#include <math.h>
#include "data/casereader.h"
+#include "data/casewriter.h"
#include "data/dataset.h"
#include "data/dictionary.h"
#include "data/mrset.h"
+#include "data/subcase.h"
+#include "data/value-labels.h"
#include "language/command.h"
#include "language/lexer/format-parser.h"
#include "language/lexer/lexer.h"
#include "libpspp/assertion.h"
#include "libpspp/hash-functions.h"
#include "libpspp/hmap.h"
+#include "libpspp/i18n.h"
#include "libpspp/message.h"
#include "libpspp/string-array.h"
+#include "math/mode.h"
#include "math/moments.h"
+#include "math/percentiles.h"
+#include "math/sort.h"
#include "output/pivot-table.h"
#include "gl/minmax.h"
const struct ctables_cell *example;
- double valid;
- double missing;
+ double d_valid; /* Dictionary weight. */
+ double d_missing;
+ double e_valid; /* Effective weight */
+ double e_missing;
+ };
+
+enum ctables_summary_variant
+ {
+ CSV_CELL,
+ CSV_TOTAL
+#define N_CSVS 2
};
struct ctables_cell
{
- /* In struct ctables's 'cells' hmap. Indexed by all the values in all the
- axes (except the scalar variable, if any). */
+ /* In struct ctables_section's 'cells' hmap. Indexed by all the values in
+ all the axes (except the scalar variable, if any). */
struct hmap_node node;
- /* The domains that contains this cell. */
+ /* The domains that contain this cell. */
+ bool contributes_to_domains;
struct ctables_domain *domains[N_CTDTS];
- struct
+ bool hide;
+ enum ctables_summary_variant sv;
+
+ struct ctables_cell_axis
{
- size_t vaa_idx;
struct ctables_cell_value
{
const struct ctables_category *category;
struct ctables
{
+ const struct dictionary *dict;
struct pivot_table_look *look;
/* If this is NULL, zeros are displayed using the normal print format.
/* Indexed by variable dictionary index. */
enum ctables_vlabel *vlabels;
+ struct hmap postcomputes; /* Contains "struct ctables_postcompute"s. */
+
bool mrsets_count_duplicates; /* MRSETS. */
bool smissing_listwise; /* SMISSING. */
- struct variable *base_weight; /* WEIGHT. */
+ struct variable *e_weight; /* WEIGHT. */
int hide_threshold; /* HIDESMALLCOUNTS. */
struct ctables_table **tables;
struct ctables_postcompute
{
struct hmap_node hmap_node; /* In struct ctables's 'pcompute' hmap. */
- const char *name; /* Name, without leading &. */
+ char *name; /* Name, without leading &. */
- struct ctables_postcompute_expr *expr;
+ struct msg_location *location; /* Location of definition. */
+ struct ctables_pcexpr *expr;
char *label;
- /* XXX FORMAT */
+ struct ctables_summary_spec_set *specs;
bool hide_source_cats;
};
-struct ctables_postcompute_expr
+struct ctables_pcexpr
{
+ /* Precedence table:
+
+ ()
+ **
+ -
+ * /
+ - +
+ */
enum ctables_postcompute_op
{
/* Terminals. */
- CTPO_CAT_NUMBER,
- CTPO_CAT_STRING,
- CTPO_CAT_RANGE,
- CTPO_CAT_MISSING,
- /* XXX OTHERNM */
- /* XXX SUBTOTAL and HSUBTOTAL */
+ CTPO_CONSTANT, /* 5 */
+ CTPO_CAT_NUMBER, /* [5] */
+ CTPO_CAT_STRING, /* ["STRING"] */
+ CTPO_CAT_RANGE, /* [LO THRU 5] */
+ CTPO_CAT_MISSING, /* MISSING */
+ CTPO_CAT_OTHERNM, /* OTHERNM */
+ CTPO_CAT_SUBTOTAL, /* SUBTOTAL */
+ CTPO_CAT_TOTAL, /* TOTAL */
/* Nonterminals. */
CTPO_ADD,
CTPO_MUL,
CTPO_DIV,
CTPO_POW,
+ CTPO_NEG,
}
op;
union
{
- /* CTPO_CAT_NUMBER, CTPO_NUMBER. */
+ /* CTPO_CAT_NUMBER. */
double number;
- /* CTPO_CAT_RANGE.
