1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2021 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/>. */
22 #include "data/casegrouper.h"
23 #include "data/casereader.h"
24 #include "data/casewriter.h"
25 #include "data/data-in.h"
26 #include "data/data-out.h"
27 #include "data/dataset.h"
28 #include "data/dictionary.h"
29 #include "data/mrset.h"
30 #include "data/subcase.h"
31 #include "data/value-labels.h"
32 #include "language/command.h"
33 #include "language/dictionary/split-file.h"
34 #include "language/lexer/format-parser.h"
35 #include "language/lexer/lexer.h"
36 #include "language/lexer/token.h"
37 #include "language/lexer/variable-parser.h"
38 #include "libpspp/array.h"
39 #include "libpspp/assertion.h"
40 #include "libpspp/hash-functions.h"
41 #include "libpspp/hmap.h"
42 #include "libpspp/i18n.h"
43 #include "libpspp/message.h"
44 #include "libpspp/string-array.h"
45 #include "math/mode.h"
46 #include "math/moments.h"
47 #include "math/percentiles.h"
48 #include "math/sort.h"
49 #include "output/pivot-table.h"
51 #include "gl/minmax.h"
52 #include "gl/xalloc.h"
55 #define _(msgid) gettext (msgid)
56 #define N_(msgid) (msgid)
60 CTVL_NONE = SETTINGS_VALUE_SHOW_DEFAULT,
61 CTVL_NAME = SETTINGS_VALUE_SHOW_VALUE,
62 CTVL_LABEL = SETTINGS_VALUE_SHOW_LABEL,
63 CTVL_BOTH = SETTINGS_VALUE_SHOW_BOTH,
66 enum ctables_function_type
68 /* A function that operates on data in a single cell. The function does
69 not have an unweighted version. */
72 /* A function that operates on data in a single cell. The function has an
73 unweighted version. */
76 /* A function that operates on an area of cells. The function has an
77 unweighted version. */
88 enum ctables_function_availability
90 CTFA_ALL, /* Any variables. */
91 CTFA_SCALE, /* Only scale variables, totals, and subtotals. */
92 //CTFA_MRSETS, /* Only multiple-response sets */
95 enum ctables_summary_function
97 #define S(ENUM, NAME, TYPE, FORMAT, AVAILABILITY) ENUM,
98 #include "ctables.inc"
103 #define S(ENUM, NAME, TYPE, FORMAT, AVAILABILITY) +1
105 #include "ctables.inc"
109 struct ctables_function_info
111 struct substring basename;
112 enum ctables_function_type type;
113 enum ctables_format format;
114 enum ctables_function_availability availability;
116 bool may_be_unweighted;
119 static const struct ctables_function_info ctables_function_info[N_CTSF_FUNCTIONS] = {
120 #define S(ENUM, NAME, TYPE, FORMAT, AVAILABILITY) \
122 .basename = SS_LITERAL_INITIALIZER (NAME), \
125 .availability = AVAILABILITY, \
126 .may_be_unweighted = (TYPE) == CTFT_UCELL || (TYPE) == CTFT_AREA, \
127 .is_area = (TYPE) == CTFT_AREA \
129 #include "ctables.inc"
133 static bool ctables_summary_function_is_count (enum ctables_summary_function);
135 enum ctables_area_type
137 /* Within a section, where stacked variables divide one section from
140 Keep CTAT_LAYER after CTAT_LAYERROW and CTAT_LAYERCOL so that
141 parse_ctables_summary_function() parses correctly. */
142 CTAT_TABLE, /* All layers of a whole section. */
143 CTAT_LAYERROW, /* Row in one layer within a section. */
144 CTAT_LAYERCOL, /* Column in one layer within a section. */
145 CTAT_LAYER, /* One layer within a section. */
147 /* Within a subtable, where a subtable pairs an innermost row variable with
148 an innermost column variable within a single layer. */
149 CTAT_SUBTABLE, /* Whole subtable. */
150 CTAT_ROW, /* Row within a subtable. */
151 CTAT_COL, /* Column within a subtable. */
155 static const char *ctables_area_type_name[N_CTATS] = {
156 [CTAT_TABLE] = "TABLE",
157 [CTAT_LAYER] = "LAYER",
158 [CTAT_LAYERROW] = "LAYERROW",
159 [CTAT_LAYERCOL] = "LAYERCOL",
160 [CTAT_SUBTABLE] = "SUBTABLE",
167 struct hmap_node node;
169 const struct ctables_cell *example;
172 double d_valid; /* Dictionary weight. */
175 double e_valid; /* Effective weight */
178 double u_valid; /* Unweighted. */
181 struct ctables_sum *sums;
190 enum ctables_summary_variant
199 /* In struct ctables_section's 'cells' hmap. Indexed by all the values in
200 all the axes (except the scalar variable, if any). */
201 struct hmap_node node;
203 /* The areas that contain this cell. */
205 struct ctables_area *areas[N_CTATS];
210 enum ctables_summary_variant sv;
212 struct ctables_cell_axis
214 struct ctables_cell_value
216 const struct ctables_category *category;
224 union ctables_summary *summaries;
231 const struct dictionary *dict;
232 struct pivot_table_look *look;
234 /* CTABLES has a number of extra formats that we implement via custom
235 currency specifications on an alternate fmt_settings. */
236 #define CTEF_NEGPAREN FMT_CCA
237 #define CTEF_NEQUAL FMT_CCB
238 #define CTEF_PAREN FMT_CCC
239 #define CTEF_PCTPAREN FMT_CCD
240 struct fmt_settings ctables_formats;
242 /* If this is NULL, zeros are displayed using the normal print format.
243 Otherwise, this string is displayed. */
246 /* If this is NULL, missing values are displayed using the normal print
247 format. Otherwise, this string is displayed. */
250 /* Indexed by variable dictionary index. */
251 enum ctables_vlabel *vlabels;
253 struct hmap postcomputes; /* Contains "struct ctables_postcompute"s. */
255 bool mrsets_count_duplicates; /* MRSETS. */
256 bool smissing_listwise; /* SMISSING. */
257 struct variable *e_weight; /* WEIGHT. */
258 int hide_threshold; /* HIDESMALLCOUNTS. */
260 struct ctables_table **tables;
264 static struct ctables_postcompute *ctables_find_postcompute (struct ctables *,
267 struct ctables_postcompute
269 struct hmap_node hmap_node; /* In struct ctables's 'pcompute' hmap. */
270 char *name; /* Name, without leading &. */
272 struct msg_location *location; /* Location of definition. */
273 struct ctables_pcexpr *expr;
275 struct ctables_summary_spec_set *specs;
276 bool hide_source_cats;
279 struct ctables_pcexpr
289 enum ctables_postcompute_op
292 CTPO_CONSTANT, /* 5 */
293 CTPO_CAT_NUMBER, /* [5] */
294 CTPO_CAT_STRING, /* ["STRING"] */
295 CTPO_CAT_NRANGE, /* [LO THRU 5] */
296 CTPO_CAT_SRANGE, /* ["A" THRU "B"] */
297 CTPO_CAT_MISSING, /* MISSING */
298 CTPO_CAT_OTHERNM, /* OTHERNM */
299 CTPO_CAT_SUBTOTAL, /* SUBTOTAL */
300 CTPO_CAT_TOTAL, /* TOTAL */
314 /* CTPO_CAT_NUMBER. */
317 /* CTPO_CAT_STRING, in dictionary encoding. */
318 struct substring string;
320 /* CTPO_CAT_NRANGE. */
323 /* CTPO_CAT_SRANGE. */
324 struct substring srange[2];
326 /* CTPO_CAT_SUBTOTAL. */
327 size_t subtotal_index;
329 /* Two elements: CTPO_ADD, CTPO_SUB, CTPO_MUL, CTPO_DIV, CTPO_POW.
330 One element: CTPO_NEG. */
331 struct ctables_pcexpr *subs[2];
334 /* Source location. */
335 struct msg_location *location;
338 static void ctables_pcexpr_destroy (struct ctables_pcexpr *);
339 static struct ctables_pcexpr *ctables_pcexpr_allocate_binary (
340 enum ctables_postcompute_op, struct ctables_pcexpr *sub0,
341 struct ctables_pcexpr *sub1);
343 struct ctables_summary_spec_set
345 struct ctables_summary_spec *specs;
349 /* The variable to which the summary specs are applied. */
350 struct variable *var;
352 /* Whether the variable to which the summary specs are applied is a scale
353 variable for the purpose of summarization.
355 (VALIDN and TOTALN act differently for summarizing scale and categorical
359 /* If any of these optional additional scale variables are missing, then
360 treat 'var' as if it's missing too. This is for implementing
361 SMISSING=LISTWISE. */
362 struct variable **listwise_vars;
363 size_t n_listwise_vars;
366 static void ctables_summary_spec_set_clone (struct ctables_summary_spec_set *,
367 const struct ctables_summary_spec_set *);
368 static void ctables_summary_spec_set_uninit (struct ctables_summary_spec_set *);
370 /* A nested sequence of variables, e.g. a > b > c. */
373 struct variable **vars;
376 size_t *areas[N_CTATS];
377 size_t n_areas[N_CTATS];
380 struct ctables_summary_spec_set specs[N_CSVS];
383 /* A stack of nestings, e.g. nest1 + nest2 + ... + nestN. */
386 struct ctables_nest *nests;
390 static void ctables_stack_uninit (struct ctables_stack *);
394 struct hmap_node node;
399 struct ctables_occurrence
401 struct hmap_node node;
405 struct ctables_section
408 struct ctables_table *table;
409 struct ctables_nest *nests[PIVOT_N_AXES];
412 struct hmap *occurrences[PIVOT_N_AXES]; /* "struct ctables_occurrence"s. */
413 struct hmap cells; /* Contains "struct ctables_cell"s. */
414 struct hmap areas[N_CTATS]; /* Contains "struct ctables_area"s. */
417 static void ctables_section_uninit (struct ctables_section *);
421 struct ctables *ctables;
422 struct ctables_axis *axes[PIVOT_N_AXES];
423 struct ctables_stack stacks[PIVOT_N_AXES];
424 struct ctables_section *sections;
426 enum pivot_axis_type summary_axis;
427 struct ctables_summary_spec_set summary_specs;
428 struct variable **sum_vars;
431 enum pivot_axis_type slabels_axis;
432 bool slabels_visible;
434 /* The innermost category labels for axis 'a' appear on axis label_axis[a].
436 Most commonly, label_axis[a] == a, and in particular we always have
437 label_axis{PIVOT_AXIS_LAYER] == PIVOT_AXIS_LAYER.
439 If ROWLABELS or COLLABELS is specified, then one of
440 label_axis[PIVOT_AXIS_ROW] or label_axis[PIVOT_AXIS_COLUMN] can be the
441 opposite axis or PIVOT_AXIS_LAYER. Only one of them will differ.
443 If any category labels are moved, then 'clabels_example' is one of the
444 variables being moved (and it is otherwise NULL). All of the variables
445 being moved have the same width, value labels, and categories, so this
446 example variable can be used to find those out.
448 The remaining members in this group are relevant only if category labels
451 'clabels_values_map' holds a "struct ctables_value" for all the values
452 that appear in all of the variables in the moved categories. It is
453 accumulated as the data is read. Once the data is fully read, its
454 sorted values are put into 'clabels_values' and 'n_clabels_values'.
456 enum pivot_axis_type label_axis[PIVOT_N_AXES];
457 enum pivot_axis_type clabels_from_axis;
458 const struct variable *clabels_example;
459 struct hmap clabels_values_map;
460 struct ctables_value **clabels_values;
461 size_t n_clabels_values;
463 /* Indexed by variable dictionary index. */
464 struct ctables_categories **categories;
473 struct ctables_chisq *chisq;
474 struct ctables_pairwise *pairwise;
477 struct ctables_categories
480 struct ctables_category *cats;
485 struct ctables_category
487 enum ctables_category_type
489 /* Explicit category lists. */
492 CCT_NRANGE, /* Numerical range. */
493 CCT_SRANGE, /* String range. */
498 /* Totals and subtotals. */
502 /* Implicit category lists. */
507 /* For contributing to TOTALN. */
508 CCT_EXCLUDED_MISSING,
512 struct ctables_category *subtotal;
518 double number; /* CCT_NUMBER. */
519 struct substring string; /* CCT_STRING, in dictionary encoding. */
520 double nrange[2]; /* CCT_NRANGE. */
521 struct substring srange[2]; /* CCT_SRANGE. */
525 char *total_label; /* CCT_SUBTOTAL, CCT_TOTAL. */
526 bool hide_subcategories; /* CCT_SUBTOTAL. */
529 /* CCT_POSTCOMPUTE. */
532 const struct ctables_postcompute *pc;
533 enum fmt_type parse_format;
536 /* CCT_VALUE, CCT_LABEL, CCT_FUNCTION. */
539 bool include_missing;
543 enum ctables_summary_function sort_function;
545 enum ctables_area_type area;
546 struct variable *sort_var;
551 /* Source location. This is null for CCT_TOTAL, CCT_VALUE, CCT_LABEL,
552 CCT_FUNCTION, CCT_EXCLUDED_MISSING. */
553 struct msg_location *location;
557 ctables_category_uninit (struct ctables_category *cat)
562 msg_location_destroy (cat->location);
569 case CCT_POSTCOMPUTE:
573 ss_dealloc (&cat->string);
577 ss_dealloc (&cat->srange[0]);
578 ss_dealloc (&cat->srange[1]);
583 free (cat->total_label);
591 case CCT_EXCLUDED_MISSING:
597 nullable_substring_equal (const struct substring *a,
598 const struct substring *b)
600 return !a->string ? !b->string : b->string && ss_equals (*a, *b);
604 ctables_category_equal (const struct ctables_category *a,
605 const struct ctables_category *b)
607 if (a->type != b->type)
613 return a->number == b->number;
616 return ss_equals (a->string, b->string);
619 return a->nrange[0] == b->nrange[0] && a->nrange[1] == b->nrange[1];
622 return (nullable_substring_equal (&a->srange[0], &b->srange[0])
623 && nullable_substring_equal (&a->srange[1], &b->srange[1]));
629 case CCT_POSTCOMPUTE:
630 return a->pc == b->pc;
634 return !strcmp (a->total_label, b->total_label);
639 return (a->include_missing == b->include_missing
640 && a->sort_ascending == b->sort_ascending
641 && a->sort_function == b->sort_function
642 && a->sort_var == b->sort_var
643 && a->percentile == b->percentile);
645 case CCT_EXCLUDED_MISSING:
653 ctables_categories_unref (struct ctables_categories *c)
658 assert (c->n_refs > 0);
662 for (size_t i = 0; i < c->n_cats; i++)
663 ctables_category_uninit (&c->cats[i]);
669 ctables_categories_equal (const struct ctables_categories *a,
670 const struct ctables_categories *b)
672 if (a->n_cats != b->n_cats || a->show_empty != b->show_empty)
675 for (size_t i = 0; i < a->n_cats; i++)
676 if (!ctables_category_equal (&a->cats[i], &b->cats[i]))
682 /* Chi-square test (SIGTEST). */
690 /* Pairwise comparison test (COMPARETEST). */
691 struct ctables_pairwise
693 enum { PROP, MEAN } type;
696 bool meansvariance_allcats;
698 enum { BONFERRONI = 1, BH } adjust;
722 struct variable *var;
724 struct ctables_summary_spec_set specs[N_CSVS];
728 struct ctables_axis *subs[2];
731 struct msg_location *loc;
734 static void ctables_axis_destroy (struct ctables_axis *);
736 struct ctables_summary_spec
738 /* The calculation to be performed.
740 'function' is the function to calculate. 'weighted' specifies whether
741 to use weighted or unweighted data (for functions that do not support a
742 choice, it must be true). 'calc_area' is the area over which the
743 calculation takes place (for functions that target only an individual
744 cell, it must be 0). For CTSF_PTILE only, 'percentile' is the
745 percentile between 0 and 100 (for other functions it must be 0). */
746 enum ctables_summary_function function;
748 enum ctables_area_type calc_area;
749 double percentile; /* CTSF_PTILE only. */
751 /* How to display the result of the calculation.
753 'label' is a user-specified label, NULL if the user didn't specify
756 'user_area' is usually the same as 'calc_area', but when category labels
757 are rotated from one axis to another it swaps rows and columns.
