v, width);
1039 case FOUT | FSTRING:
1040 v->f = i->dbl[1] ? i->dbl[0] / i->dbl[1] : SYSMIS;
1049 case POUT | FSTRING:
1050 v->f = i->dbl[1] ? i->dbl[0] / i->dbl[1] * 100.0 : SYSMIS;
1062 v->f = i->int1 ? i->dbl[0] : SYSMIS;
1064 case FIRST | FSTRING:
1065 case LAST | FSTRING:
1067 memcpy (value_str_rw (v, width), i->string, width);
1069 value_set_missing (v, width);
1072 case NMISS | FSTRING:
1076 case NUMISS | FSTRING:
1085 casewriter_write (output, c);
1088 /* Resets the state for all the aggregate functions. */
1090 initialize_aggregate_info (struct agr_proc *agr)
1092 struct agr_var *iter;
1094 for (iter = agr->agr_vars; iter; iter = iter->next)
1096 iter->saw_missing = false;
1097 iter->dbl[0] = iter->dbl[1] = iter->dbl[2] = 0.0;
1098 iter->int1 = iter->int2 = 0;
1099 switch (iter->function)
1102 iter->dbl[0] = DBL_MAX;
1105 memset (iter->string, 255, var_get_width (iter->src));
1108 iter->dbl[0] = -DBL_MAX;
1111 memset (iter->string, 0, var_get_width (iter->src));
1115 struct caseproto *proto;
1116 struct subcase ordering;
1118 proto = caseproto_create ();
1119 proto = caseproto_add_width (proto, 0);
1120 proto = caseproto_add_width (proto, 0);
1122 if ( ! iter->subject)
1123 iter->subject = dict_create_internal_var (0, 0);
1125 if ( ! iter->weight)
1126 iter->weight = dict_create_internal_var (1, 0);
1128 subcase_init_var (&ordering, iter->subject, SC_ASCEND);
1129 iter->writer = sort_create_writer (&ordering, proto);
1130 subcase_destroy (&ordering);
1131 caseproto_unref (proto);
1137 if (iter->moments == NULL)
1138 iter->moments = moments1_create (MOMENT_VARIANCE);
1140 moments1_clear (iter->moments);