+ /* CTPO_CAT_STRING. */
+ char *string;
- XXX what about string ranges? */
+ /* CTPO_CAT_RANGE. */
double range[2];
- /* CTPO_ADD, CTPO_SUB, CTPO_MUL, CTPO_DIV, CTPO_POW. */
- struct ctables_postcompute_expr *subs[2];
+ /* CTPO_CAT_SUBTOTAL. */
+ size_t subtotal_index;
+
+ /* Two elements: CTPO_ADD, CTPO_SUB, CTPO_MUL, CTPO_DIV, CTPO_POW.
+ One element: CTPO_NEG. */
+ struct ctables_pcexpr *subs[2];
};
+
+ /* Source location. */
+ int ofs[2];
+ struct msg_location *location;
};
-enum ctables_label_position
+static void ctables_pcexpr_destroy (struct ctables_pcexpr *);
+static struct ctables_pcexpr *ctables_pcexpr_allocate_binary (
+ enum ctables_postcompute_op, struct ctables_pcexpr *sub0,
+ struct ctables_pcexpr *sub1);
+
+struct ctables_summary_spec_set
{
- CTLP_NORMAL,
- CTLP_OPPOSITE,
- CTLP_LAYER,
+ struct ctables_summary_spec *specs;
+ size_t n;
+ size_t allocated;
+
+ struct variable *var;
};
-struct var_array
+static void ctables_summary_spec_set_clone (struct ctables_summary_spec_set *,
+ const struct ctables_summary_spec_set *);
+static void ctables_summary_spec_set_uninit (struct ctables_summary_spec_set *);
+
+/* A nested sequence of variables, e.g. a > b > c. */
+struct ctables_nest
{
struct variable **vars;
size_t n;
size_t *domains[N_CTDTS];
size_t n_domains[N_CTDTS];
- struct ctables_summary_spec *summaries;
- size_t n_summaries;
- struct variable *summary_var;
+ struct ctables_summary_spec_set specs[N_CSVS];
};
-struct var_array2
+/* A stack of nestings, e.g. nest1 + nest2 + ... + nestN. */
+struct ctables_stack
{
- struct var_array *vas;
+ struct ctables_nest *nests;
size_t n;
};
+struct ctables_value
+ {
+ struct hmap_node node;
+ union value value;
+ int leaf;
+ };
+
+struct ctables_section_value
+ {
+ struct hmap_node node;
+ union value value;
+ };
+
+struct ctables_section
+ {
+ struct ctables_table *table;
+ struct ctables_nest *nests[PIVOT_N_AXES];
+ struct hmap *occurrences[PIVOT_N_AXES];
+ struct hmap cells; /* Contains "struct ctable_cell"s. */
+ struct hmap domains[N_CTDTS]; /* Contains "struct ctable_domain"s. */
+ };
+
struct ctables_table
{
+ struct ctables *ctables;
struct ctables_axis *axes[PIVOT_N_AXES];
- struct var_array2 vaas[PIVOT_N_AXES];
+ struct ctables_stack stacks[PIVOT_N_AXES];
+ struct ctables_section *sections;
+ size_t n_sections;
enum pivot_axis_type summary_axis;
- struct hmap cells;
- struct hmap domains[N_CTDTS];
+ struct ctables_summary_spec_set summary_specs;
+
+ const struct variable *clabels_example;
+ struct hmap clabels_values_map;
+ struct ctables_value **clabels_values;
+ size_t n_clabels_values;
- enum pivot_axis_type slabels_position;
+ enum pivot_axis_type slabels_axis;
bool slabels_visible;
- enum ctables_label_position row_labels;
- enum ctables_label_position col_labels;
+ /* The innermost category labels for axis 'a' appear on axis label_axis[a].