759 'format' is the format for displaying the output. If
760 'is_ctables_format' is true, then 'format.type' is one of the special
761 CTEF_* formats instead of the standard ones. */
763 enum ctables_area_type user_area;
764 struct fmt_spec format;
765 bool is_ctables_format; /* Is 'format' one of CTEF_*? */
772 ctables_summary_spec_clone (struct ctables_summary_spec *dst,
773 const struct ctables_summary_spec *src)
776 dst->label = xstrdup_if_nonnull (src->label);
780 ctables_summary_spec_uninit (struct ctables_summary_spec *s)
787 ctables_summary_spec_set_clone (struct ctables_summary_spec_set *dst,
788 const struct ctables_summary_spec_set *src)
790 struct ctables_summary_spec *specs
791 = (src->n ? xnmalloc (src->n, sizeof *specs) : NULL);
792 for (size_t i = 0; i < src->n; i++)
793 ctables_summary_spec_clone (&specs[i], &src->specs[i]);
795 *dst = (struct ctables_summary_spec_set) {
800 .is_scale = src->is_scale,
805 ctables_summary_spec_set_uninit (struct ctables_summary_spec_set *set)
807 for (size_t i = 0; i < set->n; i++)
808 ctables_summary_spec_uninit (&set->specs[i]);
809 free (set->listwise_vars);
814 parse_col_width (struct lexer *lexer, const char *name, double *width)
816 lex_match (lexer, T_EQUALS);
817 if (lex_match_id (lexer, "DEFAULT"))
819 else if (lex_force_num_range_closed (lexer, name, 0, DBL_MAX))
821 *width = lex_number (lexer);
831 parse_bool (struct lexer *lexer, bool *b)
833 if (lex_match_id (lexer, "NO"))
835 else if (lex_match_id (lexer, "YES"))
839 lex_error_expecting (lexer, "YES", "NO");
845 static enum ctables_function_availability
846 ctables_function_availability (enum ctables_summary_function f)
848 static enum ctables_function_availability availability[] = {
849 #define S(ENUM, NAME, TYPE, FORMAT, AVAILABILITY) [ENUM] = AVAILABILITY,
850 #include "ctables.inc"
854 return availability[f];
858 ctables_summary_function_is_count (enum ctables_summary_function f)
860 return f == CTSF_COUNT || f == CTSF_ECOUNT;
864 parse_ctables_summary_function (struct lexer *lexer,
865 enum ctables_summary_function *function,
867 enum ctables_area_type *area)
869 if (!lex_force_id (lexer))
872 struct substring name = lex_tokss (lexer);
873 *weighted = !(ss_match_byte (&name, 'U') || ss_match_byte (&name, 'u'));
875 bool has_area = false;
877 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
878 if (ss_match_string_case (&name, ss_cstr (ctables_area_type_name[at])))
883 if (ss_equals_case (name, ss_cstr ("PCT")))
885 /* Special case where .COUNT suffix is omitted. */
886 *function = CTSF_areaPCT_COUNT;
893 for (int f = 0; f < N_CTSF_FUNCTIONS; f++)
895 const struct ctables_function_info *cfi = &ctables_function_info[f];
896 if (ss_equals_case (cfi->basename, name))
899 if (!*weighted && !cfi->may_be_unweighted)
901 if (has_area != cfi->is_area)
909 lex_error (lexer, _("Expecting summary function name."));
914 ctables_axis_destroy (struct ctables_axis *axis)
922 for (size_t i = 0; i < N_CSVS; i++)
923 ctables_summary_spec_set_uninit (&axis->specs[i]);
928 ctables_axis_destroy (axis->subs[0]);
929 ctables_axis_destroy (axis->subs[1]);
932 msg_location_destroy (axis->loc);
936 static struct ctables_axis *
937 ctables_axis_new_nonterminal (enum ctables_axis_op op,
938 struct ctables_axis *sub0,
939 struct ctables_axis *sub1,
940 struct lexer *lexer, int start_ofs)
942 struct ctables_axis *axis = xmalloc (sizeof *axis);
943 *axis = (struct ctables_axis) {
945 .subs = { sub0, sub1 },
946 .loc = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1),
951 struct ctables_axis_parse_ctx
954 struct dictionary *dict;
956 struct ctables_table *t;
959 static struct fmt_spec
960 ctables_summary_default_format (enum ctables_summary_function function,
961 const struct variable *var)
963 static const enum ctables_format default_formats[] = {
964 #define S(ENUM, NAME, TYPE, FORMAT, AVAILABILITY) [ENUM] = FORMAT,
965 #include "ctables.inc"
968 switch (default_formats[function])
971 return (struct fmt_spec) { .type = FMT_F, .w = 40 };
974 return (struct fmt_spec) { .type = FMT_PCT, .w = 40, .d = 1 };
977 return *var_get_print_format (var);
985 ctables_summary_label__ (const struct ctables_summary_spec *spec)
987 bool w = spec->weighted;
988 enum ctables_area_type a = spec->user_area;
989 switch (spec->function)
992 return w ? N_("Count") : N_("Unweighted Count");
995 return N_("Adjusted Count");
997 case CTSF_areaPCT_COUNT:
1000 case CTAT_TABLE: return w ? N_("Table %") : N_("Unweighted Table %");
1001 case CTAT_LAYER: return w ? N_("Layer %") : N_("Unweighted Layer %");
1002 case CTAT_LAYERROW: return w ? N_("Layer Row %") : N_("Unweighted Layer Row %");
1003 case CTAT_LAYERCOL: return w ? N_("Layer Column %") : N_("Unweighted Layer Column %");
1004 case CTAT_SUBTABLE: return w ? N_("Subtable %") : N_("Unweighted Subtable %");
1005 case CTAT_ROW: return w ? N_("Row %") : N_("Unweighted Row %");
1006 case CTAT_COL: return w ? N_("Column %") : N_("Unweighted Column %");
1010 case CTSF_areaPCT_VALIDN:
1013 case CTAT_TABLE: return w ? N_("Table Valid N %") : N_("Unweighted Table Valid N %");
1014 case CTAT_LAYER: return w ? N_("Layer Valid N %") : N_("Unweighted Layer Valid N %");
1015 case CTAT_LAYERROW: return w ? N_("Layer Row Valid N %") : N_("Unweighted Layer Row Valid N %");
1016 case CTAT_LAYERCOL: return w ? N_("Layer Column Valid N %") : N_("Unweighted Layer Column Valid N %");
1017 case CTAT_SUBTABLE: return w ? N_("Subtable Valid N %") : N_("Unweighted Subtable Valid N %");
1018 case CTAT_ROW: return w ? N_("Row Valid N %") : N_("Unweighted Row Valid N %");
1019 case CTAT_COL: return w ? N_("Column Valid N %") : N_("Unweighted Column Valid N %");
1023 case CTSF_areaPCT_TOTALN:
1026 case CTAT_TABLE: return w ? N_("Table Total N %") : N_("Unweighted Table Total N %");
1027 case CTAT_LAYER: return w ? N_("Layer Total N %") : N_("Unweighted Layer Total N %");
1028 case CTAT_LAYERROW: return w ? N_("Layer Row Total N %") : N_("Unweighted Layer Row Total N %");
1029 case CTAT_LAYERCOL: return w ? N_("Layer Column Total N %") : N_("Unweighted Layer Column Total N %");
1030 case CTAT_SUBTABLE: return w ? N_("Subtable Total N %") : N_("Unweighted Subtable Total N %");
1031 case CTAT_ROW: return w ? N_("Row Total N %") : N_("Unweighted Row Total N %");
1032 case CTAT_COL: return w ? N_("Column Total N %") : N_("Unweighted Column Total N %");
1036 case CTSF_MAXIMUM: return N_("Maximum");
1037 case CTSF_MEAN: return w ? N_("Mean") : N_("Unweighted Mean");
1038 case CTSF_MEDIAN: return N_("Median");
1039 case CTSF_MINIMUM: return N_("Minimum");
1040 case CTSF_MISSING: return N_("Missing");
1041 case CTSF_MODE: return N_("Mode");
1042 case CTSF_PTILE: NOT_REACHED ();
1043 case CTSF_RANGE: return N_("Range");
1044 case CTSF_SEMEAN: return N_("Std Error of Mean");
1045 case CTSF_STDDEV: return N_("Std Deviation");
1046 case CTSF_SUM: return N_("Sum");
1047 case CTSF_TOTALN: return N_("Total N");
1048 case CTSF_ETOTALN: return N_("Adjusted Total N");
1049 case CTSF_VALIDN: return N_("Valid N");
1050 case CTSF_EVALIDN: return N_("Adjusted Valid N");
1051 case CTSF_VARIANCE: return N_("Variance");
1052 case CTSF_areaPCT_SUM:
1055 case CTAT_TABLE: return w ? N_("Table Sum %") : N_("Unweighted Table Sum %");
1056 case CTAT_LAYER: return w ? N_("Layer Sum %") : N_("Unweighted Layer Sum %");
1057 case CTAT_LAYERROW: return w ? N_("Layer Row Sum %") : N_("Unweighted Layer Row Sum %");
1058 case CTAT_LAYERCOL: return w ? N_("Layer Column Sum %") : N_("Unweighted Layer Column Sum %");
1059 case CTAT_SUBTABLE: return w ? N_("Subtable Sum %") : N_("Unweighted Subtable Sum %");
1060 case CTAT_ROW: return w ? N_("Row Sum %") : N_("Unweighted Row Sum %");
1061 case CTAT_COL: return w ? N_("Column Sum %") : N_("Unweighted Column Sum %");
1068 /* Don't bother translating these: they are for developers only. */
1069 case CTAT_TABLE: return "Table ID";
1070 case CTAT_LAYER: return "Layer ID";
1071 case CTAT_LAYERROW: return "Layer Row ID";
1072 case CTAT_LAYERCOL: return "Layer Column ID";
1073 case CTAT_SUBTABLE: return "Subtable ID";
1074 case CTAT_ROW: return "Row ID";
1075 case CTAT_COL: return "Column ID";
1083 static struct pivot_value *
1084 ctables_summary_label (const struct ctables_summary_spec *spec, double cilevel)
1088 if (spec->function == CTSF_PTILE)
1090 double p = spec->percentile;
1091 char *s = (spec->weighted
1092 ? xasprintf (_("Percentile %.2f"), p)
1093 : xasprintf (_("Unweighted Percentile %.2f"), p));
1094 return pivot_value_new_user_text_nocopy (s);
1097 return pivot_value_new_text (ctables_summary_label__ (spec));
1101 struct substring in = ss_cstr (spec->label);
1102 struct substring target = ss_cstr (")CILEVEL");
1104 struct string out = DS_EMPTY_INITIALIZER;
1107 size_t chunk = ss_find_substring (in, target);
1108 ds_put_substring (&out, ss_head (in, chunk));
1109 ss_advance (&in, chunk);
1111 return pivot_value_new_user_text_nocopy (ds_steal_cstr (&out));
1113 ss_advance (&in, target.length);
1114 ds_put_format (&out, "%g", cilevel);
1120 ctables_summary_function_name (enum ctables_summary_function function,
1122 enum ctables_area_type area,
1123 char *buffer, size_t bufsize)
1125 const struct ctables_function_info *cfi = &ctables_function_info[function];
1126 snprintf (buffer, bufsize, "%s%s%s",
1127 weighted ? "" : "U",
1128 cfi->is_area ? ctables_area_type_name[area] : "",
1129 cfi->basename.string);
1134 add_summary_spec (struct ctables_axis *axis,
1135 enum ctables_summary_function function, bool weighted,
1136 enum ctables_area_type area, double percentile,
1137 const char *label, const struct fmt_spec *format,
1138 bool is_ctables_format, const struct msg_location *loc,
1139 enum ctables_summary_variant sv)
1141 if (axis->op == CTAO_VAR)
1143 char function_name[128];
1144 ctables_summary_function_name (function, weighted, area,
1145 function_name, sizeof function_name);
1146 const char *var_name = var_get_name (axis->var);
1147 switch (ctables_function_availability (function))
1151 msg_at (SE, loc, _("Summary function %s applies only to multiple "
1152 "response sets."), function_name);
1153 msg_at (SN, axis->loc, _("'%s' is not a multiple response set."),
1159 if (!axis->scale && sv != CSV_TOTAL)
1162 _("Summary function %s applies only to scale variables."),
1164 msg_at (SN, axis->loc, _("'%s' is not a scale variable."),
1174 struct ctables_summary_spec_set *set = &axis->specs[sv];
1175 if (set->n >= set->allocated)
1176 set->specs = x2nrealloc (set->specs, &set->allocated,
1177 sizeof *set->specs);
1179 struct ctables_summary_spec *dst = &set->specs[set->n++];
1180 *dst = (struct ctables_summary_spec) {
1181 .function = function,
1182 .weighted = weighted,
1185 .percentile = percentile,
1186 .label = xstrdup_if_nonnull (label),
1187 .format = (format ? *format
1188 : ctables_summary_default_format (function, axis->var)),
1189 .is_ctables_format = is_ctables_format,
1195 for (size_t i = 0; i < 2; i++)
1196 if (!add_summary_spec (axis->subs[i], function, weighted, area,
1197 percentile, label, format, is_ctables_format,
1204 static struct ctables_axis *ctables_axis_parse_stack (
1205 struct ctables_axis_parse_ctx *);
1208 static struct ctables_axis *
1209 ctables_axis_parse_primary (struct ctables_axis_parse_ctx *ctx)
1211 if (lex_match (ctx->lexer, T_LPAREN))
1213 struct ctables_axis *sub = ctables_axis_parse_stack (ctx);
1214 if (!sub || !lex_force_match (ctx->lexer, T_RPAREN))
1216 ctables_axis_destroy (sub);
1222 if (!lex_force_id (ctx->lexer))
1225 int start_ofs = lex_ofs (ctx->lexer);
1226 struct variable *var = parse_variable (ctx->lexer, ctx->dict);
1230 struct ctables_axis *axis = xmalloc (sizeof *axis);
1231 *axis = (struct ctables_axis) { .op = CTAO_VAR, .var = var };
1233 axis->scale = (lex_match_phrase (ctx->lexer, "[S]") ? true
1234 : lex_match_phrase (ctx->lexer, "[C]") ? false
1235 : var_get_measure (var) == MEASURE_SCALE);
1236 axis->loc = lex_ofs_location (ctx->lexer, start_ofs,
1237 lex_ofs (ctx->lexer) - 1);
1238 if (axis->scale && var_is_alpha (var))
1240 msg_at (SE, axis->loc, _("Cannot use string variable %s as a scale "
1242 var_get_name (var));
1243 ctables_axis_destroy (axis);
1251 has_digit (const char *s)
1253 return s[strcspn (s, "0123456789")] != '\0';
1257 parse_ctables_format_specifier (struct lexer *lexer, struct fmt_spec *format,
1258 bool *is_ctables_format)
1260 char type[FMT_TYPE_LEN_MAX + 1];
1261 if (!parse_abstract_format_specifier__ (lexer, type, &format->w, &format->d))
1264 if (!strcasecmp (type, "NEGPAREN"))
1265 format->type = CTEF_NEGPAREN;
1266 else if (!strcasecmp (type, "NEQUAL"))
1267 format->type = CTEF_NEQUAL;
1268 else if (!strcasecmp (type, "PAREN"))
1269 format->type = CTEF_PAREN;
1270 else if (!strcasecmp (type, "PCTPAREN"))
1271 format->type = CTEF_PCTPAREN;
1274 *is_ctables_format = false;
1275 return (parse_format_specifier (lexer, format)
1276 && fmt_check_output (format)
1277 && fmt_check_type_compat (format, VAL_NUMERIC));
1283 lex_next_error (lexer, -1, -1,
1284 _("Output format %s requires width 2 or greater."), type);
1287 else if (format->d > format->w - 1)
1289 lex_next_error (lexer, -1, -1, _("Output format %s requires width "
1290 "greater than decimals."), type);
1295 *is_ctables_format = true;
1300 static struct ctables_axis *
1301 ctables_axis_parse_postfix (struct ctables_axis_parse_ctx *ctx)
1303 struct ctables_axis *sub = ctables_axis_parse_primary (ctx);
1304 if (!sub || !lex_match (ctx->lexer, T_LBRACK))
1307 enum ctables_summary_variant sv = CSV_CELL;
1310 int start_ofs = lex_ofs (ctx->lexer);
1312 /* Parse function. */
1313 enum ctables_summary_function function;
1315 enum ctables_area_type area;
1316 if (!parse_ctables_summary_function (ctx->lexer, &function, &weighted,
1320 /* Parse percentile. */
1321 double percentile = 0;
1322 if (function == CTSF_PTILE)
1324 if (!lex_force_num_range_closed (ctx->lexer, "PTILE", 0, 100))
1326 percentile = lex_number (ctx->lexer);
1327 lex_get (ctx->lexer);
1332 if (lex_is_string (ctx->lexer))
1334 label = ss_xstrdup (lex_tokss (ctx->lexer));
1335 lex_get (ctx->lexer);
1339 struct fmt_spec format;
1340 const struct fmt_spec *formatp;
1341 bool is_ctables_format = false;
1342 if (lex_token (ctx->lexer) == T_ID
1343 && has_digit (lex_tokcstr (ctx->lexer)))
1345 if (!parse_ctables_format_specifier (ctx->lexer, &format,
1346 &is_ctables_format))
1356 struct msg_location *loc = lex_ofs_location (ctx->lexer, start_ofs,
1357 lex_ofs (ctx->lexer) - 1);
1358 add_summary_spec (sub, function, weighted, area, percentile, label,
1359 formatp, is_ctables_format, loc, sv);
1361 msg_location_destroy (loc);
1363 lex_match (ctx->lexer, T_COMMA);
1364 if (sv == CSV_CELL && lex_match_id (ctx->lexer, "TOTALS"))
1366 if (!lex_force_match (ctx->lexer, T_LBRACK))
1370 else if (lex_match (ctx->lexer, T_RBRACK))
1372 if (sv == CSV_TOTAL && !lex_force_match (ctx->lexer, T_RBRACK))
1379 ctables_axis_destroy (sub);
1383 static const struct ctables_axis *
1384 find_scale (const struct ctables_axis *axis)
1388 else if (axis->op == CTAO_VAR)
1389 return axis->scale ? axis : NULL;
1392 for (size_t i = 0; i < 2; i++)
1394 const struct ctables_axis *scale = find_scale (axis->subs[i]);
1402 static const struct ctables_axis *
1403 find_categorical_summary_spec (const struct ctables_axis *axis)
1407 else if (axis->op == CTAO_VAR)
1408 return !axis->scale && axis->specs[CSV_CELL].n ? axis : NULL;
1411 for (size_t i = 0; i < 2; i++)
1413 const struct ctables_axis *sum
1414 = find_categorical_summary_spec (axis->subs[i]);
1422 static struct ctables_axis *
1423 ctables_axis_parse_nest (struct ctables_axis_parse_ctx *ctx)
1425 int start_ofs = lex_ofs (ctx->lexer);
1426 struct ctables_axis *lhs = ctables_axis_parse_postfix (ctx);
1430 while (lex_match (ctx->lexer, T_GT))
1432 struct ctables_axis *rhs = ctables_axis_parse_postfix (ctx);
1435 ctables_axis_destroy (lhs);
1439 struct ctables_axis *nest = ctables_axis_new_nonterminal (
1440 CTAO_NEST, lhs, rhs, ctx->lexer, start_ofs);
1442 const struct ctables_axis *outer_scale = find_scale (lhs);
1443 const struct ctables_axis *inner_scale = find_scale (rhs);
1444 if (outer_scale && inner_scale)
1446 msg_at (SE, nest->loc, _("Cannot nest scale variables."));
1447 msg_at (SN, outer_scale->loc, _("This is an outer scale variable."));
1448 msg_at (SN, inner_scale->loc, _("This is an inner scale variable."));
1449 ctables_axis_destroy (nest);
1453 const struct ctables_axis *outer_sum = find_categorical_summary_spec (lhs);
1456 msg_at (SE, nest->loc,
1457 _("Summaries may only be requested for categorical variables "
1458 "at the innermost nesting level."));
1459 msg_at (SN, outer_sum->loc,
1460 _("This outer categorical variable has a summary."));
1461 ctables_axis_destroy (nest);
1471 static struct ctables_axis *
1472 ctables_axis_parse_stack (struct ctables_axis_parse_ctx *ctx)
1474 int start_ofs = lex_ofs (ctx->lexer);
1475 struct ctables_axis *lhs = ctables_axis_parse_nest (ctx);
1479 while (lex_match (ctx->lexer, T_PLUS))
1481 struct ctables_axis *rhs = ctables_axis_parse_nest (ctx);
1484 ctables_axis_destroy (lhs);
1488 lhs = ctables_axis_new_nonterminal (CTAO_STACK, lhs, rhs,
1489 ctx->lexer, start_ofs);
1496 ctables_axis_parse (struct lexer *lexer, struct dictionary *dict,
1497 struct ctables *ct, struct ctables_table *t,
1498 enum pivot_axis_type a)
1500 if (lex_token (lexer) == T_BY
1501 || lex_token (lexer) == T_SLASH
1502 || lex_token (lexer) == T_ENDCMD)
1505 struct ctables_axis_parse_ctx ctx = {
1511 t->axes[a] = ctables_axis_parse_stack (&ctx);
1512 return t->axes[a] != NULL;
1516 ctables_chisq_destroy (struct ctables_chisq *chisq)
1522 ctables_pairwise_destroy (struct ctables_pairwise *pairwise)
1528 ctables_table_destroy (struct ctables_table *t)
1533 for (size_t i = 0; i < t->n_sections; i++)
1534 ctables_section_uninit (&t->sections[i]);
1537 for (size_t i = 0; i < t->n_categories; i++)
1538 ctables_categories_unref (t->categories[i]);
1539 free (t->categories);
1541 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
1543 ctables_axis_destroy (t->axes[a]);
1544 ctables_stack_uninit (&t->stacks[a]);
1546 free (t->summary_specs.specs);
1548 struct ctables_value *ctv, *next_ctv;
1549 HMAP_FOR_EACH_SAFE (ctv, next_ctv, struct ctables_value, node,
1550 &t->clabels_values_map)
1552 value_destroy (&ctv->value, var_get_width (t->clabels_example));
1553 hmap_delete (&t->clabels_values_map, &ctv->node);
1556 hmap_destroy (&t->clabels_values_map);
1557 free (t->clabels_values);
1563 ctables_chisq_destroy (t->chisq);
1564 ctables_pairwise_destroy (t->pairwise);
1569 ctables_destroy (struct ctables *ct)
1574 struct ctables_postcompute *pc, *next_pc;
1575 HMAP_FOR_EACH_SAFE (pc, next_pc, struct ctables_postcompute, hmap_node,
1579 msg_location_destroy (pc->location);
1580 ctables_pcexpr_destroy (pc->expr);
1584 ctables_summary_spec_set_uninit (pc->specs);
1587 hmap_delete (&ct->postcomputes, &pc->hmap_node);
1590 hmap_destroy (&ct->postcomputes);
1592 fmt_settings_uninit (&ct->ctables_formats);
1593 pivot_table_look_unref (ct->look);
1597 for (size_t i = 0; i < ct->n_tables; i++)
1598 ctables_table_destroy (ct->tables[i]);
1603 static struct ctables_category
1604 cct_nrange (double low, double high)
1606 return (struct ctables_category) {
1608 .nrange = { low, high }
1612 static struct ctables_category
1613 cct_srange (struct substring low, struct substring high)
1615 return (struct ctables_category) {
1617 .srange = { low, high }
1622 ctables_table_parse_subtotal (struct lexer *lexer, bool hide_subcategories,
1623 struct ctables_category *cat)
1626 if (lex_match (lexer, T_EQUALS))
1628 if (!lex_force_string (lexer))
1631 total_label = ss_xstrdup (lex_tokss (lexer));
1635 total_label = xstrdup (_("Subtotal"));
1637 *cat = (struct ctables_category) {
1638 .type = CCT_SUBTOTAL,
1639 .hide_subcategories = hide_subcategories,
1640 .total_label = total_label
1645 static struct substring
1646 parse_substring (struct lexer *lexer, struct dictionary *dict)
1648 struct substring s = recode_substring_pool (
1649 dict_get_encoding (dict), "UTF-8", lex_tokss (lexer), NULL);
1650 ss_rtrim (&s, ss_cstr (" "));
1656 ctables_table_parse_explicit_category (struct lexer *lexer,
1657 struct dictionary *dict,
1659 struct ctables_category *cat)
1661 if (lex_match_id (lexer, "OTHERNM"))
1662 *cat = (struct ctables_category) { .type = CCT_OTHERNM };
1663 else if (lex_match_id (lexer, "MISSING"))
1664 *cat = (struct ctables_category) { .type = CCT_MISSING };
1665 else if (lex_match_id (lexer, "SUBTOTAL"))
1666 return ctables_table_parse_subtotal (lexer, false, cat);
1667 else if (lex_match_id (lexer, "HSUBTOTAL"))
1668 return ctables_table_parse_subtotal (lexer, true, cat);
1669 else if (lex_match_id (lexer, "LO"))
1671 if (!lex_force_match_id (lexer, "THRU"))
1673 if (lex_is_string (lexer))
1675 struct substring sr0 = { .string = NULL };
1676 struct substring sr1 = parse_substring (lexer, dict);
1677 *cat = cct_srange (sr0, sr1);
1679 else if (lex_force_num (lexer))
1681 *cat = cct_nrange (-DBL_MAX, lex_number (lexer));
1687 else if (lex_is_number (lexer))
1689 double number = lex_number (lexer);
1691 if (lex_match_id (lexer, "THRU"))
1693 if (lex_match_id (lexer, "HI"))
1694 *cat = cct_nrange (number, DBL_MAX);
1697 if (!lex_force_num (lexer))
1699 *cat = cct_nrange (number, lex_number (lexer));
1704 *cat = (struct ctables_category) {
1709 else if (lex_is_string (lexer))
1711 struct substring s = parse_substring (lexer, dict);
1712 if (lex_match_id (lexer, "THRU"))
1714 if (lex_match_id (lexer, "HI"))
1716 struct substring sr1 = { .string = NULL };
1717 *cat = cct_srange (s, sr1);
1721 if (!lex_force_string (lexer))
1726 struct substring sr1 = parse_substring (lexer, dict);
1727 *cat = cct_srange (s, sr1);
1731 *cat = (struct ctables_category) { .type = CCT_STRING, .string = s };
1733 else if (lex_match (lexer, T_AND))
1735 if (!lex_force_id (lexer))
1737 struct ctables_postcompute *pc = ctables_find_postcompute (
1738 ct, lex_tokcstr (lexer));
1741 struct msg_location *loc = lex_get_location (lexer, -1, 0);
1742 msg_at (SE, loc, _("Unknown postcompute &%s."),
1743 lex_tokcstr (lexer));
1744 msg_location_destroy (loc);
1749 *cat = (struct ctables_category) { .type = CCT_POSTCOMPUTE, .pc = pc };
1753 lex_error (lexer, NULL);
1761 parse_category_string (struct msg_location *location,
1762 struct substring s, const struct dictionary *dict,
1763 enum fmt_type format, double *n)
1766 char *error = data_in (s, dict_get_encoding (dict), format,
1767 settings_get_fmt_settings (), &v, 0, NULL);
1770 msg_at (SE, location,
1771 _("Failed to parse category specification as format %s: %s."),
1772 fmt_name (format), error);
1781 static struct ctables_category *
1782 ctables_find_category_for_postcompute__ (const struct ctables_categories *cats,
1783 const struct ctables_pcexpr *e)
1785 struct ctables_category *best = NULL;
1786 size_t n_subtotals = 0;
1787 for (size_t i = 0; i < cats->n_cats; i++)
1789 struct ctables_category *cat = &cats->cats[i];
1792 case CTPO_CAT_NUMBER:
1793 if (cat->type == CCT_NUMBER && cat->number == e->number)
1797 case CTPO_CAT_STRING:
1798 if (cat->type == CCT_STRING && ss_equals (cat->string, e->string))
1802 case CTPO_CAT_NRANGE:
1803 if (cat->type == CCT_NRANGE
1804 && cat->nrange[0] == e->nrange[0]
1805 && cat->nrange[1] == e->nrange[1])
1809 case CTPO_CAT_SRANGE:
1810 if (cat->type == CCT_SRANGE
1811 && nullable_substring_equal (&cat->srange[0], &e->srange[0])
1812 && nullable_substring_equal (&cat->srange[1], &e->srange[1]))
1816 case CTPO_CAT_MISSING:
1817 if (cat->type == CCT_MISSING)
1821 case CTPO_CAT_OTHERNM:
1822 if (cat->type == CCT_OTHERNM)
1826 case CTPO_CAT_SUBTOTAL:
1827 if (cat->type == CCT_SUBTOTAL)
1830 if (e->subtotal_index == n_subtotals)
1832 else if (e->subtotal_index == 0)
1837 case CTPO_CAT_TOTAL:
1838 if (cat->type == CCT_TOTAL)
1852 if (e->op == CTPO_CAT_SUBTOTAL && e->subtotal_index == 0 && n_subtotals > 1)
1857 static struct ctables_category *
1858 ctables_find_category_for_postcompute (const struct dictionary *dict,
1859 const struct ctables_categories *cats,
1860 enum fmt_type parse_format,
1861 const struct ctables_pcexpr *e)
1863 if (parse_format != FMT_F)
1865 if (e->op == CTPO_CAT_STRING)
1868 if (!parse_category_string (e->location, e->string, dict,
1869 parse_format, &number))
1872 struct ctables_pcexpr e2 = {
1873 .op = CTPO_CAT_NUMBER,
1875 .location = e->location,
1877 return ctables_find_category_for_postcompute__ (cats, &e2);
1879 else if (e->op == CTPO_CAT_SRANGE)
1882 if (!e->srange[0].string)
1883 nrange[0] = -DBL_MAX;
1884 else if (!parse_category_string (e->location, e->srange[0], dict,
1885 parse_format, &nrange[0]))
1888 if (!e->srange[1].string)
1889 nrange[1] = DBL_MAX;
1890 else if (!parse_category_string (e->location, e->srange[1], dict,
1891 parse_format, &nrange[1]))
1894 struct ctables_pcexpr e2 = {
1895 .op = CTPO_CAT_NRANGE,
1896 .nrange = { nrange[0], nrange[1] },
1897 .location = e->location,
1899 return ctables_find_category_for_postcompute__ (cats, &e2);
1902 return ctables_find_category_for_postcompute__ (cats, e);
1906 ctables_recursive_check_postcompute (struct dictionary *dict,
1907 const struct ctables_pcexpr *e,
1908 struct ctables_category *pc_cat,
1909 const struct ctables_categories *cats,
1910 const struct msg_location *cats_location)
1914 case CTPO_CAT_NUMBER:
1915 case CTPO_CAT_STRING:
1916 case CTPO_CAT_NRANGE:
1917 case CTPO_CAT_SRANGE:
1918 case CTPO_CAT_MISSING:
1919 case CTPO_CAT_OTHERNM:
1920 case CTPO_CAT_SUBTOTAL:
1921 case CTPO_CAT_TOTAL:
1923 struct ctables_category *cat = ctables_find_category_for_postcompute (
1924 dict, cats, pc_cat->parse_format, e);
1927 if (e->op == CTPO_CAT_SUBTOTAL && e->subtotal_index == 0)
1929 size_t n_subtotals = 0;
1930 for (size_t i = 0; i < cats->n_cats; i++)
1931 n_subtotals += cats->cats[i].type == CCT_SUBTOTAL;
1932 if (n_subtotals > 1)
1934 msg_at (SE, cats_location,
1935 ngettext ("These categories include %zu instance "
1936 "of SUBTOTAL or HSUBTOTAL, so references "
1937 "from computed categories must refer to "
1938 "subtotals by position, "
1939 "e.g. SUBTOTAL[1].",
1940 "These categories include %zu instances "
1941 "of SUBTOTAL or HSUBTOTAL, so references "
1942 "from computed categories must refer to "
1943 "subtotals by position, "
1944 "e.g. SUBTOTAL[1].",
1947 msg_at (SN, e->location,
1948 _("This is the reference that lacks a position."));
1953 msg_at (SE, pc_cat->location,
1954 _("Computed category &%s references a category not included "
1955 "in the category list."),
1957 msg_at (SN, e->location, _("This is the missing category."));
1958 if (e->op == CTPO_CAT_SUBTOTAL)
1959 msg_at (SN, cats_location,
1960 _("To fix the problem, add subtotals to the "
1961 "list of categories here."));
1962 else if (e->op == CTPO_CAT_TOTAL)
1963 msg (SN, _("To fix the problem, add TOTAL=YES to the variable's "
1964 "CATEGORIES specification."));
1966 msg_at (SN, cats_location,
1967 _("To fix the problem, add the missing category to the "
1968 "list of categories here."));
1971 if (pc_cat->pc->hide_source_cats)
1985 for (size_t i = 0; i < 2; i++)
1986 if (e->subs[i] && !ctables_recursive_check_postcompute (
1987 dict, e->subs[i], pc_cat, cats, cats_location))
1996 all_strings (struct variable **vars, size_t n_vars,
1997 const struct ctables_category *cat)
1999 for (size_t j = 0; j < n_vars; j++)
2000 if (var_is_numeric (vars[j]))
2002 msg_at (SE, cat->location,
2003 _("This category specification may be applied only to string "
2004 "variables, but this subcommand tries to apply it to "
2005 "numeric variable %s."),
2006 var_get_name (vars[j]));
2013 ctables_table_parse_categories (struct lexer *lexer, struct dictionary *dict,
2014 struct ctables *ct, struct ctables_table *t)
2016 if (!lex_match_id (lexer, "VARIABLES"))
2018 lex_match (lexer, T_EQUALS);
2020 struct variable **vars;
2022 if (!parse_variables (lexer, dict, &vars, &n_vars, PV_NO_SCRATCH))
2025 const struct fmt_spec *common_format = var_get_print_format (vars[0]);
2026 for (size_t i = 1; i < n_vars; i++)
2028 const struct fmt_spec *f = var_get_print_format (vars[i]);
2029 if (f->type != common_format->type)
2031 common_format = NULL;
2037 && (fmt_get_category (common_format->type)
2038 & (FMT_CAT_DATE | FMT_CAT_TIME | FMT_CAT_DATE_COMPONENT)));
2040 struct ctables_categories *c = xmalloc (sizeof *c);
2041 *c = (struct ctables_categories) { .n_refs = n_vars, .show_empty = true };
2042 for (size_t i = 0; i < n_vars; i++)
2044 struct ctables_categories **cp
2045 = &t->categories[var_get_dict_index (vars[i])];
2046 ctables_categories_unref (*cp);
2050 size_t allocated_cats = 0;
2051 int cats_start_ofs = -1;
2052 int cats_end_ofs = -1;
2053 if (lex_match (lexer, T_LBRACK))
2055 cats_start_ofs = lex_ofs (lexer);
2058 if (c->n_cats >= allocated_cats)
2059 c->cats = x2nrealloc (c->cats, &allocated_cats, sizeof *c->cats);
2061 int start_ofs = lex_ofs (lexer);
2062 struct ctables_category *cat = &c->cats[c->n_cats];
2063 if (!ctables_table_parse_explicit_category (lexer, dict, ct, cat))
2065 cat->location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1);
2068 lex_match (lexer, T_COMMA);
2070 while (!lex_match (lexer, T_RBRACK));
2071 cats_end_ofs = lex_ofs (lexer) - 1;
2074 struct ctables_category cat = {
2076 .include_missing = false,
2077 .sort_ascending = true,
2079 bool show_totals = false;
2080 char *total_label = NULL;
2081 bool totals_before = false;
2082 while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
2084 if (!c->n_cats && lex_match_id (lexer, "ORDER"))
2086 lex_match (lexer, T_EQUALS);
2087 if (lex_match_id (lexer, "A"))
2088 cat.sort_ascending = true;
2089 else if (lex_match_id (lexer, "D"))
2090 cat.sort_ascending = false;
2093 lex_error_expecting (lexer, "A", "D");
2097 else if (!c->n_cats && lex_match_id (lexer, "KEY"))
2099 lex_match (lexer, T_EQUALS);
2100 if (lex_match_id (lexer, "VALUE"))
2101 cat.type = CCT_VALUE;
2102 else if (lex_match_id (lexer, "LABEL"))
2103 cat.type = CCT_LABEL;
2106 cat.type = CCT_FUNCTION;
2107 if (!parse_ctables_summary_function (lexer, &cat.sort_function,
2108 &cat.weighted, &cat.area))
2111 if (lex_match (lexer, T_LPAREN))
2113 cat.sort_var = parse_variable (lexer, dict);
2117 if (cat.sort_function == CTSF_PTILE)
2119 lex_match (lexer, T_COMMA);
2120 if (!lex_force_num_range_closed (lexer, "PTILE", 0, 100))
2122 cat.percentile = lex_number (lexer);
2126 if (!lex_force_match (lexer, T_RPAREN))
2129 else if (ctables_function_availability (cat.sort_function)
2132 bool UNUSED b = lex_force_match (lexer, T_LPAREN);
2137 else if (!c->n_cats && lex_match_id (lexer, "MISSING"))
2139 lex_match (lexer, T_EQUALS);
2140 if (lex_match_id (lexer, "INCLUDE"))
2141 cat.include_missing = true;
2142 else if (lex_match_id (lexer, "EXCLUDE"))
2143 cat.include_missing = false;
2146 lex_error_expecting (lexer, "INCLUDE", "EXCLUDE");
2150 else if (lex_match_id (lexer, "TOTAL"))
2152 lex_match (lexer, T_EQUALS);
2153 if (!parse_bool (lexer, &show_totals))
2156 else if (lex_match_id (lexer, "LABEL"))
2158 lex_match (lexer, T_EQUALS);
2159 if (!lex_force_string (lexer))
2162 total_label = ss_xstrdup (lex_tokss (lexer));
2165 else if (lex_match_id (lexer, "POSITION"))
2167 lex_match (lexer, T_EQUALS);
2168 if (lex_match_id (lexer, "BEFORE"))
2169 totals_before = true;
2170 else if (lex_match_id (lexer, "AFTER"))
2171 totals_before = false;
2174 lex_error_expecting (lexer, "BEFORE", "AFTER");
2178 else if (lex_match_id (lexer, "EMPTY"))
2180 lex_match (lexer, T_EQUALS);
2181 if (lex_match_id (lexer, "INCLUDE"))
2182 c->show_empty = true;
2183 else if (lex_match_id (lexer, "EXCLUDE"))
2184 c->show_empty = false;
2187 lex_error_expecting (lexer, "INCLUDE", "EXCLUDE");
2194 lex_error_expecting (lexer, "ORDER", "KEY", "MISSING",
2195 "TOTAL", "LABEL", "POSITION", "EMPTY");
2197 lex_error_expecting (lexer, "TOTAL", "LABEL", "POSITION", "EMPTY");
2204 if (c->n_cats >= allocated_cats)
2205 c->cats = x2nrealloc (c->cats, &allocated_cats, sizeof *c->cats);
2206 c->cats[c->n_cats++] = cat;
2211 if (c->n_cats >= allocated_cats)
2212 c->cats = x2nrealloc (c->cats, &allocated_cats, sizeof *c->cats);
2214 struct ctables_category *totals;
2217 insert_element (c->cats, c->n_cats, sizeof *c->cats, 0);
2218 totals = &c->cats[0];
2221 totals = &c->cats[c->n_cats];
2224 *totals = (struct ctables_category) {
2226 .total_label = total_label ? total_label : xstrdup (_("Total")),
2230 struct ctables_category *subtotal = NULL;
2231 for (size_t i = totals_before ? 0 : c->n_cats;
2232 totals_before ? i < c->n_cats : i-- > 0;
2233 totals_before ? i++ : 0)
2235 struct ctables_category *cat = &c->cats[i];
2244 cat->subtotal = subtotal;
2247 case CCT_POSTCOMPUTE:
2258 case CCT_EXCLUDED_MISSING:
2263 if (cats_start_ofs != -1)
2265 for (size_t i = 0; i < c->n_cats; i++)
2267 struct ctables_category *cat = &c->cats[i];
2270 case CCT_POSTCOMPUTE:
2271 cat->parse_format = parse_strings ? common_format->type : FMT_F;
2272 struct msg_location *cats_location
2273 = lex_ofs_location (lexer, cats_start_ofs, cats_end_ofs);
2274 bool ok = ctables_recursive_check_postcompute (
2275 dict, cat->pc->expr, cat, c, cats_location);
2276 msg_location_destroy (cats_location);
2283 for (size_t j = 0; j < n_vars; j++)
2284 if (var_is_alpha (vars[j]))
2286 msg_at (SE, cat->location,
2287 _("This category specification may be applied "
2288 "only to numeric variables, but this "
2289 "subcommand tries to apply it to string "
2291 var_get_name (vars[j]));
2300 if (!parse_category_string (cat->location, cat->string, dict,
2301 common_format->type, &n))
2304 ss_dealloc (&cat->string);
2306 cat->type = CCT_NUMBER;
2309 else if (!all_strings (vars, n_vars, cat))
2318 if (!cat->srange[0].string)
2320 else if (!parse_category_string (cat->location,
2321 cat->srange[0], dict,
2322 common_format->type, &n[0]))
2325 if (!cat->srange[1].string)
2327 else if (!parse_category_string (cat->location,
2328 cat->srange[1], dict,
2329 common_format->type, &n[1]))
2332 ss_dealloc (&cat->srange[0]);
2333 ss_dealloc (&cat->srange[1]);
2335 cat->type = CCT_NRANGE;
2336 cat->nrange[0] = n[0];
2337 cat->nrange[1] = n[1];
2339 else if (!all_strings (vars, n_vars, cat))
2350 case CCT_EXCLUDED_MISSING:
2365 ctables_nest_uninit (struct ctables_nest *nest)
2368 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
2369 ctables_summary_spec_set_uninit (&nest->specs[sv]);
2370 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
2371 free (nest->areas[at]);
2375 ctables_stack_uninit (struct ctables_stack *stack)
2379 for (size_t i = 0; i < stack->n; i++)
2380 ctables_nest_uninit (&stack->nests[i]);
2381 free (stack->nests);
2385 static struct ctables_stack
2386 nest_fts (struct ctables_stack s0, struct ctables_stack s1)
2393 struct ctables_stack stack = { .nests = xnmalloc (s0.n, s1.n * sizeof *stack.nests) };
2394 for (size_t i = 0; i < s0.n; i++)
2395 for (size_t j = 0; j < s1.n; j++)
2397 const struct ctables_nest *a = &s0.nests[i];
2398 const struct ctables_nest *b = &s1.nests[j];
2400 size_t allocate = a->n + b->n;
2401 struct variable **vars = xnmalloc (allocate, sizeof *vars);
2403 for (size_t k = 0; k < a->n; k++)
2404 vars[n++] = a->vars[k];
2405 for (size_t k = 0; k < b->n; k++)
2406 vars[n++] = b->vars[k];
2407 assert (n == allocate);
2409 const struct ctables_nest *summary_src;
2410 if (!a->specs[CSV_CELL].var)
2412 else if (!b->specs[CSV_CELL].var)
2417 struct ctables_nest *new = &stack.nests[stack.n++];
2418 *new = (struct ctables_nest) {
2420 .scale_idx = (a->scale_idx != SIZE_MAX ? a->scale_idx
2421 : b->scale_idx != SIZE_MAX ? a->n + b->scale_idx
2425 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
2426 ctables_summary_spec_set_clone (&new->specs[sv], &summary_src->specs[sv]);
2428 ctables_stack_uninit (&s0);
2429 ctables_stack_uninit (&s1);
2433 static struct ctables_stack
2434 stack_fts (struct ctables_stack s0, struct ctables_stack s1)
2436 struct ctables_stack stack = { .nests = xnmalloc (s0.n + s1.n, sizeof *stack.nests) };
2437 for (size_t i = 0; i < s0.n; i++)
2438 stack.nests[stack.n++] = s0.nests[i];
2439 for (size_t i = 0; i < s1.n; i++)
2441 stack.nests[stack.n] = s1.nests[i];
2442 stack.nests[stack.n].group_head += s0.n;
2445 assert (stack.n == s0.n + s1.n);
2451 static struct ctables_stack
2452 var_fts (const struct ctables_axis *a)
2454 struct variable **vars = xmalloc (sizeof *vars);
2457 struct ctables_nest *nest = xmalloc (sizeof *nest);
2458 *nest = (struct ctables_nest) {
2461 .scale_idx = a->scale ? 0 : SIZE_MAX,
2463 if (a->specs[CSV_CELL].n || a->scale)
2464 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
2466 ctables_summary_spec_set_clone (&nest->specs[sv], &a->specs[sv]);
2467 nest->specs[sv].var = a->var;
2468 nest->specs[sv].is_scale = a->scale;
2470 return (struct ctables_stack) { .nests = nest, .n = 1 };
2473 static struct ctables_stack
2474 enumerate_fts (enum pivot_axis_type axis_type, const struct ctables_axis *a)
2477 return (struct ctables_stack) { .n = 0 };
2485 return stack_fts (enumerate_fts (axis_type, a->subs[0]),
2486 enumerate_fts (axis_type, a->subs[1]));
2489 /* This should consider any of the scale variables found in the result to
2490 be linked to each other listwise for SMISSING=LISTWISE. */
2491 return nest_fts (enumerate_fts (axis_type, a->subs[0]),
2492 enumerate_fts (axis_type, a->subs[1]));
2498 union ctables_summary
2500 /* COUNT, VALIDN, TOTALN. */
2503 /* MINIMUM, MAXIMUM, RANGE. */
2510 /* MEAN, SEMEAN, STDDEV, SUM, VARIANCE, *.SUM. */
2511 struct moments1 *moments;
2513 /* MEDIAN, MODE, PTILE. */
2516 struct casewriter *writer;
2521 /* XXX multiple response */
2525 ctables_summary_init (union ctables_summary *s,
2526 const struct ctables_summary_spec *ss)
2528 switch (ss->function)
2532 case CTSF_areaPCT_COUNT:
2533 case CTSF_areaPCT_VALIDN:
2534 case CTSF_areaPCT_TOTALN:
2549 s->min = s->max = SYSMIS;
2557 case CTSF_areaPCT_SUM:
2558 s->moments = moments1_create (MOMENT_VARIANCE);
2565 struct caseproto *proto = caseproto_create ();
2566 proto = caseproto_add_width (proto, 0);
2567 proto = caseproto_add_width (proto, 0);
2569 struct subcase ordering;
2570 subcase_init (&ordering, 0, 0, SC_ASCEND);
2571 s->writer = sort_create_writer (&ordering, proto);
2572 subcase_uninit (&ordering);
2573 caseproto_unref (proto);
2583 ctables_summary_uninit (union ctables_summary *s,
2584 const struct ctables_summary_spec *ss)
2586 switch (ss->function)
2590 case CTSF_areaPCT_COUNT:
2591 case CTSF_areaPCT_VALIDN:
2592 case CTSF_areaPCT_TOTALN:
2613 case CTSF_areaPCT_SUM:
2614 moments1_destroy (s->moments);
2620 casewriter_destroy (s->writer);
2626 ctables_summary_add (union ctables_summary *s,
2627 const struct ctables_summary_spec *ss,
2628 const struct variable *var, const union value *value,
2629 bool is_scale, bool is_scale_missing,
2630 bool is_missing, bool excluded_missing,
2631 double d_weight, double e_weight)
2633 /* To determine whether a case is included in a given table for a particular
2634 kind of summary, consider the following charts for each variable in the
2635 table. Only if "yes" appears for every variable for the summary is the
2638 Categorical variables: VALIDN COUNT TOTALN
2639 Valid values in included categories yes yes yes
2640 Missing values in included categories --- yes yes
2641 Missing values in excluded categories --- --- yes
2642 Valid values in excluded categories --- --- ---
2644 Scale variables: VALIDN COUNT TOTALN
2645 Valid value yes yes yes
2646 Missing value --- yes yes
2648 Missing values include both user- and system-missing. (The system-missing
2649 value is always in an excluded category.)
2651 switch (ss->function)
2654 case CTSF_areaPCT_TOTALN:
2655 s->count += ss->weighted ? d_weight : 1.0;
2659 case CTSF_areaPCT_COUNT:
2660 if (is_scale || !excluded_missing)
2661 s->count += ss->weighted ? d_weight : 1.0;
2665 case CTSF_areaPCT_VALIDN:
2669 s->count += ss->weighted ? d_weight : 1.0;
2679 s->count += ss->weighted ? d_weight : 1.0;
2683 if (is_scale || !excluded_missing)
2684 s->count += e_weight;
2691 s->count += e_weight;
2695 s->count += e_weight;
2701 if (!is_scale_missing)
2703 assert (!var_is_alpha (var)); /* XXX? */
2704 if (s->min == SYSMIS || value->f < s->min)
2706 if (s->max == SYSMIS || value->f > s->max)
2716 case CTSF_areaPCT_SUM:
2717 if (!is_scale_missing)
2718 moments1_add (s->moments, value->f, ss->weighted ? e_weight : 1.0);
2724 if (!is_scale_missing)
2726 double w = ss->weighted ? e_weight : 1.0;
2729 struct ccase *c = case_create (casewriter_get_proto (s->writer));
2730 *case_num_rw_idx (c, 0) = value->f;
2731 *case_num_rw_idx (c, 1) = w;
2732 casewriter_write (s->writer, c);
2739 ctables_summary_value (const struct ctables_cell *cell,
2740 union ctables_summary *s,
2741 const struct ctables_summary_spec *ss)
2743 switch (ss->function)
2750 return cell->areas[ss->calc_area]->sequence;
2752 case CTSF_areaPCT_COUNT:
2754 const struct ctables_area *a = cell->areas[ss->calc_area];
2755 double a_count = ss->weighted ? a->e_count : a->u_count;
2756 return a_count ? s->count / a_count * 100 : SYSMIS;
2759 case CTSF_areaPCT_VALIDN:
2761 const struct ctables_area *a = cell->areas[ss->calc_area];
2762 double a_valid = ss->weighted ? a->e_valid : a->u_valid;
2763 return a_valid ? s->count / a_valid * 100 : SYSMIS;
2766 case CTSF_areaPCT_TOTALN:
2768 const struct ctables_area *a = cell->areas[ss->calc_area];
2769 double a_total = ss->weighted ? a->e_total : a->u_total;
2770 return a_total ? s->count / a_total * 100 : SYSMIS;
2787 return s->max != SYSMIS && s->min != SYSMIS ? s->max - s->min : SYSMIS;
2792 moments1_calculate (s->moments, NULL, &mean, NULL, NULL, NULL);
2798 double weight, variance;
2799 moments1_calculate (s->moments, &weight, NULL, &variance, NULL, NULL);
2800 return calc_semean (variance, weight);
2806 moments1_calculate (s->moments, NULL, NULL, &variance, NULL, NULL);
2807 return variance != SYSMIS ? sqrt (variance) : SYSMIS;
2812 double weight, mean;
2813 moments1_calculate (s->moments, &weight, &mean, NULL, NULL, NULL);
2814 return weight != SYSMIS && mean != SYSMIS ? weight * mean : SYSMIS;
2820 moments1_calculate (s->moments, NULL, NULL, &variance, NULL, NULL);
2824 case CTSF_areaPCT_SUM:
2826 double weight, mean;
2827 moments1_calculate (s->moments, &weight, &mean, NULL, NULL, NULL);
2828 if (weight == SYSMIS || mean == SYSMIS)
2831 const struct ctables_area *a = cell->areas[ss->calc_area];
2832 const struct ctables_sum *sum = &a->sums[ss->sum_var_idx];
2833 double denom = ss->weighted ? sum->e_sum : sum->u_sum;
2834 return denom != 0 ? weight * mean / denom * 100 : SYSMIS;
2841 struct casereader *reader = casewriter_make_reader (s->writer);
2844 struct percentile *ptile = percentile_create (
2845 ss->function == CTSF_PTILE ? ss->percentile : 0.5, s->ovalid);
2846 struct order_stats *os = &ptile->parent;
2847 order_stats_accumulate_idx (&os, 1, reader, 1, 0);
2848 s->ovalue = percentile_calculate (ptile, PC_HAVERAGE);
2849 statistic_destroy (&ptile->parent.parent);
2856 struct casereader *reader = casewriter_make_reader (s->writer);
2859 struct mode *mode = mode_create ();
2860 struct order_stats *os = &mode->parent;
2861 order_stats_accumulate_idx (&os, 1, reader, 1, 0);
2862 s->ovalue = mode->mode;
2863 statistic_destroy (&mode->parent.parent);
2871 struct ctables_cell_sort_aux
2873 const struct ctables_nest *nest;
2874 enum pivot_axis_type a;
2878 ctables_cell_compare_3way (const void *a_, const void *b_, const void *aux_)
2880 const struct ctables_cell_sort_aux *aux = aux_;
2881 struct ctables_cell *const *ap = a_;
2882 struct ctables_cell *const *bp = b_;
2883 const struct ctables_cell *a = *ap;
2884 const struct ctables_cell *b = *bp;
2886 const struct ctables_nest *nest = aux->nest;
2887 for (size_t i = 0; i < nest->n; i++)
2888 if (i != nest->scale_idx)
2890 const struct variable *var = nest->vars[i];
2891 const struct ctables_cell_value *a_cv = &a->axes[aux->a].cvs[i];
2892 const struct ctables_cell_value *b_cv = &b->axes[aux->a].cvs[i];
2893 if (a_cv->category != b_cv->category)
2894 return a_cv->category > b_cv->category ? 1 : -1;
2896 const union value *a_val = &a_cv->value;
2897 const union value *b_val = &b_cv->value;
2898 switch (a_cv->category->type)
2904 case CCT_POSTCOMPUTE:
2905 case CCT_EXCLUDED_MISSING:
2906 /* Must be equal. */
2914 int cmp = value_compare_3way (a_val, b_val, var_get_width (var));
2922 int cmp = value_compare_3way (a_val, b_val, var_get_width (var));
2924 return a_cv->category->sort_ascending ? cmp : -cmp;
2930 const char *a_label = var_lookup_value_label (var, a_val);
2931 const char *b_label = var_lookup_value_label (var, b_val);
2937 cmp = strcmp (a_label, b_label);
2943 cmp = value_compare_3way (a_val, b_val, var_get_width (var));
2946 return a_cv->category->sort_ascending ? cmp : -cmp;
2958 ctables_cell_compare_leaf_3way (const void *a_, const void *b_,
2959 const void *aux UNUSED)
2961 struct ctables_cell *const *ap = a_;
2962 struct ctables_cell *const *bp = b_;
2963 const struct ctables_cell *a = *ap;
2964 const struct ctables_cell *b = *bp;
2966 for (enum pivot_axis_type axis = 0; axis < PIVOT_N_AXES; axis++)
2968 int al = a->axes[axis].leaf;
2969 int bl = b->axes[axis].leaf;
2971 return al > bl ? 1 : -1;