+
+ Most commonly, label_axis[a] == a, and in particular we always have
+ label_axis{PIVOT_AXIS_LAYER] == PIVOT_AXIS_LAYER.
+
+ If ROWLABELS or COLLABELS is specified, then one of
+ label_axis[PIVOT_AXIS_ROW] or label_axis[PIVOT_AXIS_COLUMN] can be the
+ opposite axis or PIVOT_AXIS_LAYER. Only one of them will differ.
+ */
+ enum pivot_axis_type label_axis[PIVOT_N_AXES];
+ enum pivot_axis_type clabels_from_axis;
/* Indexed by variable dictionary index. */
struct ctables_categories **categories;
{
enum ctables_category_type
{
+ /* Explicit category lists. */
CCT_NUMBER,
CCT_STRING,
CCT_RANGE,
CCT_MISSING,
CCT_OTHERNM,
+ /* Totals and subtotals. */
CCT_SUBTOTAL,
CCT_HSUBTOTAL,
CCT_TOTAL,
+ /* Implicit category lists. */
CCT_VALUE,
CCT_LABEL,
CCT_FUNCTION,
}
type;
+ struct ctables_category *subtotal;
+
union
{
double number; /* CCT_NUMBER. */
}
}
+static bool
+ctables_category_equal (const struct ctables_category *a,
+ const struct ctables_category *b)
+{
+ if (a->type != b->type)
+ return false;
+
+ switch (a->type)
+ {
+ case CCT_NUMBER:
+ return a->number == b->number;
+
+ case CCT_STRING:
+ return strcmp (a->string, b->string);
+
+ case CCT_RANGE:
+ return a->range[0] == b->range[0] && a->range[1] == b->range[1];
+
+ case CCT_MISSING:
+ case CCT_OTHERNM:
+ return true;
+
+ case CCT_SUBTOTAL:
+ case CCT_HSUBTOTAL:
+ case CCT_TOTAL:
+ return !strcmp (a->total_label, b->total_label);
+
+ case CCT_VALUE:
+ case CCT_LABEL:
+ case CCT_FUNCTION:
+ return (a->include_missing == b->include_missing
+ && a->sort_ascending == b->sort_ascending
+ && a->sort_function == b->sort_function
+ && a->sort_var == b->sort_var
+ && a->percentile == b->percentile);
+ }
+
+ NOT_REACHED ();
+}
+
static void
ctables_categories_unref (struct ctables_categories *c)
{
free (c);
}
+static bool
+ctables_categories_equal (const struct ctables_categories *a,
+ const struct ctables_categories *b)
+{
+ if (a->n_cats != b->n_cats || a->show_empty != b->show_empty)
+ return false;
+
+ for (size_t i = 0; i < a->n_cats; i++)
+ if (!ctables_category_equal (&a->cats[i], &b->cats[i]))
+ return false;
+
+ return true;
+}
+
/* Chi-square test (SIGTEST). */
struct ctables_chisq
{
{
struct ctables_var var;
bool scale;
- struct ctables_summary_spec *summaries;
- size_t n_summaries;
- size_t allocated_summaries;
+ struct ctables_summary_spec_set specs[N_CSVS];
};
/* Nonterminals. */
double percentile; /* CTSF_PTILE only. */
char *label;
struct fmt_spec format; /* XXX extra CTABLES formats */
+ size_t axis_idx;
};
+static void
+ctables_summary_spec_clone (struct ctables_summary_spec *dst,
+ const struct ctables_summary_spec *src)
+{
+ *dst = *src;
+ dst->label = xstrdup (src->label);
+}
+
static void
ctables_summary_spec_uninit (struct ctables_summary_spec *s)
{
free (s->label);
}
+static void
+ctables_summary_spec_set_clone (struct ctables_summary_spec_set *dst,
+ const struct ctables_summary_spec_set *src)
+{
+ struct ctables_summary_spec *specs = xnmalloc (src->n, sizeof *specs);
+ for (size_t i = 0; i < src->n; i++)
+ ctables_summary_spec_clone (&specs[i], &src->specs[i]);
+