2979 For each ctables_table:
2980 For each combination of row vars:
2981 For each combination of column vars:
2982 For each combination of layer vars:
2984 Make a table of row values:
2985 Sort entries by row values
2986 Assign a 0-based index to each actual value
2987 Construct a dimension
2988 Make a table of column values
2989 Make a table of layer values
2991 Fill the table entry using the indexes from before.
2994 static struct ctables_area *
2995 ctables_area_insert (struct ctables_section *s, struct ctables_cell *cell,
2996 enum ctables_area_type area)
2999 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3001 const struct ctables_nest *nest = s->nests[a];
3002 for (size_t i = 0; i < nest->n_areas[area]; i++)
3004 size_t v_idx = nest->areas[area][i];
3005 struct ctables_cell_value *cv = &cell->axes[a].cvs[v_idx];
3006 hash = hash_pointer (cv->category, hash);
3007 if (cv->category->type != CCT_TOTAL
3008 && cv->category->type != CCT_SUBTOTAL
3009 && cv->category->type != CCT_POSTCOMPUTE)
3010 hash = value_hash (&cv->value,
3011 var_get_width (nest->vars[v_idx]), hash);
3015 struct ctables_area *a;
3016 HMAP_FOR_EACH_WITH_HASH (a, struct ctables_area, node, hash, &s->areas[area])
3018 const struct ctables_cell *df = a->example;
3019 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3021 const struct ctables_nest *nest = s->nests[a];
3022 for (size_t i = 0; i < nest->n_areas[area]; i++)
3024 size_t v_idx = nest->areas[area][i];
3025 struct ctables_cell_value *cv1 = &df->axes[a].cvs[v_idx];
3026 struct ctables_cell_value *cv2 = &cell->axes[a].cvs[v_idx];
3027 if (cv1->category != cv2->category
3028 || (cv1->category->type != CCT_TOTAL
3029 && cv1->category->type != CCT_SUBTOTAL
3030 && cv1->category->type != CCT_POSTCOMPUTE
3031 && !value_equal (&cv1->value, &cv2->value,
3032 var_get_width (nest->vars[v_idx]))))
3041 struct ctables_sum *sums = (s->table->n_sum_vars
3042 ? xzalloc (s->table->n_sum_vars * sizeof *sums)
3045 a = xmalloc (sizeof *a);
3046 *a = (struct ctables_area) { .example = cell, .sums = sums };
3047 hmap_insert (&s->areas[area], &a->node, hash);
3051 static struct substring
3052 rtrim_value (const union value *v, const struct variable *var)
3054 struct substring s = ss_buffer (CHAR_CAST (char *, v->s),
3055 var_get_width (var));
3056 ss_rtrim (&s, ss_cstr (" "));
3061 in_string_range (const union value *v, const struct variable *var,
3062 const struct substring *srange)
3064 struct substring s = rtrim_value (v, var);
3065 return ((!srange[0].string || ss_compare (s, srange[0]) >= 0)
3066 && (!srange[1].string || ss_compare (s, srange[1]) <= 0));
3069 static const struct ctables_category *
3070 ctables_categories_match (const struct ctables_categories *c,
3071 const union value *v, const struct variable *var)
3073 if (var_is_numeric (var) && v->f == SYSMIS)
3076 const struct ctables_category *othernm = NULL;
3077 for (size_t i = c->n_cats; i-- > 0; )
3079 const struct ctables_category *cat = &c->cats[i];
3083 if (cat->number == v->f)
3088 if (ss_equals (cat->string, rtrim_value (v, var)))
3093 if ((cat->nrange[0] == -DBL_MAX || v->f >= cat->nrange[0])
3094 && (cat->nrange[1] == DBL_MAX || v->f <= cat->nrange[1]))
3099 if (in_string_range (v, var, cat->srange))
3104 if (var_is_value_missing (var, v))
3108 case CCT_POSTCOMPUTE:
3123 return (cat->include_missing || !var_is_value_missing (var, v) ? cat
3126 case CCT_EXCLUDED_MISSING:
3131 return var_is_value_missing (var, v) ? NULL : othernm;
3134 static const struct ctables_category *
3135 ctables_categories_total (const struct ctables_categories *c)
3137 const struct ctables_category *first = &c->cats[0];
3138 const struct ctables_category *last = &c->cats[c->n_cats - 1];
3139 return (first->type == CCT_TOTAL ? first
3140 : last->type == CCT_TOTAL ? last
3144 static struct ctables_cell *
3145 ctables_cell_insert__ (struct ctables_section *s, const struct ccase *c,
3146 const struct ctables_category *cats[PIVOT_N_AXES][10])
3149 enum ctables_summary_variant sv = CSV_CELL;
3150 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3152 const struct ctables_nest *nest = s->nests[a];
3153 for (size_t i = 0; i < nest->n; i++)
3154 if (i != nest->scale_idx)
3156 hash = hash_pointer (cats[a][i], hash);
3157 if (cats[a][i]->type != CCT_TOTAL
3158 && cats[a][i]->type != CCT_SUBTOTAL
3159 && cats[a][i]->type != CCT_POSTCOMPUTE)
3160 hash = value_hash (case_data (c, nest->vars[i]),
3161 var_get_width (nest->vars[i]), hash);
3167 struct ctables_cell *cell;
3168 HMAP_FOR_EACH_WITH_HASH (cell, struct ctables_cell, node, hash, &s->cells)
3170 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3172 const struct ctables_nest *nest = s->nests[a];
3173 for (size_t i = 0; i < nest->n; i++)
3174 if (i != nest->scale_idx
3175 && (cats[a][i] != cell->axes[a].cvs[i].category
3176 || (cats[a][i]->type != CCT_TOTAL
3177 && cats[a][i]->type != CCT_SUBTOTAL
3178 && cats[a][i]->type != CCT_POSTCOMPUTE
3179 && !value_equal (case_data (c, nest->vars[i]),
3180 &cell->axes[a].cvs[i].value,
3181 var_get_width (nest->vars[i])))))
3190 cell = xmalloc (sizeof *cell);
3193 cell->omit_areas = 0;
3194 cell->postcompute = false;
3195 //struct string name = DS_EMPTY_INITIALIZER;
3196 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3198 const struct ctables_nest *nest = s->nests[a];
3199 cell->axes[a].cvs = (nest->n
3200 ? xnmalloc (nest->n, sizeof *cell->axes[a].cvs)
3202 for (size_t i = 0; i < nest->n; i++)
3204 const struct ctables_category *cat = cats[a][i];
3205 const struct variable *var = nest->vars[i];
3206 const union value *value = case_data (c, var);
3207 if (i != nest->scale_idx)
3209 const struct ctables_category *subtotal = cat->subtotal;
3210 if (cat->hide || (subtotal && subtotal->hide_subcategories))
3213 if (cat->type == CCT_TOTAL
3214 || cat->type == CCT_SUBTOTAL
3215 || cat->type == CCT_POSTCOMPUTE)
3217 /* XXX these should be more encompassing I think.*/
3221 case PIVOT_AXIS_COLUMN:
3222 cell->omit_areas |= ((1u << CTAT_TABLE) |
3223 (1u << CTAT_LAYER) |
3224 (1u << CTAT_LAYERCOL) |
3225 (1u << CTAT_SUBTABLE) |
3228 case PIVOT_AXIS_ROW:
3229 cell->omit_areas |= ((1u << CTAT_TABLE) |
3230 (1u << CTAT_LAYER) |
3231 (1u << CTAT_LAYERROW) |
3232 (1u << CTAT_SUBTABLE) |
3235 case PIVOT_AXIS_LAYER:
3236 cell->omit_areas |= ((1u << CTAT_TABLE) |
3237 (1u << CTAT_LAYER));
3241 if (cat->type == CCT_POSTCOMPUTE)
3242 cell->postcompute = true;
3245 cell->axes[a].cvs[i].category = cat;
3246 value_clone (&cell->axes[a].cvs[i].value, value, var_get_width (var));
3249 if (i != nest->scale_idx)
3251 if (!ds_is_empty (&name))
3252 ds_put_cstr (&name, ", ");
3253 char *value_s = data_out (value, var_get_encoding (var),
3254 var_get_print_format (var),
3255 settings_get_fmt_settings ());
3256 if (cat->type == CCT_TOTAL
3257 || cat->type == CCT_SUBTOTAL
3258 || cat->type == CCT_POSTCOMPUTE)
3259 ds_put_format (&name, "%s=total", var_get_name (var));
3261 ds_put_format (&name, "%s=%s", var_get_name (var),
3262 value_s + strspn (value_s, " "));
3268 //cell->name = ds_steal_cstr (&name);
3270 const struct ctables_nest *ss = s->nests[s->table->summary_axis];
3271 const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
3272 cell->summaries = xmalloc (specs->n * sizeof *cell->summaries);
3273 for (size_t i = 0; i < specs->n; i++)
3274 ctables_summary_init (&cell->summaries[i], &specs->specs[i]);
3275 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
3276 cell->areas[at] = ctables_area_insert (s, cell, at);
3277 hmap_insert (&s->cells, &cell->node, hash);
3282 is_scale_missing (const struct ctables_summary_spec_set *specs,
3283 const struct ccase *c)
3285 if (!specs->is_scale)
3288 if (var_is_num_missing (specs->var, case_num (c, specs->var)))
3291 for (size_t i = 0; i < specs->n_listwise_vars; i++)
3293 const struct variable *var = specs->listwise_vars[i];
3294 if (var_is_num_missing (var, case_num (c, var)))
3302 ctables_cell_add__ (struct ctables_section *s, const struct ccase *c,
3303 const struct ctables_category *cats[PIVOT_N_AXES][10],
3304 bool is_missing, bool excluded_missing,
3305 double d_weight, double e_weight)
3307 struct ctables_cell *cell = ctables_cell_insert__ (s, c, cats);
3308 const struct ctables_nest *ss = s->nests[s->table->summary_axis];
3310 const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
3312 bool scale_missing = is_scale_missing (specs, c);
3313 for (size_t i = 0; i < specs->n; i++)
3314 ctables_summary_add (&cell->summaries[i], &specs->specs[i],
3315 specs->var, case_data (c, specs->var), specs->is_scale,
3316 scale_missing, is_missing, excluded_missing,
3317 d_weight, e_weight);
3318 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
3319 if (!(cell->omit_areas && (1u << at)))
3321 struct ctables_area *a = cell->areas[at];
3322 a->d_total += d_weight;
3323 a->e_total += e_weight;
3325 if (!excluded_missing)
3327 a->d_count += d_weight;
3328 a->e_count += e_weight;
3333 a->d_valid += d_weight;
3334 a->e_valid += e_weight;
3337 for (size_t i = 0; i < s->table->n_sum_vars; i++)
3339 /* XXX listwise_missing??? */
3340 const struct variable *var = s->table->sum_vars[i];
3341 double addend = case_num (c, var);
3342 if (!var_is_num_missing (var, addend))
3344 struct ctables_sum *sum = &a->sums[i];
3345 sum->e_sum += addend * e_weight;
3346 sum->u_sum += addend;
3354 recurse_totals (struct ctables_section *s, const struct ccase *c,
3355 const struct ctables_category *cats[PIVOT_N_AXES][10],
3356 bool is_missing, bool excluded_missing,
3357 double d_weight, double e_weight,
3358 enum pivot_axis_type start_axis, size_t start_nest)
3360 for (enum pivot_axis_type a = start_axis; a < PIVOT_N_AXES; a++)
3362 const struct ctables_nest *nest = s->nests[a];
3363 for (size_t i = start_nest; i < nest->n; i++)
3365 if (i == nest->scale_idx)
3368 const struct variable *var = nest->vars[i];
3370 const struct ctables_category *total = ctables_categories_total (
3371 s->table->categories[var_get_dict_index (var)]);
3374 const struct ctables_category *save = cats[a][i];
3376 ctables_cell_add__ (s, c, cats, is_missing, excluded_missing,
3377 d_weight, e_weight);
3378 recurse_totals (s, c, cats, is_missing, excluded_missing,
3379 d_weight, e_weight, a, i + 1);
3388 recurse_subtotals (struct ctables_section *s, const struct ccase *c,
3389 const struct ctables_category *cats[PIVOT_N_AXES][10],
3390 bool is_missing, bool excluded_missing,
3391 double d_weight, double e_weight,
3392 enum pivot_axis_type start_axis, size_t start_nest)
3394 for (enum pivot_axis_type a = start_axis; a < PIVOT_N_AXES; a++)
3396 const struct ctables_nest *nest = s->nests[a];
3397 for (size_t i = start_nest; i < nest->n; i++)
3399 if (i == nest->scale_idx)
3402 const struct ctables_category *save = cats[a][i];
3405 cats[a][i] = save->subtotal;
3406 ctables_cell_add__ (s, c, cats, is_missing, excluded_missing,
3407 d_weight, e_weight);
3408 recurse_subtotals (s, c, cats, is_missing, excluded_missing,
3409 d_weight, e_weight, a, i + 1);
3418 ctables_add_occurrence (const struct variable *var,
3419 const union value *value,
3420 struct hmap *occurrences)
3422 int width = var_get_width (var);
3423 unsigned int hash = value_hash (value, width, 0);
3425 struct ctables_occurrence *o;
3426 HMAP_FOR_EACH_WITH_HASH (o, struct ctables_occurrence, node, hash,
3428 if (value_equal (value, &o->value, width))
3431 o = xmalloc (sizeof *o);
3432 value_clone (&o->value, value, width);
3433 hmap_insert (occurrences, &o->node, hash);
3437 ctables_cell_insert (struct ctables_section *s,
3438 const struct ccase *c,
3439 double d_weight, double e_weight)
3441 const struct ctables_category *cats[PIVOT_N_AXES][10]; /* XXX */
3443 /* Does at least one categorical variable have a missing value in an included
3444 or excluded category? */
3445 bool is_missing = false;
3447 /* Does at least one categorical variable have a missing value in an excluded
3449 bool excluded_missing = false;
3451 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3453 const struct ctables_nest *nest = s->nests[a];
3454 for (size_t i = 0; i < nest->n; i++)
3456 if (i == nest->scale_idx)
3459 const struct variable *var = nest->vars[i];
3460 const union value *value = case_data (c, var);
3462 bool var_missing = var_is_value_missing (var, value) != 0;
3466 cats[a][i] = ctables_categories_match (
3467 s->table->categories[var_get_dict_index (var)], value, var);
3473 static const struct ctables_category cct_excluded_missing = {
3474 .type = CCT_EXCLUDED_MISSING,
3477 cats[a][i] = &cct_excluded_missing;
3478 excluded_missing = true;
3483 if (!excluded_missing)
3484 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3486 const struct ctables_nest *nest = s->nests[a];
3487 for (size_t i = 0; i < nest->n; i++)
3488 if (i != nest->scale_idx)
3490 const struct variable *var = nest->vars[i];
3491 const union value *value = case_data (c, var);
3492 ctables_add_occurrence (var, value, &s->occurrences[a][i]);
3496 ctables_cell_add__ (s, c, cats, is_missing, excluded_missing,
3497 d_weight, e_weight);
3499 //if (!excluded_missing)
3501 recurse_totals (s, c, cats, is_missing, excluded_missing,
3502 d_weight, e_weight, 0, 0);
3503 recurse_subtotals (s, c, cats, is_missing, excluded_missing,
3504 d_weight, e_weight, 0, 0);
3510 const struct ctables_summary_spec_set *set;
3515 merge_item_compare_3way (const struct merge_item *a, const struct merge_item *b)
3517 const struct ctables_summary_spec *as = &a->set->specs[a->ofs];
3518 const struct ctables_summary_spec *bs = &b->set->specs[b->ofs];
3519 if (as->function != bs->function)
3520 return as->function > bs->function ? 1 : -1;
3521 else if (as->weighted != bs->weighted)
3522 return as->weighted > bs->weighted ? 1 : -1;
3523 else if (as->calc_area != bs->calc_area)
3524 return as->calc_area > bs->calc_area ? 1 : -1;
3525 else if (as->percentile != bs->percentile)
3526 return as->percentile < bs->percentile ? 1 : -1;
3528 const char *as_label = as->label ? as->label : "";
3529 const char *bs_label = bs->label ? bs->label : "";
3530 return strcmp (as_label, bs_label);
3534 ctables_category_format_number (double number, const struct variable *var,
3537 struct pivot_value *pv = pivot_value_new_var_value (
3538 var, &(union value) { .f = number });
3539 pivot_value_format (pv, NULL, s);
3540 pivot_value_destroy (pv);
3544 ctables_category_format_string (struct substring string,
3545 const struct variable *var, struct string *out)
3547 int width = var_get_width (var);
3548 char *s = xmalloc (width);
3549 buf_copy_rpad (s, width, string.string, string.length, ' ');
3550 struct pivot_value *pv = pivot_value_new_var_value (
3551 var, &(union value) { .s = CHAR_CAST (uint8_t *, s) });
3552 pivot_value_format (pv, NULL, out);
3553 pivot_value_destroy (pv);
3558 ctables_category_format_label (const struct ctables_category *cat,
3559 const struct variable *var,
3565 ctables_category_format_number (cat->number, var, s);
3569 ctables_category_format_string (cat->string, var, s);
3573 ctables_category_format_number (cat->nrange[0], var, s);
3574 ds_put_format (s, " THRU ");
3575 ctables_category_format_number (cat->nrange[1], var, s);
3579 ctables_category_format_string (cat->srange[0], var, s);
3580 ds_put_format (s, " THRU ");
3581 ctables_category_format_string (cat->srange[1], var, s);
3585 ds_put_cstr (s, "MISSING");
3589 ds_put_cstr (s, "OTHERNM");
3592 case CCT_POSTCOMPUTE:
3593 ds_put_format (s, "&%s", cat->pc->name);
3598 ds_put_cstr (s, cat->total_label);
3604 case CCT_EXCLUDED_MISSING:
3611 static struct pivot_value *
3612 ctables_postcompute_label (const struct ctables_categories *cats,
3613 const struct ctables_category *cat,
3614 const struct variable *var)
3616 struct substring in = ss_cstr (cat->pc->label);
3617 struct substring target = ss_cstr (")LABEL[");
3619 struct string out = DS_EMPTY_INITIALIZER;
3622 size_t chunk = ss_find_substring (in, target);
3623 if (chunk == SIZE_MAX)
3625 if (ds_is_empty (&out))
3626 return pivot_value_new_user_text (in.string, in.length);
3629 ds_put_substring (&out, in);
3630 return pivot_value_new_user_text_nocopy (ds_steal_cstr (&out));
3634 ds_put_substring (&out, ss_head (in, chunk));
3635 ss_advance (&in, chunk + target.length);
3637 struct substring idx_s;
3638 if (!ss_get_until (&in, ']', &idx_s))
3641 long int idx = strtol (idx_s.string, &tail, 10);
3642 if (idx < 1 || idx > cats->n_cats || tail != ss_end (idx_s))
3645 struct ctables_category *cat2 = &cats->cats[idx - 1];
3646 if (!ctables_category_format_label (cat2, var, &out))
3652 return pivot_value_new_user_text (cat->pc->label, SIZE_MAX);
3655 static struct pivot_value *
3656 ctables_category_create_value_label (const struct ctables_categories *cats,
3657 const struct ctables_category *cat,
3658 const struct variable *var,
3659 const union value *value)
3661 return (cat->type == CCT_POSTCOMPUTE && cat->pc->label
3662 ? ctables_postcompute_label (cats, cat, var)
3663 : cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL
3664 ? pivot_value_new_user_text (cat->total_label, SIZE_MAX)
3665 : pivot_value_new_var_value (var, value));
3668 static struct ctables_value *
3669 ctables_value_find__ (struct ctables_table *t, const union value *value,
3670 int width, unsigned int hash)
3672 struct ctables_value *clv;
3673 HMAP_FOR_EACH_WITH_HASH (clv, struct ctables_value, node,
3674 hash, &t->clabels_values_map)
3675 if (value_equal (value, &clv->value, width))
3681 ctables_value_insert (struct ctables_table *t, const union value *value,
3684 unsigned int hash = value_hash (value, width, 0);
3685 struct ctables_value *clv = ctables_value_find__ (t, value, width, hash);
3688 clv = xmalloc (sizeof *clv);
3689 value_clone (&clv->value, value, width);
3690 hmap_insert (&t->clabels_values_map, &clv->node, hash);
3694 static struct ctables_value *
3695 ctables_value_find (struct ctables_table *t,
3696 const union value *value, int width)
3698 return ctables_value_find__ (t, value, width,
3699 value_hash (value, width, 0));
3703 ctables_table_add_section (struct ctables_table *t, enum pivot_axis_type a,
3704 size_t ix[PIVOT_N_AXES])
3706 if (a < PIVOT_N_AXES)
3708 size_t limit = MAX (t->stacks[a].n, 1);
3709 for (ix[a] = 0; ix[a] < limit; ix[a]++)
3710 ctables_table_add_section (t, a + 1, ix);
3714 struct ctables_section *s = &t->sections[t->n_sections++];
3715 *s = (struct ctables_section) {
3717 .cells = HMAP_INITIALIZER (s->cells),
3719 for (a = 0; a < PIVOT_N_AXES; a++)
3722 struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
3724 s->occurrences[a] = xnmalloc (nest->n, sizeof *s->occurrences[a]);
3725 for (size_t i = 0; i < nest->n; i++)
3726 hmap_init (&s->occurrences[a][i]);
3728 for (size_t i = 0; i < N_CTATS; i++)
3729 hmap_init (&s->areas[i]);
3734 ctpo_add (double a, double b)
3740 ctpo_sub (double a, double b)
3746 ctpo_mul (double a, double b)
3752 ctpo_div (double a, double b)
3754 return b ? a / b : SYSMIS;
3758 ctpo_pow (double a, double b)
3760 int save_errno = errno;
3762 double result = pow (a, b);
3770 ctpo_neg (double a, double b UNUSED)
3775 struct ctables_pcexpr_evaluate_ctx
3777 const struct ctables_cell *cell;
3778 const struct ctables_section *section;
3779 const struct ctables_categories *cats;
3780 enum pivot_axis_type pc_a;
3783 enum fmt_type parse_format;
3786 static double ctables_pcexpr_evaluate (
3787 const struct ctables_pcexpr_evaluate_ctx *, const struct ctables_pcexpr *);
3790 ctables_pcexpr_evaluate_nonterminal (
3791 const struct ctables_pcexpr_evaluate_ctx *ctx,
3792 const struct ctables_pcexpr *e, size_t n_args,
3793 double evaluate (double, double))
3795 double args[2] = { 0, 0 };
3796 for (size_t i = 0; i < n_args; i++)
3798 args[i] = ctables_pcexpr_evaluate (ctx, e->subs[i]);
3799 if (!isfinite (args[i]) || args[i] == SYSMIS)
3802 return evaluate (args[0], args[1]);
3806 ctables_pcexpr_evaluate_category (const struct ctables_pcexpr_evaluate_ctx *ctx,
3807 const struct ctables_cell_value *pc_cv)
3809 const struct ctables_section *s = ctx->section;
3812 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3814 const struct ctables_nest *nest = s->nests[a];
3815 for (size_t i = 0; i < nest->n; i++)
3816 if (i != nest->scale_idx)
3818 const struct ctables_cell_value *cv
3819 = (a == ctx->pc_a && i == ctx->pc_a_idx ? pc_cv
3820 : &ctx->cell->axes[a].cvs[i]);
3821 hash = hash_pointer (cv->category, hash);
3822 if (cv->category->type != CCT_TOTAL
3823 && cv->category->type != CCT_SUBTOTAL
3824 && cv->category->type != CCT_POSTCOMPUTE)
3825 hash = value_hash (&cv->value,
3826 var_get_width (nest->vars[i]), hash);
3830 struct ctables_cell *tc;
3831 HMAP_FOR_EACH_WITH_HASH (tc, struct ctables_cell, node, hash, &s->cells)
3833 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
3835 const struct ctables_nest *nest = s->nests[a];
3836 for (size_t i = 0; i < nest->n; i++)
3837 if (i != nest->scale_idx)
3839 const struct ctables_cell_value *p_cv
3840 = (a == ctx->pc_a && i == ctx->pc_a_idx ? pc_cv
3841 : &ctx->cell->axes[a].cvs[i]);
3842 const struct ctables_cell_value *t_cv = &tc->axes[a].cvs[i];
3843 if (p_cv->category != t_cv->category
3844 || (p_cv->category->type != CCT_TOTAL
3845 && p_cv->category->type != CCT_SUBTOTAL
3846 && p_cv->category->type != CCT_POSTCOMPUTE
3847 && !value_equal (&p_cv->value,
3849 var_get_width (nest->vars[i]))))
3861 const struct ctables_table *t = s->table;
3862 const struct ctables_nest *specs_nest = s->nests[t->summary_axis];
3863 const struct ctables_summary_spec_set *specs = &specs_nest->specs[tc->sv];
3864 return ctables_summary_value (tc, &tc->summaries[ctx->summary_idx],
3865 &specs->specs[ctx->summary_idx]);
3869 ctables_pcexpr_evaluate (const struct ctables_pcexpr_evaluate_ctx *ctx,
3870 const struct ctables_pcexpr *e)
3877 case CTPO_CAT_NRANGE:
3878 case CTPO_CAT_SRANGE:
3879 case CTPO_CAT_MISSING:
3880 case CTPO_CAT_OTHERNM:
3882 struct ctables_cell_value cv = {
3883 .category = ctables_find_category_for_postcompute (ctx->section->table->ctables->dict, ctx->cats, ctx->parse_format, e)
3885 assert (cv.category != NULL);
3887 struct hmap *occurrences = &ctx->section->occurrences[ctx->pc_a][ctx->pc_a_idx];
3888 const struct ctables_occurrence *o;
3891 const struct variable *var = ctx->section->nests[ctx->pc_a]->vars[ctx->pc_a_idx];
3892 HMAP_FOR_EACH (o, struct ctables_occurrence, node, occurrences)
3893 if (ctables_categories_match (ctx->cats, &o->value, var) == cv.category)
3895 cv.value = o->value;
3896 sum += ctables_pcexpr_evaluate_category (ctx, &cv);
3901 case CTPO_CAT_NUMBER:
3902 case CTPO_CAT_SUBTOTAL:
3903 case CTPO_CAT_TOTAL:
3905 struct ctables_cell_value cv = {
3906 .category = ctables_find_category_for_postcompute (ctx->section->table->ctables->dict, ctx->cats, ctx->parse_format, e),
3907 .value = { .f = e->number },
3909 assert (cv.category != NULL);
3910 return ctables_pcexpr_evaluate_category (ctx, &cv);
3913 case CTPO_CAT_STRING:
3915 int width = var_get_width (ctx->section->nests[ctx->pc_a]->vars[ctx->pc_a_idx]);
3917 if (width > e->string.length)
3919 s = xmalloc (width);
3920 buf_copy_rpad (s, width, e->string.string, e->string.length, ' ');
3923 const struct ctables_category *category
3924 = ctables_find_category_for_postcompute (
3925 ctx->section->table->ctables->dict,
3926 ctx->cats, ctx->parse_format, e);
3927 assert (category != NULL);
3929 struct ctables_cell_value cv = { .category = category };
3930 if (category->type == CCT_NUMBER)
3931 cv.value.f = category->number;
3932 else if (category->type == CCT_STRING)
3933 cv.value.s = CHAR_CAST (uint8_t *, s ? s : e->string.string);
3937 double retval = ctables_pcexpr_evaluate_category (ctx, &cv);
3943 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 2, ctpo_add);
3946 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 2, ctpo_sub);
3949 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 2, ctpo_mul);
3952 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 2, ctpo_div);
3955 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 2, ctpo_pow);
3958 return ctables_pcexpr_evaluate_nonterminal (ctx, e, 1, ctpo_neg);
3964 static const struct ctables_category *
3965 ctables_cell_postcompute (const struct ctables_section *s,
3966 const struct ctables_cell *cell,
3967 enum pivot_axis_type *pc_a_p,
3970 assert (cell->postcompute);
3971 const struct ctables_category *pc_cat = NULL;
3972 for (enum pivot_axis_type pc_a = 0; pc_a < PIVOT_N_AXES; pc_a++)
3973 for (size_t pc_a_idx = 0; pc_a_idx < s->nests[pc_a]->n; pc_a_idx++)
3975 const struct ctables_cell_value *cv = &cell->axes[pc_a].cvs[pc_a_idx];
3976 if (cv->category->type == CCT_POSTCOMPUTE)
3980 /* Multiple postcomputes cross each other. The value is
3985 pc_cat = cv->category;
3989 *pc_a_idx_p = pc_a_idx;
3993 assert (pc_cat != NULL);
3998 ctables_cell_calculate_postcompute (const struct ctables_section *s,
3999 const struct ctables_cell *cell,
4000 const struct ctables_summary_spec *ss,
4001 struct fmt_spec *format,
4002 bool *is_ctables_format,
4005 enum pivot_axis_type pc_a = 0;
4006 size_t pc_a_idx = 0;
4007 const struct ctables_category *pc_cat = ctables_cell_postcompute (
4008 s, cell, &pc_a, &pc_a_idx);
4012 const struct ctables_postcompute *pc = pc_cat->pc;
4015 for (size_t i = 0; i < pc->specs->n; i++)
4017 const struct ctables_summary_spec *ss2 = &pc->specs->specs[i];
4018 if (ss->function == ss2->function
4019 && ss->weighted == ss2->weighted
4020 && ss->calc_area == ss2->calc_area
4021 && ss->percentile == ss2->percentile)
4023 *format = ss2->format;
4024 *is_ctables_format = ss2->is_ctables_format;
4030 const struct variable *var = s->nests[pc_a]->vars[pc_a_idx];
4031 const struct ctables_categories *cats = s->table->categories[
4032 var_get_dict_index (var)];
4033 struct ctables_pcexpr_evaluate_ctx ctx = {
4038 .pc_a_idx = pc_a_idx,
4039 .summary_idx = summary_idx,
4040 .parse_format = pc_cat->parse_format,
4042 return ctables_pcexpr_evaluate (&ctx, pc->expr);
4046 ctables_format (double d, const struct fmt_spec *format,
4047 const struct fmt_settings *settings)
4049 const union value v = { .f = d };
4050 char *s = data_out_stretchy (&v, "UTF-8", format, settings, NULL);
4052 /* The custom-currency specifications for NEQUAL, PAREN, and PCTPAREN don't
4053 produce the results we want for negative numbers, putting the negative
4054 sign in the wrong spot, before the prefix instead of after it. We can't,
4055 in fact, produce the desired results using a custom-currency
4056 specification. Instead, we postprocess the output, moving the negative
4059 NEQUAL: "-N=3" => "N=-3"
4060 PAREN: "-(3)" => "(-3)"
4061 PCTPAREN: "-(3%)" => "(-3%)"
4063 This transformation doesn't affect NEGPAREN. */
4064 char *minus_src = strchr (s, '-');
4065 if (minus_src && (minus_src == s || minus_src[-1] != 'E'))
4067 char *n_equals = strstr (s, "N=");
4068 char *lparen = strchr (s, '(');
4069 char *minus_dst = n_equals ? n_equals + 1 : lparen;
4071 move_element (s, minus_dst - s + 1, 1, minus_src - s, minus_dst - s);
4077 all_hidden_vlabels (const struct ctables_table *t, enum pivot_axis_type a)
4079 for (size_t i = 0; i < t->stacks[a].n; i++)
4081 struct ctables_nest *nest = &t->stacks[a].nests[i];
4082 if (nest->n != 1 || nest->scale_idx != 0)
4085 enum ctables_vlabel vlabel
4086 = t->ctables->vlabels[var_get_dict_index (nest->vars[0])];
4087 if (vlabel != CTVL_NONE)
4094 ctables_table_output (struct ctables *ct, struct ctables_table *t)
4096 struct pivot_table *pt = pivot_table_create__ (
4098 ? pivot_value_new_user_text (t->title, SIZE_MAX)
4099 : pivot_value_new_text (N_("Custom Tables"))),
4102 pivot_table_set_caption (
4103 pt, pivot_value_new_user_text (t->caption, SIZE_MAX));
4105 pivot_table_set_corner_text (
4106 pt, pivot_value_new_user_text (t->corner, SIZE_MAX));
4108 bool summary_dimension = (t->summary_axis != t->slabels_axis
4109 || (!t->slabels_visible
4110 && t->summary_specs.n > 1));
4111 if (summary_dimension)
4113 struct pivot_dimension *d = pivot_dimension_create (
4114 pt, t->slabels_axis, N_("Statistics"));
4115 const struct ctables_summary_spec_set *specs = &t->summary_specs;
4116 if (!t->slabels_visible)
4117 d->hide_all_labels = true;
4118 for (size_t i = 0; i < specs->n; i++)
4119 pivot_category_create_leaf (
4120 d->root, ctables_summary_label (&specs->specs[i], t->cilevel));
4123 bool categories_dimension = t->clabels_example != NULL;
4124 if (categories_dimension)
4126 struct pivot_dimension *d = pivot_dimension_create (
4127 pt, t->label_axis[t->clabels_from_axis],
4128 t->clabels_from_axis == PIVOT_AXIS_ROW
4129 ? N_("Row Categories")
4130 : N_("Column Categories"));
4131 const struct variable *var = t->clabels_example;
4132 const struct ctables_categories *c = t->categories[var_get_dict_index (var)];
4133 for (size_t i = 0; i < t->n_clabels_values; i++)
4135 const struct ctables_value *value = t->clabels_values[i];
4136 const struct ctables_category *cat = ctables_categories_match (c, &value->value, var);
4137 assert (cat != NULL);
4138 pivot_category_create_leaf (
4139 d->root, ctables_category_create_value_label (c, cat,
4145 pivot_table_set_look (pt, ct->look);
4146 struct pivot_dimension *d[PIVOT_N_AXES];
4147 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
4149 static const char *names[] = {
4150 [PIVOT_AXIS_ROW] = N_("Rows"),
4151 [PIVOT_AXIS_COLUMN] = N_("Columns"),
4152 [PIVOT_AXIS_LAYER] = N_("Layers"),
4154 d[a] = (t->axes[a] || a == t->summary_axis
4155 ? pivot_dimension_create (pt, a, names[a])
4160 assert (t->axes[a]);
4162 for (size_t i = 0; i < t->stacks[a].n; i++)
4164 struct ctables_nest *nest = &t->stacks[a].nests[i];
4165 struct ctables_section **sections = xnmalloc (t->n_sections,
4167 size_t n_sections = 0;
4169 size_t n_total_cells = 0;
4170 size_t max_depth = 0;
4171 for (size_t j = 0; j < t->n_sections; j++)
4172 if (t->sections[j].nests[a] == nest)
4174 struct ctables_section *s = &t->sections[j];
4175 sections[n_sections++] = s;
4176 n_total_cells += hmap_count (&s->cells);
4178 size_t depth = s->nests[a]->n;
4179 max_depth = MAX (depth, max_depth);
4182 struct ctables_cell **sorted = xnmalloc (n_total_cells,
4184 size_t n_sorted = 0;
4186 for (size_t j = 0; j < n_sections; j++)
4188 struct ctables_section *s = sections[j];
4190 struct ctables_cell *cell;
4191 HMAP_FOR_EACH (cell, struct ctables_cell, node, &s->cells)
4193 sorted[n_sorted++] = cell;
4194 assert (n_sorted <= n_total_cells);
4197 struct ctables_cell_sort_aux aux = { .nest = nest, .a = a };
4198 sort (sorted, n_sorted, sizeof *sorted, ctables_cell_compare_3way, &aux);
4201 if (a == PIVOT_AXIS_ROW)
4203 size_t ids[N_CTATS];
4204 memset (ids, 0, sizeof ids);
4205 for (size_t j = 0; j < n_sorted; j++)
4207 struct ctables_cell *cell = sorted[j];
4208 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
4210 struct ctables_area *area = cell->areas[at];
4211 if (!area->sequence)
4212 area->sequence = ++ids[at];
4219 for (size_t j = 0; j < n_sorted; j++)
4221 printf ("%s (%s): %f/%f = %.1f%%\n", sorted[j]->name, sorted[j]->contributes_to_areas ? "y" : "n", sorted[j]->summaries[0].count, sorted[j]->areas[CTAT_COL]->e_count, sorted[j]->summaries[0].count / sorted[j]->areas[CTAT_COL]->e_count * 100.0);
4226 struct ctables_level
4228 enum ctables_level_type
4230 CTL_VAR, /* Variable label for nest->vars[var_idx]. */
4231 CTL_CATEGORY, /* Category for nest->vars[var_idx]. */
4232 CTL_SUMMARY, /* Summary functions. */
4236 enum settings_value_show vlabel; /* CTL_VAR only. */
4239 struct ctables_level *levels = xnmalloc (1 + 2 * max_depth, sizeof *levels);
4240 size_t n_levels = 0;
4241 for (size_t k = 0; k < nest->n; k++)
4243 enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[k])];
4244 if (vlabel == CTVL_NONE && nest->scale_idx == k)
4246 if (vlabel != CTVL_NONE)
4248 levels[n_levels++] = (struct ctables_level) {
4250 .vlabel = (enum settings_value_show) vlabel,
4255 if (nest->scale_idx != k
4256 && (k != nest->n - 1 || t->label_axis[a] == a))
4258 levels[n_levels++] = (struct ctables_level) {
4259 .type = CTL_CATEGORY,
4265 if (!summary_dimension && a == t->slabels_axis)
4267 levels[n_levels++] = (struct ctables_level) {
4268 .type = CTL_SUMMARY,
4269 .var_idx = SIZE_MAX,
4273 /* Pivot categories:
4275 - variable label for nest->vars[0], if vlabel != CTVL_NONE
4276 - category for nest->vars[0], if nest->scale_idx != 0
4277 - variable label for nest->vars[1], if vlabel != CTVL_NONE
4278 - category for nest->vars[1], if nest->scale_idx != 1
4280 - variable label for nest->vars[n - 1], if vlabel != CTVL_NONE
4281 - category for nest->vars[n - 1], if t->label_axis[a] == a && nest->scale_idx != n - 1.