+ *dst = (struct ctables_summary_spec_set) {
+ .specs = specs,
+ .n = src->n,
+ .allocated = src->n,
+ .var = src->var
+ };
+}
+
+static void
+ctables_summary_spec_set_uninit (struct ctables_summary_spec_set *set)
+{
+ for (size_t i = 0; i < set->n; i++)
+ ctables_summary_spec_uninit (&set->specs[i]);
+ free (set->specs);
+}
+
static bool
parse_col_width (struct lexer *lexer, const char *name, double *width)
{
switch (axis->op)
{
case CTAO_VAR:
- for (size_t i = 0; i < axis->n_summaries; i++)
- ctables_summary_spec_uninit (&axis->summaries[i]);
- free (axis->summaries);
+ for (size_t i = 0; i < N_CSVS; i++)
+ ctables_summary_spec_set_uninit (&axis->specs[i]);
break;
case CTAO_STACK:
add_summary_spec (struct ctables_axis *axis,
enum ctables_summary_function function, double percentile,
const char *label, const struct fmt_spec *format,
- const struct msg_location *loc)
+ const struct msg_location *loc, enum ctables_summary_variant sv)
{
if (axis->op == CTAO_VAR)
{
- if (axis->n_summaries >= axis->allocated_summaries)
- axis->summaries = x2nrealloc (axis->summaries,
- &axis->allocated_summaries,
- sizeof *axis->summaries);
-
const char *function_name = ctables_summary_function_name (function);
const char *var_name = ctables_var_name (&axis->var);
switch (ctables_function_availability (function))
break;
}
- struct ctables_summary_spec *dst = &axis->summaries[axis->n_summaries++];
+ struct ctables_summary_spec_set *set = &axis->specs[sv];
+ if (set->n >= set->allocated)
+ set->specs = x2nrealloc (set->specs, &set->allocated,
+ sizeof *set->specs);
+
+ struct ctables_summary_spec *dst = &set->specs[set->n++];
*dst = (struct ctables_summary_spec) {
.function = function,
.percentile = percentile,
{
for (size_t i = 0; i < 2; i++)
if (!add_summary_spec (axis->subs[i], function, percentile, label,
- format, loc))
+ format, loc, sv))
return false;
return true;
}
return axis;
}
+static bool
+has_digit (const char *s)
+{
+ return s[strcspn (s, "0123456789")] != '\0';
+}
+
static struct ctables_axis *
ctables_axis_parse_postfix (struct ctables_axis_parse_ctx *ctx)
{
if (!sub || !lex_match (ctx->lexer, T_LBRACK))
return sub;
- do
+ enum ctables_summary_variant sv = CSV_CELL;
+ for (;;)
{
int start_ofs = lex_ofs (ctx->lexer);
/* Parse format. */
struct fmt_spec format;
const struct fmt_spec *formatp;
- if (lex_token (ctx->lexer) == T_ID)
+ if (lex_token (ctx->lexer) == T_ID
+ && has_digit (lex_tokcstr (ctx->lexer)))
{
if (!parse_format_specifier (ctx->lexer, &format)
|| !fmt_check_output (&format)
struct msg_location *loc = lex_ofs_location (ctx->lexer, start_ofs,
lex_ofs (ctx->lexer) - 1);
- add_summary_spec (sub, function, percentile, label, formatp, loc);
+ add_summary_spec (sub, function, percentile, label, formatp, loc, sv);
free (label);
msg_location_destroy (loc);
lex_match (ctx->lexer, T_COMMA);
+ if (sv == CSV_CELL && lex_match_id (ctx->lexer, "TOTALS"))
+ {
+ if (!lex_force_match (ctx->lexer, T_LBRACK))
+ goto error;
+ sv = CSV_TOTAL;
+ }
+ else if (lex_match (ctx->lexer, T_RBRACK))
+ {
+ if (sv == CSV_TOTAL && !lex_force_match (ctx->lexer, T_RBRACK))