4282 - summary function, if 'a == t->slabels_axis && a ==
4285 Additional dimensions:
4287 - If 'a == t->slabels_axis && a != t->summary_axis', add a summary
4289 - If 't->label_axis[b] == a' for some 'b != a', add a category
4294 struct pivot_category **groups = xnmalloc (1 + 2 * max_depth, sizeof *groups);
4296 for (size_t j = 0; j < n_sorted; j++)
4298 struct ctables_cell *cell = sorted[j];
4299 struct ctables_cell *prev = j > 0 ? sorted[j - 1] : NULL;
4301 size_t n_common = 0;
4304 for (; n_common < n_levels; n_common++)
4306 const struct ctables_level *level = &levels[n_common];
4307 if (level->type == CTL_CATEGORY)
4309 size_t var_idx = level->var_idx;
4310 const struct ctables_category *c = cell->axes[a].cvs[var_idx].category;
4311 if (prev->axes[a].cvs[var_idx].category != c)
4313 else if (c->type != CCT_SUBTOTAL
4314 && c->type != CCT_TOTAL
4315 && c->type != CCT_POSTCOMPUTE
4316 && !value_equal (&prev->axes[a].cvs[var_idx].value,
4317 &cell->axes[a].cvs[var_idx].value,
4318 var_get_type (nest->vars[var_idx])))
4324 for (size_t k = n_common; k < n_levels; k++)
4326 const struct ctables_level *level = &levels[k];
4327 struct pivot_category *parent = k ? groups[k - 1] : d[a]->root;
4328 if (level->type == CTL_SUMMARY)
4330 assert (k == n_levels - 1);
4332 const struct ctables_summary_spec_set *specs = &t->summary_specs;
4333 for (size_t m = 0; m < specs->n; m++)
4335 int leaf = pivot_category_create_leaf (
4336 parent, ctables_summary_label (&specs->specs[m],
4344 const struct variable *var = nest->vars[level->var_idx];
4345 struct pivot_value *label;
4346 if (level->type == CTL_VAR)
4348 label = pivot_value_new_variable (var);
4349 label->variable.show = level->vlabel;
4351 else if (level->type == CTL_CATEGORY)
4353 const struct ctables_cell_value *cv = &cell->axes[a].cvs[level->var_idx];
4354 label = ctables_category_create_value_label (
4355 t->categories[var_get_dict_index (var)],
4356 cv->category, var, &cv->value);
4361 if (k == n_levels - 1)
4362 prev_leaf = pivot_category_create_leaf (parent, label);
4364 groups[k] = pivot_category_create_group__ (parent, label);
4368 cell->axes[a].leaf = prev_leaf;
4377 d[a]->hide_all_labels = all_hidden_vlabels (t, a);
4381 size_t n_total_cells = 0;
4382 for (size_t j = 0; j < t->n_sections; j++)
4383 n_total_cells += hmap_count (&t->sections[j].cells);
4385 struct ctables_cell **sorted = xnmalloc (n_total_cells, sizeof *sorted);
4386 size_t n_sorted = 0;
4387 for (size_t j = 0; j < t->n_sections; j++)
4389 const struct ctables_section *s = &t->sections[j];
4390 struct ctables_cell *cell;
4391 HMAP_FOR_EACH (cell, struct ctables_cell, node, &s->cells)
4393 sorted[n_sorted++] = cell;
4395 assert (n_sorted <= n_total_cells);
4396 sort (sorted, n_sorted, sizeof *sorted, ctables_cell_compare_leaf_3way,
4398 size_t ids[N_CTATS];
4399 memset (ids, 0, sizeof ids);
4400 for (size_t j = 0; j < n_sorted; j++)
4402 struct ctables_cell *cell = sorted[j];
4403 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
4405 struct ctables_area *area = cell->areas[at];
4406 if (!area->sequence)
4407 area->sequence = ++ids[at];
4414 for (size_t i = 0; i < t->n_sections; i++)
4416 struct ctables_section *s = &t->sections[i];
4418 struct ctables_cell *cell;
4419 HMAP_FOR_EACH (cell, struct ctables_cell, node, &s->cells)
4424 const struct ctables_nest *specs_nest = s->nests[t->summary_axis];
4425 const struct ctables_summary_spec_set *specs = &specs_nest->specs[cell->sv];
4426 for (size_t j = 0; j < specs->n; j++)
4429 size_t n_dindexes = 0;
4431 if (summary_dimension)
4432 dindexes[n_dindexes++] = specs->specs[j].axis_idx;
4434 if (categories_dimension)
4436 const struct ctables_nest *clabels_nest = s->nests[t->clabels_from_axis];
4437 const struct variable *var = clabels_nest->vars[clabels_nest->n - 1];
4438 const union value *value = &cell->axes[t->clabels_from_axis].cvs[clabels_nest->n - 1].value;
4439 const struct ctables_value *ctv = ctables_value_find (t, value, var_get_width (var));
4442 dindexes[n_dindexes++] = ctv->leaf;
4445 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
4448 int leaf = cell->axes[a].leaf;
4449 if (a == t->summary_axis && !summary_dimension)
4451 dindexes[n_dindexes++] = leaf;
4454 const struct ctables_summary_spec *ss = &specs->specs[j];
4456 struct fmt_spec format = specs->specs[j].format;
4457 bool is_ctables_format = ss->is_ctables_format;
4458 double d = (cell->postcompute
4459 ? ctables_cell_calculate_postcompute (
4460 s, cell, ss, &format, &is_ctables_format, j)
4461 : ctables_summary_value (cell, &cell->summaries[j],
4464 struct pivot_value *value;
4465 if (ct->hide_threshold != 0
4466 && d < ct->hide_threshold
4467 && ctables_summary_function_is_count (ss->function))
4469 value = pivot_value_new_user_text_nocopy (
4470 xasprintf ("<%d", ct->hide_threshold));
4472 else if (d == 0 && ct->zero)
4473 value = pivot_value_new_user_text (ct->zero, SIZE_MAX);
4474 else if (d == SYSMIS && ct->missing)
4475 value = pivot_value_new_user_text (ct->missing, SIZE_MAX);
4476 else if (is_ctables_format)
4477 value = pivot_value_new_user_text_nocopy (
4478 ctables_format (d, &format, &ct->ctables_formats));
4481 value = pivot_value_new_number (d);
4482 value->numeric.format = format;
4484 /* XXX should text values be right-justified? */
4485 pivot_table_put (pt, dindexes, n_dindexes, value);
4490 pivot_table_submit (pt);
4494 ctables_check_label_position (struct ctables_table *t, enum pivot_axis_type a)
4496 enum pivot_axis_type label_pos = t->label_axis[a];
4500 t->clabels_from_axis = a;
4502 const char *subcommand_name = a == PIVOT_AXIS_ROW ? "ROWLABELS" : "COLLABELS";
4503 const char *pos_name = label_pos == PIVOT_AXIS_LAYER ? "LAYER" : "OPPOSITE";
4505 const struct ctables_stack *stack = &t->stacks[a];
4509 const struct ctables_nest *n0 = &stack->nests[0];
4512 assert (stack->n == 1);
4516 const struct variable *v0 = n0->vars[n0->n - 1];
4517 struct ctables_categories *c0 = t->categories[var_get_dict_index (v0)];
4518 t->clabels_example = v0;
4520 for (size_t i = 0; i < c0->n_cats; i++)
4521 if (c0->cats[i].type == CCT_FUNCTION)
4523 msg (SE, _("%s=%s is not allowed with sorting based "
4524 "on a summary function."),
4525 subcommand_name, pos_name);
4528 if (n0->n - 1 == n0->scale_idx)
4530 msg (SE, _("%s=%s requires the variables to be moved to be categorical, "
4531 "but %s is a scale variable."),
4532 subcommand_name, pos_name, var_get_name (v0));
4536 for (size_t i = 1; i < stack->n; i++)
4538 const struct ctables_nest *ni = &stack->nests[i];
4540 const struct variable *vi = ni->vars[ni->n - 1];
4541 struct ctables_categories *ci = t->categories[var_get_dict_index (vi)];
4543 if (ni->n - 1 == ni->scale_idx)
4545 msg (SE, _("%s=%s requires the variables to be moved to be "
4546 "categorical, but %s is a scale variable."),
4547 subcommand_name, pos_name, var_get_name (vi));
4550 if (var_get_width (v0) != var_get_width (vi))
4552 msg (SE, _("%s=%s requires the variables to be "
4553 "moved to have the same width, but %s has "
4554 "width %d and %s has width %d."),
4555 subcommand_name, pos_name,
4556 var_get_name (v0), var_get_width (v0),
4557 var_get_name (vi), var_get_width (vi));
4560 if (!val_labs_equal (var_get_value_labels (v0),
4561 var_get_value_labels (vi)))
4563 msg (SE, _("%s=%s requires the variables to be "
4564 "moved to have the same value labels, but %s "
4565 "and %s have different value labels."),
4566 subcommand_name, pos_name,
4567 var_get_name (v0), var_get_name (vi));
4570 if (!ctables_categories_equal (c0, ci))
4572 msg (SE, _("%s=%s requires the variables to be "
4573 "moved to have the same category "
4574 "specifications, but %s and %s have different "
4575 "category specifications."),
4576 subcommand_name, pos_name,
4577 var_get_name (v0), var_get_name (vi));
4586 add_sum_var (struct variable *var,
4587 struct variable ***sum_vars, size_t *n, size_t *allocated)
4589 for (size_t i = 0; i < *n; i++)
4590 if (var == (*sum_vars)[i])
4593 if (*n >= *allocated)
4594 *sum_vars = x2nrealloc (*sum_vars, allocated, sizeof **sum_vars);
4595 (*sum_vars)[*n] = var;
4599 static enum ctables_area_type
4600 rotate_area (enum ctables_area_type area)
4611 return CTAT_LAYERCOL;
4614 return CTAT_LAYERROW;
4627 enumerate_sum_vars (const struct ctables_axis *a,
4628 struct variable ***sum_vars, size_t *n, size_t *allocated)
4636 for (size_t i = 0; i < N_CSVS; i++)
4637 for (size_t j = 0; j < a->specs[i].n; j++)
4639 struct ctables_summary_spec *spec = &a->specs[i].specs[j];
4640 if (spec->function == CTSF_areaPCT_SUM)
4641 spec->sum_var_idx = add_sum_var (a->var, sum_vars, n, allocated);
4647 for (size_t i = 0; i < 2; i++)
4648 enumerate_sum_vars (a->subs[i], sum_vars, n, allocated);
4654 ctables_prepare_table (struct ctables_table *t)
4656 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
4659 t->stacks[a] = enumerate_fts (a, t->axes[a]);
4661 for (size_t j = 0; j < t->stacks[a].n; j++)
4663 struct ctables_nest *nest = &t->stacks[a].nests[j];
4664 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
4666 nest->areas[at] = xmalloc (nest->n * sizeof *nest->areas[at]);
4667 nest->n_areas[at] = 0;
4669 bool add_vars = (at == CTAT_LAYER ? a == PIVOT_AXIS_LAYER
4670 : at == CTAT_LAYERROW ? a != PIVOT_AXIS_COLUMN
4671 : at == CTAT_LAYERCOL ? a != PIVOT_AXIS_ROW
4672 : at == CTAT_TABLE ? false
4675 for (size_t k = 0; k < nest->n; k++)
4677 if (k == nest->scale_idx)
4679 nest->areas[at][nest->n_areas[at]++] = k;
4681 else if (at == CTAT_LAYERCOL && a == PIVOT_AXIS_ROW && t->label_axis[a] != a)
4683 for (size_t k = nest->n - 1; k < nest->n; k--)
4685 if (k == nest->scale_idx)
4687 nest->areas[at][nest->n_areas[at]++] = k;
4692 else if (at == CTAT_LAYERROW && a == PIVOT_AXIS_COLUMN && t->label_axis[a] != a)
4694 for (size_t k = nest->n - 1; k < nest->n; k--)
4696 if (k == nest->scale_idx)
4698 nest->areas[at][nest->n_areas[at]++] = k;
4703 else if (at == CTAT_LAYER && t->label_axis[a] == PIVOT_AXIS_LAYER)
4705 for (size_t k = nest->n - 1; k < nest->n; k--)
4707 if (k == nest->scale_idx)
4709 nest->areas[at][nest->n_areas[at]++] = k;
4715 if (at == CTAT_SUBTABLE && t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_COLUMN)
4717 size_t n_drop = (a == PIVOT_AXIS_LAYER ? 0
4718 : a == PIVOT_AXIS_ROW ? 2
4720 for (size_t i = 0; i < n_drop; i++)
4721 if (nest->n_areas[at] > 0)
4722 nest->n_areas[at]--;
4725 else if (at == CTAT_SUBTABLE && t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_ROW)
4727 size_t n_drop = (a == PIVOT_AXIS_LAYER ? 0
4728 : a == PIVOT_AXIS_COLUMN ? 2
4730 for (size_t i = 0; i < n_drop; i++)
4731 if (nest->n_areas[at] > 0)
4732 nest->n_areas[at]--;
4735 else if (at == CTAT_SUBTABLE && t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_LAYER)
4737 if (a == PIVOT_AXIS_ROW)
4739 size_t n = nest->n_areas[at];
4742 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4743 nest->n_areas[at]--;
4746 else if (a == PIVOT_AXIS_COLUMN)
4748 if (nest->n_areas[at] > 0)
4749 nest->n_areas[at]--;
4753 else if (at == CTAT_SUBTABLE && t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_LAYER)
4755 if (a == PIVOT_AXIS_COLUMN)
4757 size_t n = nest->n_areas[at];
4760 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4761 nest->n_areas[at]--;
4764 else if (a == PIVOT_AXIS_ROW)
4766 if (nest->n_areas[at] > 0)
4767 nest->n_areas[at]--;
4771 else if (at == CTAT_ROW && a == PIVOT_AXIS_COLUMN && t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_COLUMN)
4773 else if (at == CTAT_COL && a == PIVOT_AXIS_ROW && t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_COLUMN)
4775 size_t n = nest->n_areas[at];
4778 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4779 nest->n_areas[at]--;
4783 else if (at == CTAT_COL && a == PIVOT_AXIS_ROW && t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_ROW)
4785 else if (at == CTAT_ROW && a == PIVOT_AXIS_COLUMN && t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_ROW)
4787 size_t n = nest->n_areas[at];
4790 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4791 nest->n_areas[at]--;
4795 else if (at == CTAT_COL && a == PIVOT_AXIS_ROW && t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_LAYER)
4797 size_t n = nest->n_areas[at];
4800 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4801 nest->n_areas[at]--;
4805 else if (at == CTAT_ROW && a == PIVOT_AXIS_COLUMN && t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_LAYER)
4807 size_t n = nest->n_areas[at];
4810 nest->areas[at][n - 2] = nest->areas[at][n - 1];
4811 nest->n_areas[at]--;
4816 bool drop_last = (at == CTAT_SUBTABLE ? a != PIVOT_AXIS_LAYER
4817 : at == CTAT_ROW ? a == PIVOT_AXIS_COLUMN
4818 : at == CTAT_COL ? a == PIVOT_AXIS_ROW
4820 if (drop_last && nest->n_areas[at] > 0)
4821 nest->n_areas[at]--;
4823 bool drop_additional
4824 = ((t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_COLUMN && (at == CTAT_ROW || at == CTAT_LAYERROW) && a == PIVOT_AXIS_ROW)
4825 || (t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_ROW && (at == CTAT_COL || at == CTAT_LAYERCOL) && a == PIVOT_AXIS_COLUMN));
4826 if (drop_additional && nest->n_areas[at] > 0)
4827 nest->n_areas[at]--;
4833 struct ctables_nest *nest = xmalloc (sizeof *nest);
4834 *nest = (struct ctables_nest) { .n = 0 };
4835 t->stacks[a] = (struct ctables_stack) { .nests = nest, .n = 1 };
4837 /* There's no point in moving labels away from an axis that has no
4838 labels, so avoid dealing with the special cases around that. */
4839 t->label_axis[a] = a;
4842 struct ctables_stack *stack = &t->stacks[t->summary_axis];
4843 for (size_t i = 0; i < stack->n; i++)
4845 struct ctables_nest *nest = &stack->nests[i];
4846 if (!nest->specs[CSV_CELL].n)
4848 struct ctables_summary_spec_set *specs = &nest->specs[CSV_CELL];
4849 specs->specs = xmalloc (sizeof *specs->specs);
4852 enum ctables_summary_function function
4853 = specs->is_scale ? CTSF_MEAN : CTSF_COUNT;
4855 *specs->specs = (struct ctables_summary_spec) {
4856 .function = function,
4858 .format = ctables_summary_default_format (function, specs->var),
4861 specs->var = nest->vars[0];
4863 ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL],
4864 &nest->specs[CSV_CELL]);
4866 else if (!nest->specs[CSV_TOTAL].n)
4867 ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL],
4868 &nest->specs[CSV_CELL]);
4870 if (t->label_axis[PIVOT_AXIS_ROW] == PIVOT_AXIS_COLUMN
4871 || t->label_axis[PIVOT_AXIS_COLUMN] == PIVOT_AXIS_ROW)
4873 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
4874 for (size_t i = 0; i < nest->specs[sv].n; i++)
4876 struct ctables_summary_spec *ss = &nest->specs[sv].specs[i];
4877 const struct ctables_function_info *cfi =
4878 &ctables_function_info[ss->function];
4880 ss->calc_area = rotate_area (ss->calc_area);
4884 if (t->ctables->smissing_listwise)
4886 struct variable **listwise_vars = NULL;
4888 size_t allocated = 0;
4890 for (size_t j = nest->group_head; j < stack->n; j++)
4892 const struct ctables_nest *other_nest = &stack->nests[j];
4893 if (other_nest->group_head != nest->group_head)
4896 if (nest != other_nest && other_nest->scale_idx < other_nest->n)
4899 listwise_vars = x2nrealloc (listwise_vars, &allocated,
4900 sizeof *listwise_vars);
4901 listwise_vars[n++] = other_nest->vars[other_nest->scale_idx];
4904 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
4907 listwise_vars = xmemdup (listwise_vars,
4908 n * sizeof *listwise_vars);
4909 nest->specs[sv].listwise_vars = listwise_vars;
4910 nest->specs[sv].n_listwise_vars = n;
4915 struct ctables_summary_spec_set *merged = &t->summary_specs;
4916 struct merge_item *items = xnmalloc (N_CSVS * stack->n, sizeof *items);
4918 for (size_t j = 0; j < stack->n; j++)
4920 const struct ctables_nest *nest = &stack->nests[j];
4922 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
4923 items[n_left++] = (struct merge_item) { .set = &nest->specs[sv] };
4928 struct merge_item min = items[0];
4929 for (size_t j = 1; j < n_left; j++)
4930 if (merge_item_compare_3way (&items[j], &min) < 0)
4933 if (merged->n >= merged->allocated)
4934 merged->specs = x2nrealloc (merged->specs, &merged->allocated,
4935 sizeof *merged->specs);
4936 merged->specs[merged->n++] = min.set->specs[min.ofs];
4938 for (size_t j = 0; j < n_left; )
4940 if (merge_item_compare_3way (&items[j], &min) == 0)
4942 struct merge_item *item = &items[j];
4943 item->set->specs[item->ofs].axis_idx = merged->n - 1;
4944 if (++item->ofs >= item->set->n)
4946 items[j] = items[--n_left];
4956 for (size_t j = 0; j < merged->n; j++)
4957 printf ("%s\n", ctables_summary_function_name (merged->specs[j].function));
4959 for (size_t j = 0; j < stack->n; j++)
4961 const struct ctables_nest *nest = &stack->nests[j];
4962 for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
4964 const struct ctables_summary_spec_set *specs = &nest->specs[sv];
4965 for (size_t k = 0; k < specs->n; k++)
4966 printf ("(%s, %zu) ", ctables_summary_function_name (specs->specs[k].function),
4967 specs->specs[k].axis_idx);
4973 size_t allocated_sum_vars = 0;
4974 enumerate_sum_vars (t->axes[t->summary_axis],
4975 &t->sum_vars, &t->n_sum_vars, &allocated_sum_vars);
4977 return (ctables_check_label_position (t, PIVOT_AXIS_ROW)
4978 && ctables_check_label_position (t, PIVOT_AXIS_COLUMN));
4982 ctables_insert_clabels_values (struct ctables_table *t, const struct ccase *c,
4983 enum pivot_axis_type a)
4985 struct ctables_stack *stack = &t->stacks[a];
4986 for (size_t i = 0; i < stack->n; i++)
4988 const struct ctables_nest *nest = &stack->nests[i];
4989 const struct variable *var = nest->vars[nest->n - 1];
4990 const union value *value = case_data (c, var);
4992 if (var_is_numeric (var) && value->f == SYSMIS)
4995 if (ctables_categories_match (t->categories [var_get_dict_index (var)],
4997 ctables_value_insert (t, value, var_get_width (var));
5002 compare_clabels_values_3way (const void *a_, const void *b_, const void *width_)
5004 const struct ctables_value *const *ap = a_;
5005 const struct ctables_value *const *bp = b_;
5006 const struct ctables_value *a = *ap;
5007 const struct ctables_value *b = *bp;
5008 const int *width = width_;
5009 return value_compare_3way (&a->value, &b->value, *width);
5013 ctables_sort_clabels_values (struct ctables_table *t)
5015 const struct variable *v0 = t->clabels_example;
5016 int width = var_get_width (v0);
5018 struct ctables_categories *c0 = t->categories[var_get_dict_index (v0)];
5021 const struct val_labs *val_labs = var_get_value_labels (v0);
5022 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5023 vl = val_labs_next (val_labs, vl))
5024 if (ctables_categories_match (c0, &vl->value, v0))
5025 ctables_value_insert (t, &vl->value, width);
5028 size_t n = hmap_count (&t->clabels_values_map);
5029 t->clabels_values = xnmalloc (n, sizeof *t->clabels_values);
5031 struct ctables_value *clv;
5033 HMAP_FOR_EACH (clv, struct ctables_value, node, &t->clabels_values_map)
5034 t->clabels_values[i++] = clv;
5035 t->n_clabels_values = n;
5038 sort (t->clabels_values, n, sizeof *t->clabels_values,
5039 compare_clabels_values_3way, &width);
5041 for (size_t i = 0; i < n; i++)
5042 t->clabels_values[i]->leaf = i;
5046 ctables_add_category_occurrences (const struct variable *var,
5047 struct hmap *occurrences,
5048 const struct ctables_categories *cats)
5050 const struct val_labs *val_labs = var_get_value_labels (var);
5052 for (size_t i = 0; i < cats->n_cats; i++)
5054 const struct ctables_category *c = &cats->cats[i];
5058 ctables_add_occurrence (var, &(const union value) { .f = c->number },
5064 int width = var_get_width (var);
5066 value_init (&value, width);
5067 value_copy_buf_rpad (&value, width,
5068 CHAR_CAST (uint8_t *, c->string.string),
5069 c->string.length, ' ');
5070 ctables_add_occurrence (var, &value, occurrences);
5071 value_destroy (&value, width);
5076 assert (var_is_numeric (var));
5077 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5078 vl = val_labs_next (val_labs, vl))
5079 if (vl->value.f >= c->nrange[0] && vl->value.f <= c->nrange[1])
5080 ctables_add_occurrence (var, &vl->value, occurrences);
5084 assert (var_is_alpha (var));
5085 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5086 vl = val_labs_next (val_labs, vl))
5087 if (in_string_range (&vl->value, var, c->srange))
5088 ctables_add_occurrence (var, &vl->value, occurrences);
5092 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5093 vl = val_labs_next (val_labs, vl))
5094 if (var_is_value_missing (var, &vl->value))
5095 ctables_add_occurrence (var, &vl->value, occurrences);
5099 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5100 vl = val_labs_next (val_labs, vl))
5101 ctables_add_occurrence (var, &vl->value, occurrences);
5104 case CCT_POSTCOMPUTE:
5114 for (const struct val_lab *vl = val_labs_first (val_labs); vl;
5115 vl = val_labs_next (val_labs, vl))
5116 if (c->include_missing || !var_is_value_missing (var, &vl->value))
5117 ctables_add_occurrence (var, &vl->value, occurrences);
5120 case CCT_EXCLUDED_MISSING:
5127 ctables_section_recurse_add_empty_categories (
5128 struct ctables_section *s,
5129 const struct ctables_category *cats[PIVOT_N_AXES][10], struct ccase *c,
5130 enum pivot_axis_type a, size_t a_idx)
5132 if (a >= PIVOT_N_AXES)
5133 ctables_cell_insert__ (s, c, cats);
5134 else if (!s->nests[a] || a_idx >= s->nests[a]->n)
5135 ctables_section_recurse_add_empty_categories (s, cats, c, a + 1, 0);
5138 const struct variable *var = s->nests[a]->vars[a_idx];
5139 const struct ctables_categories *categories = s->table->categories[
5140 var_get_dict_index (var)];
5141 int width = var_get_width (var);
5142 const struct hmap *occurrences = &s->occurrences[a][a_idx];
5143 const struct ctables_occurrence *o;
5144 HMAP_FOR_EACH (o, struct ctables_occurrence, node, occurrences)
5146 union value *value = case_data_rw (c, var);
5147 value_destroy (value, width);
5148 value_clone (value, &o->value, width);
5149 cats[a][a_idx] = ctables_categories_match (categories, value, var);
5150 assert (cats[a][a_idx] != NULL);
5151 ctables_section_recurse_add_empty_categories (s, cats, c, a, a_idx + 1);
5154 for (size_t i = 0; i < categories->n_cats; i++)
5156 const struct ctables_category *cat = &categories->cats[i];
5157 if (cat->type == CCT_POSTCOMPUTE)
5159 cats[a][a_idx] = cat;
5160 ctables_section_recurse_add_empty_categories (s, cats, c, a, a_idx + 1);
5167 ctables_section_add_empty_categories (struct ctables_section *s)
5169 bool show_empty = false;
5170 for (size_t a = 0; a < PIVOT_N_AXES; a++)
5172 for (size_t k = 0; k < s->nests[a]->n; k++)
5173 if (k != s->nests[a]->scale_idx)
5175 const struct variable *var = s->nests[a]->vars[k];
5176 const struct ctables_categories *cats = s->table->categories[
5177 var_get_dict_index (var)];
5178 if (cats->show_empty)
5181 ctables_add_category_occurrences (var, &s->occurrences[a][k], cats);
5187 const struct ctables_category *cats[PIVOT_N_AXES][10]; /* XXX */
5188 struct ccase *c = case_create (dict_get_proto (s->table->ctables->dict));
5189 ctables_section_recurse_add_empty_categories (s, cats, c, 0, 0);
5194 ctables_section_clear (struct ctables_section *s)
5196 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
5198 const struct ctables_nest *nest = s->nests[a];
5199 for (size_t i = 0; i < nest->n; i++)
5200 if (i != nest->scale_idx)
5202 const struct variable *var = nest->vars[i];
5203 int width = var_get_width (var);
5204 struct ctables_occurrence *o, *next;
5205 struct hmap *map = &s->occurrences[a][i];
5206 HMAP_FOR_EACH_SAFE (o, next, struct ctables_occurrence, node, map)
5208 value_destroy (&o->value, width);
5209 hmap_delete (map, &o->node);
5216 struct ctables_cell *cell, *next_cell;
5217 HMAP_FOR_EACH_SAFE (cell, next_cell, struct ctables_cell, node, &s->cells)
5219 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
5221 const struct ctables_nest *nest = s->nests[a];
5222 for (size_t i = 0; i < nest->n; i++)
5223 if (i != nest->scale_idx)
5224 value_destroy (&cell->axes[a].cvs[i].value,
5225 var_get_width (nest->vars[i]));
5226 free (cell->axes[a].cvs);
5229 const struct ctables_nest *ss = s->nests[s->table->summary_axis];
5230 const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
5231 for (size_t i = 0; i < specs->n; i++)
5232 ctables_summary_uninit (&cell->summaries[i], &specs->specs[i]);
5233 free (cell->summaries);
5235 hmap_delete (&s->cells, &cell->node);
5238 hmap_shrink (&s->cells);
5240 for (enum ctables_area_type at = 0; at < N_CTATS; at++)
5242 struct ctables_area *area, *next_area;
5243 HMAP_FOR_EACH_SAFE (area, next_area, struct ctables_area, node,
5247 hmap_delete (&s->areas[at], &area->node);
5250 hmap_shrink (&s->areas[at]);
5255 ctables_section_uninit (struct ctables_section *s)
5257 ctables_section_clear (s);
5259 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
5261 struct ctables_nest *nest = s->nests[a];
5262 for (size_t i = 0; i < nest->n; i++)
5263 hmap_destroy (&s->occurrences[a][i]);
5264 free (s->occurrences[a]);
5267 hmap_destroy (&s->cells);
5268 for (size_t i = 0; i < N_CTATS; i++)
5269 hmap_destroy (&s->areas[i]);
5273 ctables_table_clear (struct ctables_table *t)
5275 for (size_t i = 0; i < t->n_sections; i++)
5276 ctables_section_clear (&t->sections[i]);
5278 if (t->clabels_example)
5280 int width = var_get_width (t->clabels_example);
5281 struct ctables_value *value, *next_value;
5282 HMAP_FOR_EACH_SAFE (value, next_value, struct ctables_value, node,
5283 &t->clabels_values_map)
5285 value_destroy (&value->value, width);
5286 hmap_delete (&t->clabels_values_map, &value->node);
5289 hmap_shrink (&t->clabels_values_map);
5291 free (t->clabels_values);
5292 t->clabels_values = NULL;
5293 t->n_clabels_values = 0;
5298 ctables_execute (struct dataset *ds, struct casereader *input,
5301 for (size_t i = 0; i < ct->n_tables; i++)
5303 struct ctables_table *t = ct->tables[i];
5304 t->sections = xnmalloc (MAX (1, t->stacks[PIVOT_AXIS_ROW].n) *
5305 MAX (1, t->stacks[PIVOT_AXIS_COLUMN].n) *
5306 MAX (1, t->stacks[PIVOT_AXIS_LAYER].n),
5307 sizeof *t->sections);
5308 size_t ix[PIVOT_N_AXES];
5309 ctables_table_add_section (t, 0, ix);
5312 struct dictionary *dict = dataset_dict (ds);
5314 bool splitting = dict_get_split_type (dict) == SPLIT_SEPARATE;
5315 struct casegrouper *grouper
5317 ? casegrouper_create_splits (input, dict)
5318 : casegrouper_create_vars (input, NULL, 0));
5319 struct casereader *group;
5320 while (casegrouper_get_next_group (grouper, &group))
5324 struct ccase *c = casereader_peek (group, 0);
5327 output_split_file_values (ds, c);
5332 bool warn_on_invalid = true;
5333 for (struct ccase *c = casereader_read (group); c;
5334 case_unref (c), c = casereader_read (group))
5336 double d_weight = dict_get_case_weight (dict, c, &warn_on_invalid);
5337 double e_weight = (ct->e_weight
5338 ? var_force_valid_weight (ct->e_weight,
5339 case_num (c, ct->e_weight),
5343 for (size_t i = 0; i < ct->n_tables; i++)
5345 struct ctables_table *t = ct->tables[i];
5347 for (size_t j = 0; j < t->n_sections; j++)
5348 ctables_cell_insert (&t->sections[j], c, d_weight, e_weight);
5350 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
5351 if (t->label_axis[a] != a)
5352 ctables_insert_clabels_values (t, c, a);
5355 casereader_destroy (group);
5357 for (size_t i = 0; i < ct->n_tables; i++)
5359 struct ctables_table *t = ct->tables[i];
5361 if (t->clabels_example)
5362 ctables_sort_clabels_values (t);
5364 for (size_t j = 0; j < t->n_sections; j++)
5365 ctables_section_add_empty_categories (&t->sections[j]);
5367 ctables_table_output (ct, t);
5368 ctables_table_clear (t);
5371 return casegrouper_destroy (grouper);
5376 typedef struct ctables_pcexpr *parse_recursively_func (struct lexer *,
5377 struct dictionary *);
5380 ctables_pcexpr_destroy (struct ctables_pcexpr *e)
5386 case CTPO_CAT_STRING:
5387 ss_dealloc (&e->string);
5390 case CTPO_CAT_SRANGE:
5391 for (size_t i = 0; i < 2; i++)
5392 ss_dealloc (&e->srange[i]);
5401 for (size_t i = 0; i < 2; i++)
5402 ctables_pcexpr_destroy (e->subs[i]);
5406 case CTPO_CAT_NUMBER:
5407 case CTPO_CAT_NRANGE:
5408 case CTPO_CAT_MISSING:
5409 case CTPO_CAT_OTHERNM:
5410 case CTPO_CAT_SUBTOTAL:
5411 case CTPO_CAT_TOTAL:
5415 msg_location_destroy (e->location);
5420 static struct ctables_pcexpr *
5421 ctables_pcexpr_allocate_binary (enum ctables_postcompute_op op,
5422 struct ctables_pcexpr *sub0,
5423 struct ctables_pcexpr *sub1)
5425 struct ctables_pcexpr *e = xmalloc (sizeof *e);
5426 *e = (struct ctables_pcexpr) {
5428 .subs = { sub0, sub1 },
5429 .location = msg_location_merged (sub0->location, sub1->location),
5434 /* How to parse an operator. */
5437 enum token_type token;
5438 enum ctables_postcompute_op op;
5441 static const struct operator *
5442 ctables_pcexpr_match_operator (struct lexer *lexer,
5443 const struct operator ops[], size_t n_ops)
5445 for (const struct operator *op = ops; op < ops + n_ops; op++)
5446 if (lex_token (lexer) == op->token)
5448 if (op->token != T_NEG_NUM)
5457 static struct ctables_pcexpr *
5458 ctables_pcexpr_parse_binary_operators__ (
5459 struct lexer *lexer, struct dictionary *dict,
5460 const struct operator ops[], size_t n_ops,
5461 parse_recursively_func *parse_next_level,
5462 const char *chain_warning, struct ctables_pcexpr *lhs)
5464 for (int op_count = 0; ; op_count++)
5466 const struct operator *op
5467 = ctables_pcexpr_match_operator (lexer, ops, n_ops);
5470 if (op_count > 1 && chain_warning)
5471 msg_at (SW, lhs->location, "%s", chain_warning);
5476 struct ctables_pcexpr *rhs = parse_next_level (lexer, dict);
5479 ctables_pcexpr_destroy (lhs);
5483 lhs = ctables_pcexpr_allocate_binary (op->op, lhs, rhs);
5487 static struct ctables_pcexpr *
5488 ctables_pcexpr_parse_binary_operators (
5489 struct lexer *lexer, struct dictionary *dict,
5490 const struct operator ops[], size_t n_ops,
5491 parse_recursively_func *parse_next_level, const char *chain_warning)
5493 struct ctables_pcexpr *lhs = parse_next_level (lexer, dict);
5497 return ctables_pcexpr_parse_binary_operators__ (lexer, dict, ops, n_ops,
5499 chain_warning, lhs);
5502 static struct ctables_pcexpr *ctables_pcexpr_parse_add (struct lexer *,
5503 struct dictionary *);
5505 static struct ctables_pcexpr
5506 ctpo_cat_nrange (double low, double high)
5508 return (struct ctables_pcexpr) {
5509 .op = CTPO_CAT_NRANGE,
5510 .nrange = { low, high },
5514 static struct ctables_pcexpr
5515 ctpo_cat_srange (struct substring low, struct substring high)
5517 return (struct ctables_pcexpr) {
5518 .op = CTPO_CAT_SRANGE,
5519 .srange = { low, high },
5523 static struct ctables_pcexpr *
5524 ctables_pcexpr_parse_primary (struct lexer *lexer, struct dictionary *dict)
5526 int start_ofs = lex_ofs (lexer);
5527 struct ctables_pcexpr e;
5528 if (lex_is_number (lexer))
5530 e = (struct ctables_pcexpr) { .op = CTPO_CONSTANT,
5531 .number = lex_number (lexer) };
5534 else if (lex_match_id (lexer, "MISSING"))
5535 e = (struct ctables_pcexpr) { .op = CTPO_CAT_MISSING };
5536 else if (lex_match_id (lexer, "OTHERNM"))
5537 e = (struct ctables_pcexpr) { .op = CTPO_CAT_OTHERNM };
5538 else if (lex_match_id (lexer, "TOTAL"))
5539 e = (struct ctables_pcexpr) { .op = CTPO_CAT_TOTAL };
5540 else if (lex_match_id (lexer, "SUBTOTAL"))
5542 size_t subtotal_index = 0;
5543 if (lex_match (lexer, T_LBRACK))
5545 if (!lex_force_int_range (lexer, "SUBTOTAL", 1, LONG_MAX))
5547 subtotal_index = lex_integer (lexer);
5549 if (!lex_force_match (lexer, T_RBRACK))
5552 e = (struct ctables_pcexpr) { .op = CTPO_CAT_SUBTOTAL,
5553 .subtotal_index = subtotal_index };
5555 else if (lex_match (lexer, T_LBRACK))
5557 if (lex_match_id (lexer, "LO"))
5559 if (!lex_force_match_id (lexer, "THRU"))
5562 if (lex_is_string (lexer))
5564 struct substring low = { .string = NULL };
5565 struct substring high = parse_substring (lexer, dict);
5566 e = ctpo_cat_srange (low, high);
5570 if (!lex_force_num (lexer))
5572 e = ctpo_cat_nrange (-DBL_MAX, lex_number (lexer));
5576 else if (lex_is_number (lexer))
5578 double number = lex_number (lexer);
5580 if (lex_match_id (lexer, "THRU"))
5582 if (lex_match_id (lexer, "HI"))
5583 e = ctpo_cat_nrange (number, DBL_MAX);
5586 if (!lex_force_num (lexer))
5588 e = ctpo_cat_nrange (number, lex_number (lexer));
5593 e = (struct ctables_pcexpr) { .op = CTPO_CAT_NUMBER,
5596 else if (lex_is_string (lexer))
5598 struct substring s = parse_substring (lexer, dict);
5600 if (lex_match_id (lexer, "THRU"))
5602 struct substring high;
5604 if (lex_match_id (lexer, "HI"))
5605 high = (struct substring) { .string = NULL };
5608 if (!lex_force_string (lexer))
5613 high = parse_substring (lexer, dict);
5616 e = ctpo_cat_srange (s, high);
5619 e = (struct ctables_pcexpr) { .op = CTPO_CAT_STRING, .string = s };
5623 lex_error (lexer, NULL);
5627 if (!lex_force_match (lexer, T_RBRACK))
5629 if (e.op == CTPO_CAT_STRING)
5630 ss_dealloc (&e.string);
5631 else if (e.op == CTPO_CAT_SRANGE)
5633 ss_dealloc (&e.srange[0]);
5634 ss_dealloc (&e.srange[1]);
5639 else if (lex_match (lexer, T_LPAREN))
5641 struct ctables_pcexpr *ep = ctables_pcexpr_parse_add (lexer, dict);
5644 if (!lex_force_match (lexer, T_RPAREN))
5646 ctables_pcexpr_destroy (ep);
5653 lex_error (lexer, NULL);
5657 e.location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1);
5658 return xmemdup (&e, sizeof e);
5661 static struct ctables_pcexpr *
5662 ctables_pcexpr_allocate_neg (struct ctables_pcexpr *sub,
5663 struct lexer *lexer, int start_ofs)
5665 struct ctables_pcexpr *e = xmalloc (sizeof *e);
5666 *e = (struct ctables_pcexpr) {
5669 .location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1),
5674 static struct ctables_pcexpr *
5675 ctables_pcexpr_parse_exp (struct lexer *lexer, struct dictionary *dict)
5677 static const struct operator op = { T_EXP, CTPO_POW };
5679 const char *chain_warning =
5680 _("The exponentiation operator (`**') is left-associative: "
5681 "`a**b**c' equals `(a**b)**c', not `a**(b**c)'. "
5682 "To disable this warning, insert parentheses.");
5684 if (lex_token (lexer) != T_NEG_NUM || lex_next_token (lexer, 1) != T_EXP)
5685 return ctables_pcexpr_parse_binary_operators (lexer, dict, &op, 1,
5686 ctables_pcexpr_parse_primary,
5689 /* Special case for situations like "-5**6", which must be parsed as
5692 int start_ofs = lex_ofs (lexer);
5693 struct ctables_pcexpr *lhs = xmalloc (sizeof *lhs);
5694 *lhs = (struct ctables_pcexpr) {
5695 .op = CTPO_CONSTANT,
5696 .number = -lex_tokval (lexer),
5697 .location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer)),
5701 struct ctables_pcexpr *node = ctables_pcexpr_parse_binary_operators__ (
5702 lexer, dict, &op, 1,
5703 ctables_pcexpr_parse_primary, chain_warning, lhs);
5707 return ctables_pcexpr_allocate_neg (node, lexer, start_ofs);
5710 /* Parses the unary minus level. */
5711 static struct ctables_pcexpr *
5712 ctables_pcexpr_parse_neg (struct lexer *lexer, struct dictionary *dict)
5714 int start_ofs = lex_ofs (lexer);
5715 if (!lex_match (lexer, T_DASH))
5716 return ctables_pcexpr_parse_exp (lexer, dict);
5718 struct ctables_pcexpr *inner = ctables_pcexpr_parse_neg (lexer, dict);
5722 return ctables_pcexpr_allocate_neg (inner, lexer, start_ofs);
5725 /* Parses the multiplication and division level. */
5726 static struct ctables_pcexpr *
5727 ctables_pcexpr_parse_mul (struct lexer *lexer, struct dictionary *dict)
5729 static const struct operator ops[] =
5731 { T_ASTERISK, CTPO_MUL },
5732 { T_SLASH, CTPO_DIV },
5735 return ctables_pcexpr_parse_binary_operators (lexer, dict, ops,
5736 sizeof ops / sizeof *ops,
5737 ctables_pcexpr_parse_neg, NULL);
5740 /* Parses the addition and subtraction level. */
5741 static struct ctables_pcexpr *
5742 ctables_pcexpr_parse_add (struct lexer *lexer, struct dictionary *dict)
5744 static const struct operator ops[] =
5746 { T_PLUS, CTPO_ADD },
5747 { T_DASH, CTPO_SUB },
5748 { T_NEG_NUM, CTPO_ADD },
5751 return ctables_pcexpr_parse_binary_operators (lexer, dict,
5752 ops, sizeof ops / sizeof *ops,
5753 ctables_pcexpr_parse_mul, NULL);
5756 static struct ctables_postcompute *
5757 ctables_find_postcompute (struct ctables *ct, const char *name)
5759 struct ctables_postcompute *pc;
5760 HMAP_FOR_EACH_WITH_HASH (pc, struct ctables_postcompute, hmap_node,
5761 utf8_hash_case_string (name, 0), &ct->postcomputes)
5762 if (!utf8_strcasecmp (pc->name, name))
5768 ctables_parse_pcompute (struct lexer *lexer, struct dictionary *dict,
5771 int pcompute_start = lex_ofs (lexer) - 1;
5773 if (!lex_match (lexer, T_AND))
5775 lex_error_expecting (lexer, "&");
5778 if (!lex_force_id (lexer))
5781 char *name = ss_xstrdup (lex_tokss (lexer));
5784 if (!lex_force_match (lexer, T_EQUALS)
5785 || !lex_force_match_id (lexer, "EXPR")
5786 || !lex_force_match (lexer, T_LPAREN))
5792 int expr_start = lex_ofs (lexer);
5793 struct ctables_pcexpr *expr = ctables_pcexpr_parse_add (lexer, dict);
5794 int expr_end = lex_ofs (lexer) - 1;
5795 if (!expr || !lex_force_match (lexer, T_RPAREN))
5797 ctables_pcexpr_destroy (expr);
5801 int pcompute_end = lex_ofs (lexer) - 1;
5803 struct msg_location *location = lex_ofs_location (lexer, pcompute_start,
5806 struct ctables_postcompute *pc = ctables_find_postcompute (ct, name);
5809 msg_at (SW, location, _("New definition of &%s will override the "
5810 "previous definition."),
5812 msg_at (SN, pc->location, _("This is the previous definition."));
5814 ctables_pcexpr_destroy (pc->expr);
5815 msg_location_destroy (pc->location);
5820 pc = xmalloc (sizeof *pc);
5821 *pc = (struct ctables_postcompute) { .name = name };
5822 hmap_insert (&ct->postcomputes, &pc->hmap_node,
5823 utf8_hash_case_string (pc->name, 0));
5826 pc->location = location;
5828 pc->label = lex_ofs_representation (lexer, expr_start, expr_end);
5833 ctables_parse_pproperties_format (struct lexer *lexer,
5834 struct ctables_summary_spec_set *sss)
5836 *sss = (struct ctables_summary_spec_set) { .n = 0 };
5838 while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH
5839 && !(lex_token (lexer) == T_ID
5840 && (lex_id_match (ss_cstr ("LABEL"), lex_tokss (lexer))
5841 || lex_id_match (ss_cstr ("HIDESOURCECATS"),
5842 lex_tokss (lexer)))))
5844 /* Parse function. */
5845 enum ctables_summary_function function;
5847 enum ctables_area_type area;
5848 if (!parse_ctables_summary_function (lexer, &function, &weighted, &area))
5851 /* Parse percentile. */
5852 double percentile = 0;
5853 if (function == CTSF_PTILE)
5855 if (!lex_force_num_range_closed (lexer, "PTILE", 0, 100))
5857 percentile = lex_number (lexer);
5862 struct fmt_spec format;
5863 bool is_ctables_format;
5864 if (!parse_ctables_format_specifier (lexer, &format, &is_ctables_format))
5867 if (sss->n >= sss->allocated)
5868 sss->specs = x2nrealloc (sss->specs, &sss->allocated,
5869 sizeof *sss->specs);
5870 sss->specs[sss->n++] = (struct ctables_summary_spec) {
5871 .function = function,
5872 .weighted = weighted,
5875 .percentile = percentile,
5877 .is_ctables_format = is_ctables_format,
5883 ctables_summary_spec_set_uninit (sss);
5888 ctables_parse_pproperties (struct lexer *lexer, struct ctables *ct)
5890 struct ctables_postcompute **pcs = NULL;
5892 size_t allocated_pcs = 0;
5894 while (lex_match (lexer, T_AND))
5896 if (!lex_force_id (lexer))
5898 struct ctables_postcompute *pc
5899 = ctables_find_postcompute (ct, lex_tokcstr (lexer));
5902 msg (SE, _("Unknown computed category &%s."), lex_tokcstr (lexer));
5907 if (n_pcs >= allocated_pcs)
5908 pcs = x2nrealloc (pcs, &allocated_pcs, sizeof *pcs);
5912 while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
5914 if (lex_match_id (lexer, "LABEL"))
5916 lex_match (lexer, T_EQUALS);
5917 if (!lex_force_string (lexer))
5920 for (size_t i = 0; i < n_pcs; i++)
5922 free (pcs[i]->label);
5923 pcs[i]->label = ss_xstrdup (lex_tokss (lexer));
5928 else if (lex_match_id (lexer, "FORMAT"))
5930 lex_match (lexer, T_EQUALS);
5932 struct ctables_summary_spec_set sss;
5933 if (!ctables_parse_pproperties_format (lexer, &sss))
5936 for (size_t i = 0; i < n_pcs; i++)
5939 ctables_summary_spec_set_uninit (pcs[i]->specs);
5941 pcs[i]->specs = xmalloc (sizeof *pcs[i]->specs);
5942 ctables_summary_spec_set_clone (pcs[i]->specs, &sss);
5944 ctables_summary_spec_set_uninit (&sss);
5946 else if (lex_match_id (lexer, "HIDESOURCECATS"))
5948 lex_match (lexer, T_EQUALS);
5949 bool hide_source_cats;
5950 if (!parse_bool (lexer, &hide_source_cats))
5952 for (size_t i = 0; i < n_pcs; i++)
5953 pcs[i]->hide_source_cats = hide_source_cats;
5957 lex_error_expecting (lexer, "LABEL", "FORMAT", "HIDESOURCECATS");
5970 put_strftime (struct string *out, time_t now, const char *format)
5972 const struct tm *tm = localtime (&now);
5974 strftime (value, sizeof value, format, tm);
5975 ds_put_cstr (out, value);
5979 skip_prefix (struct substring *s, struct substring prefix)
5981 if (ss_starts_with (*s, prefix))
5983 ss_advance (s, prefix.length);
5991 put_table_expression (struct string *out, struct lexer *lexer,
5992 struct dictionary *dict, int expr_start, int expr_end)
5995 for (int ofs = expr_start; ofs < expr_end; ofs++)
5997 const struct token *t = lex_ofs_token (lexer, ofs);
5998 if (t->type == T_LBRACK)
6000 else if (t->type == T_RBRACK && nest > 0)
6006 else if (t->type == T_ID)
6008 const struct variable *var
6009 = dict_lookup_var (dict, t->string.string);
6010 const char *label = var ? var_get_label (var) : NULL;
6011 ds_put_cstr (out, label ? label : t->string.string);
6015 if (ofs != expr_start && t->type != T_RPAREN && ds_last (out) != ' ')
6016 ds_put_byte (out, ' ');
6018 char *repr = lex_ofs_representation (lexer, ofs, ofs);
6019 ds_put_cstr (out, repr);
6022 if (ofs + 1 != expr_end && t->type != T_LPAREN)
6023 ds_put_byte (out, ' ');
6029 put_title_text (struct string *out, struct substring in, time_t now,
6030 struct lexer *lexer, struct dictionary *dict,
6031 int expr_start, int expr_end)
6035 size_t chunk = ss_find_byte (in, ')');
6036 ds_put_substring (out, ss_head (in, chunk));
6037 ss_advance (&in, chunk);
6038 if (ss_is_empty (in))
6041 if (skip_prefix (&in, ss_cstr (")DATE")))
6042 put_strftime (out, now, "%x");
6043 else if (skip_prefix (&in, ss_cstr (")TIME")))
6044 put_strftime (out, now, "%X");
6045 else if (skip_prefix (&in, ss_cstr (")TABLE")))
6046 put_table_expression (out, lexer, dict, expr_start, expr_end);
6049 ds_put_byte (out, ')');
6050 ss_advance (&in, 1);
6056 cmd_ctables (struct lexer *lexer, struct dataset *ds)
6058 struct casereader *input = NULL;
6060 struct measure_guesser *mg = measure_guesser_create (ds);
6063 input = proc_open (ds);
6064 measure_guesser_run (mg, input);
6065 measure_guesser_destroy (mg);
6068 size_t n_vars = dict_get_n_vars (dataset_dict (ds));
6069 enum ctables_vlabel *vlabels = xnmalloc (n_vars, sizeof *vlabels);
6070 enum settings_value_show tvars = settings_get_show_variables ();
6071 for (size_t i = 0; i < n_vars; i++)
6072 vlabels[i] = (enum ctables_vlabel) tvars;
6074 struct pivot_table_look *look = pivot_table_look_unshare (
6075 pivot_table_look_ref (pivot_table_look_get_default ()));
6076 look->omit_empty = false;
6078 struct ctables *ct = xmalloc (sizeof *ct);
6079 *ct = (struct ctables) {
6080 .dict = dataset_dict (ds),
6082 .ctables_formats = FMT_SETTINGS_INIT,
6084 .postcomputes = HMAP_INITIALIZER (ct->postcomputes),
6087 time_t now = time (NULL);
6092 const char *dot_string;
6093 const char *comma_string;
6095 static const struct ctf ctfs[4] = {
6096 { CTEF_NEGPAREN, "(,,,)", "(...)" },
6097 { CTEF_NEQUAL, "-,N=,,", "-.N=.." },
6098 { CTEF_PAREN, "-,(,),", "-.(.)." },
6099 { CTEF_PCTPAREN, "-,(,%),", "-.(.%)." },
6101 bool is_dot = settings_get_fmt_settings ()->decimal == '.';
6102 for (size_t i = 0; i < 4; i++)
6104 const char *s = is_dot ? ctfs[i].dot_string : ctfs[i].comma_string;
6105 fmt_settings_set_cc (&ct->ctables_formats, ctfs[i].type,
6106 fmt_number_style_from_string (s));
6109 if (!lex_force_match (lexer, T_SLASH))
6112 while (!lex_match_id (lexer, "TABLE"))
6114 if (lex_match_id (lexer, "FORMAT"))
6116 double widths[2] = { SYSMIS, SYSMIS };
6117 double units_per_inch = 72.0;
6119 while (lex_token (lexer) != T_SLASH)
6121 if (lex_match_id (lexer, "MINCOLWIDTH"))
6123 if (!parse_col_width (lexer, "MINCOLWIDTH", &widths[0]))
6126 else if (lex_match_id (lexer, "MAXCOLWIDTH"))
6128 if (!parse_col_width (lexer, "MAXCOLWIDTH", &widths[1]))
6131 else if (lex_match_id (lexer, "UNITS"))
6133 lex_match (lexer, T_EQUALS);
6134 if (lex_match_id (lexer, "POINTS"))
6135 units_per_inch = 72.0;
6136 else if (lex_match_id (lexer, "INCHES"))
6137 units_per_inch = 1.0;
6138 else if (lex_match_id (lexer, "CM"))
6139 units_per_inch = 2.54;
6142 lex_error_expecting (lexer, "POINTS", "INCHES", "CM");
6146 else if (lex_match_id (lexer, "EMPTY"))
6151 lex_match (lexer, T_EQUALS);
6152 if (lex_match_id (lexer, "ZERO"))
6154 /* Nothing to do. */
6156 else if (lex_match_id (lexer, "BLANK"))
6157 ct->zero = xstrdup ("");
6158 else if (lex_force_string (lexer))
6160 ct->zero = ss_xstrdup (lex_tokss (lexer));
6166 else if (lex_match_id (lexer, "MISSING"))
6168 lex_match (lexer, T_EQUALS);
6169 if (!lex_force_string (lexer))
6173 ct->missing = (strcmp (lex_tokcstr (lexer), ".")
6174 ? ss_xstrdup (lex_tokss (lexer))
6180 lex_error_expecting (lexer, "MINCOLWIDTH", "MAXCOLWIDTH",
6181 "UNITS", "EMPTY", "MISSING");
6186 if (widths[0] != SYSMIS && widths[1] != SYSMIS
6187 && widths[0] > widths[1])
6189 msg (SE, _("MINCOLWIDTH must not be greater than MAXCOLWIDTH."));
6193 for (size_t i = 0; i < 2; i++)
6194 if (widths[i] != SYSMIS)
6196 int *wr = ct->look->width_ranges[TABLE_HORZ];
6197 wr[i] = widths[i] / units_per_inch * 96.0;
6202 else if (lex_match_id (lexer, "VLABELS"))
6204 if (!lex_force_match_id (lexer, "VARIABLES"))
6206 lex_match (lexer, T_EQUALS);
6208 struct variable **vars;
6210 if (!parse_variables (lexer, dataset_dict (ds), &vars, &n_vars,
6214 if (!lex_force_match_id (lexer, "DISPLAY"))
6219 lex_match (lexer, T_EQUALS);
6221 enum ctables_vlabel vlabel;
6222 if (lex_match_id (lexer, "DEFAULT"))
6223 vlabel = (enum ctables_vlabel) settings_get_show_variables ();
6224 else if (lex_match_id (lexer, "NAME"))
6226 else if (lex_match_id (lexer, "LABEL"))
6227 vlabel = CTVL_LABEL;
6228 else if (lex_match_id (lexer, "BOTH"))
6230 else if (lex_match_id (lexer, "NONE"))
6234 lex_error_expecting (lexer, "DEFAULT", "NAME", "LABEL",
6240 for (size_t i = 0; i < n_vars; i++)
6241 ct->vlabels[var_get_dict_index (vars[i])] = vlabel;
6244 else if (lex_match_id (lexer, "MRSETS"))
6246 if (!lex_force_match_id (lexer, "COUNTDUPLICATES"))
6248 lex_match (lexer, T_EQUALS);
6249 if (!parse_bool (lexer, &ct->mrsets_count_duplicates))
6252 else if (lex_match_id (lexer, "SMISSING"))
6254 if (lex_match_id (lexer, "VARIABLE"))
6255 ct->smissing_listwise = false;
6256 else if (lex_match_id (lexer, "LISTWISE"))
6257 ct->smissing_listwise = true;
6260 lex_error_expecting (lexer, "VARIABLE", "LISTWISE");
6264 else if (lex_match_id (lexer, "PCOMPUTE"))
6266 if (!ctables_parse_pcompute (lexer, dataset_dict (ds), ct))
6269 else if (lex_match_id (lexer, "PPROPERTIES"))
6271 if (!ctables_parse_pproperties (lexer, ct))
6274 else if (lex_match_id (lexer, "WEIGHT"))
6276 if (!lex_force_match_id (lexer, "VARIABLE"))
6278 lex_match (lexer, T_EQUALS);
6279 ct->e_weight = parse_variable (lexer, dataset_dict (ds));
6283 else if (lex_match_id (lexer, "HIDESMALLCOUNTS"))
6285 if (lex_match_id (lexer, "COUNT"))
6287 lex_match (lexer, T_EQUALS);
6288 if (!lex_force_int_range (lexer, "HIDESMALLCOUNTS COUNT",
6291 ct->hide_threshold = lex_integer (lexer);
6294 else if (ct->hide_threshold == 0)
6295 ct->hide_threshold = 5;
6299 lex_error_expecting (lexer, "FORMAT", "VLABELS", "MRSETS",
6300 "SMISSING", "PCOMPUTE", "PPROPERTIES",
6301 "WEIGHT", "HIDESMALLCOUNTS", "TABLE");
6305 if (!lex_force_match (lexer, T_SLASH))
6309 size_t allocated_tables = 0;
6312 if (ct->n_tables >= allocated_tables)
6313 ct->tables = x2nrealloc (ct->tables, &allocated_tables,
6314 sizeof *ct->tables);
6316 struct ctables_category *cat = xmalloc (sizeof *cat);
6317 *cat = (struct ctables_category) {
6319 .include_missing = false,
6320 .sort_ascending = true,
6323 struct ctables_categories *c = xmalloc (sizeof *c);
6324 size_t n_vars = dict_get_n_vars (dataset_dict (ds));
6325 *c = (struct ctables_categories) {
6332 struct ctables_categories **categories = xnmalloc (n_vars,
6333 sizeof *categories);
6334 for (size_t i = 0; i < n_vars; i++)
6337 struct ctables_table *t = xmalloc (sizeof *t);
6338 *t = (struct ctables_table) {
6340 .slabels_axis = PIVOT_AXIS_COLUMN,
6341 .slabels_visible = true,
6342 .clabels_values_map = HMAP_INITIALIZER (t->clabels_values_map),
6344 [PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW,
6345 [PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN,
6346 [PIVOT_AXIS_LAYER] = PIVOT_AXIS_LAYER,
6348 .clabels_from_axis = PIVOT_AXIS_LAYER,
6349 .categories = categories,
6350 .n_categories = n_vars,
6353 ct->tables[ct->n_tables++] = t;
6355 lex_match (lexer, T_EQUALS);
6356 int expr_start = lex_ofs (lexer);
6357 if (!ctables_axis_parse (lexer, dataset_dict (ds), ct, t, PIVOT_AXIS_ROW))
6359 if (lex_match (lexer, T_BY))
6361 if (!ctables_axis_parse (lexer, dataset_dict (ds),
6362 ct, t, PIVOT_AXIS_COLUMN))
6365 if (lex_match (lexer, T_BY))
6367 if (!ctables_axis_parse (lexer, dataset_dict (ds),
6368 ct, t, PIVOT_AXIS_LAYER))
6372 int expr_end = lex_ofs (lexer);
6374 if (!t->axes[PIVOT_AXIS_ROW] && !t->axes[PIVOT_AXIS_COLUMN]
6375 && !t->axes[PIVOT_AXIS_LAYER])
6377 lex_error (lexer, _("At least one variable must be specified."));
6381 const struct ctables_axis *scales[PIVOT_N_AXES];
6382 size_t n_scales = 0;
6383 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
6385 scales[a] = find_scale (t->axes[a]);
6391 msg (SE, _("Scale variables may appear only on one axis."));
6392 if (scales[PIVOT_AXIS_ROW])
6393 msg_at (SN, scales[PIVOT_AXIS_ROW]->loc,
6394 _("This scale variable appears on the rows axis."));
6395 if (scales[PIVOT_AXIS_COLUMN])
6396 msg_at (SN, scales[PIVOT_AXIS_COLUMN]->loc,
6397 _("This scale variable appears on the columns axis."));
6398 if (scales[PIVOT_AXIS_LAYER])
6399 msg_at (SN, scales[PIVOT_AXIS_LAYER]->loc,
6400 _("This scale variable appears on the layer axis."));
6404 const struct ctables_axis *summaries[PIVOT_N_AXES];
6405 size_t n_summaries = 0;
6406 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
6408 summaries[a] = (scales[a]
6410 : find_categorical_summary_spec (t->axes[a]));
6414 if (n_summaries > 1)
6416 msg (SE, _("Summaries may appear only on one axis."));
6417 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
6420 msg_at (SN, summaries[a]->loc,
6422 ? _("This variable on the rows axis has a summary.")
6423 : a == PIVOT_AXIS_COLUMN
6424 ? _("This variable on the columns axis has a summary.")
6425 : _("This variable on the layers axis has a summary."));
6427 msg_at (SN, summaries[a]->loc,
6428 _("This is a scale variable, so it always has a "
6429 "summary even if the syntax does not explicitly "
6434 for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
6435 if (n_summaries ? summaries[a] : t->axes[a])
6437 t->summary_axis = a;
6441 if (lex_token (lexer) == T_ENDCMD)
6443 if (!ctables_prepare_table (t))
6447 if (!lex_force_match (lexer, T_SLASH))
6450 while (!lex_match_id (lexer, "TABLE") && lex_token (lexer) != T_ENDCMD)
6452 if (lex_match_id (lexer, "SLABELS"))
6454 while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
6456 if (lex_match_id (lexer, "POSITION"))
6458 lex_match (lexer, T_EQUALS);
6459 if (lex_match_id (lexer, "COLUMN"))
6460 t->slabels_axis = PIVOT_AXIS_COLUMN;
6461 else if (lex_match_id (lexer, "ROW"))
6462 t->slabels_axis = PIVOT_AXIS_ROW;
6463 else if (lex_match_id (lexer, "LAYER"))
6464 t->slabels_axis = PIVOT_AXIS_LAYER;
6467 lex_error_expecting (lexer, "COLUMN", "ROW", "LAYER");
6471 else if (lex_match_id (lexer, "VISIBLE"))
6473 lex_match (lexer, T_EQUALS);
6474 if (!parse_bool (lexer, &t->slabels_visible))
6479 lex_error_expecting (lexer, "POSITION", "VISIBLE");
6484 else if (lex_match_id (lexer, "CLABELS"))
6486 if (lex_match_id (lexer, "AUTO"))
6488 t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW;
6489 t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN;
6491 else if (lex_match_id (lexer, "ROWLABELS"))
6493 lex_match (lexer, T_EQUALS);
6494 if (lex_match_id (lexer, "OPPOSITE"))
6495 t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_COLUMN;
6496 else if (lex_match_id (lexer, "LAYER"))
6497 t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_LAYER;
6500 lex_error_expecting (lexer, "OPPOSITE", "LAYER");
6504 else if (lex_match_id (lexer, "COLLABELS"))
6506 lex_match (lexer, T_EQUALS);
6507 if (lex_match_id (lexer, "OPPOSITE"))
6508 t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_ROW;
6509 else if (lex_match_id (lexer, "LAYER"))
6510 t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_LAYER;
6513 lex_error_expecting (lexer, "OPPOSITE", "LAYER");
6519 lex_error_expecting (lexer, "AUTO", "ROWLABELS",
6524 else if (lex_match_id (lexer, "CRITERIA"))
6526 if (!lex_force_match_id (lexer, "CILEVEL"))
6528 lex_match (lexer, T_EQUALS);
6530 if (!lex_force_num_range_halfopen (lexer, "CILEVEL", 0, 100))
6532 t->cilevel = lex_number (lexer);
6535 else if (lex_match_id (lexer, "CATEGORIES"))
6537 if (!ctables_table_parse_categories (lexer, dataset_dict (ds),
6541 else if (lex_match_id (lexer, "TITLES"))
6546 if (lex_match_id (lexer, "CAPTION"))
6547 textp = &t->caption;
6548 else if (lex_match_id (lexer, "CORNER"))
6550 else if (lex_match_id (lexer, "TITLE"))
6554 lex_error_expecting (lexer, "CAPTION", "CORNER", "TITLE");
6557 lex_match (lexer, T_EQUALS);
6559 struct string s = DS_EMPTY_INITIALIZER;
6560 while (lex_is_string (lexer))
6562 if (!ds_is_empty (&s))
6563 ds_put_byte (&s, ' ');
6564 put_title_text (&s, lex_tokss (lexer), now,
6565 lexer, dataset_dict (ds),
6566 expr_start, expr_end);
6570 *textp = ds_steal_cstr (&s);
6572 while (lex_token (lexer) != T_SLASH
6573 && lex_token (lexer) != T_ENDCMD);
6575 else if (lex_match_id (lexer, "SIGTEST"))
6579 t->chisq = xmalloc (sizeof *t->chisq);
6580 *t->chisq = (struct ctables_chisq) {
6582 .include_mrsets = true,
6583 .all_visible = true,
6589 if (lex_match_id (lexer, "TYPE"))
6591 lex_match (lexer, T_EQUALS);
6592 if (!lex_force_match_id (lexer, "CHISQUARE"))
6595 else if (lex_match_id (lexer, "ALPHA"))
6597 lex_match (lexer, T_EQUALS);
6598 if (!lex_force_num_range_halfopen (lexer, "ALPHA", 0, 1))
6600 t->chisq->alpha = lex_number (lexer);
6603 else if (lex_match_id (lexer, "INCLUDEMRSETS"))
6605 lex_match (lexer, T_EQUALS);
6606 if (!parse_bool (lexer, &t->chisq->include_mrsets))
6609 else if (lex_match_id (lexer, "CATEGORIES"))
6611 lex_match (lexer, T_EQUALS);
6612 if (lex_match_id (lexer, "ALLVISIBLE"))
6613 t->chisq->all_visible = true;
6614 else if (lex_match_id (lexer, "SUBTOTALS"))
6615 t->chisq->all_visible = false;
6618 lex_error_expecting (lexer,
6619 "ALLVISIBLE", "SUBTOTALS");
6625 lex_error_expecting (lexer, "TYPE", "ALPHA",
6626 "INCLUDEMRSETS", "CATEGORIES");
6630 while (lex_token (lexer) != T_SLASH
6631 && lex_token (lexer) != T_ENDCMD);
6633 else if (lex_match_id (lexer, "COMPARETEST"))
6637 t->pairwise = xmalloc (sizeof *t->pairwise);
6638 *t->pairwise = (struct ctables_pairwise) {
6640 .alpha = { .05, .05 },
6641 .adjust = BONFERRONI,
6642 .include_mrsets = true,
6643 .meansvariance_allcats = true,
6644 .all_visible = true,
6653 if (lex_match_id (lexer, "TYPE"))
6655 lex_match (lexer, T_EQUALS);
6656 if (lex_match_id (lexer, "PROP"))
6657 t->pairwise->type = PROP;
6658 else if (lex_match_id (lexer, "MEAN"))
6659 t->pairwise->type = MEAN;
6662 lex_error_expecting (lexer, "PROP", "MEAN");
6666 else if (lex_match_id (lexer, "ALPHA"))
6668 lex_match (lexer, T_EQUALS);
6670 if (!lex_force_num_range_open (lexer, "ALPHA", 0, 1))
6672 double a0 = lex_number (lexer);
6675 lex_match (lexer, T_COMMA);
6676 if (lex_is_number (lexer))
6678 if (!lex_force_num_range_open (lexer, "ALPHA", 0, 1))
6680 double a1 = lex_number (lexer);
6683 t->pairwise->alpha[0] = MIN (a0, a1);
6684 t->pairwise->alpha[1] = MAX (a0, a1);
6687 t->pairwise->alpha[0] = t->pairwise->alpha[1] = a0;
6689 else if (lex_match_id (lexer, "ADJUST"))
6691 lex_match (lexer, T_EQUALS);
6692 if (lex_match_id (lexer, "BONFERRONI"))
6693 t->pairwise->adjust = BONFERRONI;
6694 else if (lex_match_id (lexer, "BH"))
6695 t->pairwise->adjust = BH;
6696 else if (lex_match_id (lexer, "NONE"))
6697 t->pairwise->adjust = 0;
6700 lex_error_expecting (lexer, "BONFERRONI", "BH",
6705 else if (lex_match_id (lexer, "INCLUDEMRSETS"))
6707 lex_match (lexer, T_EQUALS);
6708 if (!parse_bool (lexer, &t->pairwise->include_mrsets))
6711 else if (lex_match_id (lexer, "MEANSVARIANCE"))
6713 lex_match (lexer, T_EQUALS);
6714 if (lex_match_id (lexer, "ALLCATS"))
6715 t->pairwise->meansvariance_allcats = true;
6716 else if (lex_match_id (lexer, "TESTEDCATS"))
6717 t->pairwise->meansvariance_allcats = false;
6720 lex_error_expecting (lexer, "ALLCATS", "TESTEDCATS");
6724 else if (lex_match_id (lexer, "CATEGORIES"))
6726 lex_match (lexer, T_EQUALS);
6727 if (lex_match_id (lexer, "ALLVISIBLE"))
6728 t->pairwise->all_visible = true;
6729 else if (lex_match_id (lexer, "SUBTOTALS"))
6730 t->pairwise->all_visible = false;
6733 lex_error_expecting (lexer, "ALLVISIBLE",
6738 else if (lex_match_id (lexer, "MERGE"))
6740 lex_match (lexer, T_EQUALS);
6741 if (!parse_bool (lexer, &t->pairwise->merge))
6744 else if (lex_match_id (lexer, "STYLE"))
6746 lex_match (lexer, T_EQUALS);
6747 if (lex_match_id (lexer, "APA"))
6748 t->pairwise->apa_style = true;
6749 else if (lex_match_id (lexer, "SIMPLE"))
6750 t->pairwise->apa_style = false;
6753 lex_error_expecting (lexer, "APA", "SIMPLE");
6757 else if (lex_match_id (lexer, "SHOWSIG"))
6759 lex_match (lexer, T_EQUALS);
6760 if (!parse_bool (lexer, &t->pairwise->show_sig))
6765 lex_error_expecting (lexer, "TYPE", "ALPHA", "ADJUST",
6766 "INCLUDEMRSETS", "MEANSVARIANCE",
6767 "CATEGORIES", "MERGE", "STYLE",
6772 while (lex_token (lexer) != T_SLASH
6773 && lex_token (lexer) != T_ENDCMD);
6777 lex_error_expecting (lexer, "TABLE", "SLABELS", "CLABELS",
6778 "CRITERIA", "CATEGORIES", "TITLES",
6779 "SIGTEST", "COMPARETEST");
6783 if (!lex_match (lexer, T_SLASH))
6787 if (t->label_axis[PIVOT_AXIS_ROW] != PIVOT_AXIS_ROW
6788 && t->label_axis[PIVOT_AXIS_COLUMN] != PIVOT_AXIS_COLUMN)
6790 msg (SE, _("ROWLABELS and COLLABELS may not both be specified."));
6794 if (!ctables_prepare_table (t))
6797 while (lex_token (lexer) != T_ENDCMD);
6800 input = proc_open (ds);
6801 bool ok = ctables_execute (ds, input, ct);
6802 ok = proc_commit (ds) && ok;
6804 ctables_destroy (ct);
6805 return ok ? CMD_SUCCESS : CMD_FAILURE;
6810 ctables_destroy (ct);