+ goto error;
+ return sub;
+ }
}
- while (!lex_match (ctx->lexer, T_RBRACK));
-
- return sub;
error:
ctables_axis_destroy (sub);
if (!axis)
return NULL;
else if (axis->op == CTAO_VAR)
- return !axis->scale && axis->n_summaries ? axis : NULL;
+ return !axis->scale && axis->specs[CSV_CELL].n ? axis : NULL;
else
{
for (size_t i = 0; i < 2; i++)
return false;
struct ctables_categories *c = xmalloc (sizeof *c);
- *c = (struct ctables_categories) { .n_refs = n_vars };
+ *c = (struct ctables_categories) { .n_refs = n_vars, .show_empty = true };
for (size_t i = 0; i < n_vars; i++)
{
struct ctables_categories **cp
lex_get (lexer);
if (lex_match_id (lexer, "THRU"))
{
- cat->type = CCT_RANGE;
- cat->range[0] = number;
if (lex_match_id (lexer, "HI"))
*cat = cct_range (number, DBL_MAX);
else
return false;
}
- if ((cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
- && lex_match (lexer, T_EQUALS))
+ if (cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
{
- if (!lex_force_string (lexer))
- return false;
+ if (lex_match (lexer, T_EQUALS))
+ {
+ if (!lex_force_string (lexer))
+ return false;
- cat->total_label = ss_xstrdup (lex_tokss (lexer));
- lex_get (lexer);
+ cat->total_label = ss_xstrdup (lex_tokss (lexer));
+ lex_get (lexer);
+ }
+ else
+ cat->total_label = xstrdup (_("Subtotal"));
}
c->n_cats++;
};
}
+ struct ctables_category *subtotal = NULL;
+ for (size_t i = totals_before ? 0 : c->n_cats;
+ totals_before ? i < c->n_cats : i-- > 0;
+ totals_before ? i++ : 0)
+ {
+ struct ctables_category *cat = &c->cats[i];
+ switch (cat->type)
+ {
+ case CCT_NUMBER:
+ case CCT_STRING:
+ case CCT_RANGE:
+ case CCT_MISSING:
+ case CCT_OTHERNM:
+ cat->subtotal = subtotal;
+ break;
+
+ case CCT_SUBTOTAL:
+ case CCT_HSUBTOTAL:
+ subtotal = cat;
+ break;
+
+ case CCT_TOTAL:
+ case CCT_VALUE:
+ case CCT_LABEL:
+ case CCT_FUNCTION:
+ break;
+ }
+ }
+
return true;
}
static void
-var_array_uninit (struct var_array *va)
+ctables_nest_uninit (struct ctables_nest *nest)
{
- if (va)
- free (va->vars);
+ if (nest)
+ free (nest->vars);
}
static void
-var_array2_uninit (struct var_array2 *vaa)
+ctables_stack_uninit (struct ctables_stack *stack)
{
- if (vaa)
+ if (stack)
{
- for (size_t i = 0; i < vaa->n; i++)
- var_array_uninit (&vaa->vas[i]);
- free (vaa->vas);
+ for (size_t i = 0; i < stack->n; i++)
+ ctables_nest_uninit (&stack->nests[i]);
+ free (stack->nests);
}
}
-static struct var_array2
-nest_fts (struct var_array2 va0, struct var_array2 va1)
+static struct ctables_stack
+nest_fts (struct ctables_stack s0, struct ctables_stack s1)
{
- if (!va0.n)
- return va1;
- else if (!va1.n)
- return va0;
-
- struct var_array2 vaa = { .vas = xnmalloc (va0.n, va1.n * sizeof *vaa.vas) };
- for (size_t i = 0; i < va0.n; i++)
- for (size_t j = 0; j < va1.n; j++)
+ if (!s0.n)
+ return s1;
+ else if (!s1.n)
+ return s0;
+
+ struct ctables_stack stack = { .nests = xnmalloc (s0.n, s1.n * sizeof *stack.nests) };
+ for (size_t i = 0; i < s0.n; i++)
+ for (size_t j = 0; j < s1.n; j++)
{
- const struct var_array *a = &va0.vas[i];
- const struct var_array *b = &va1.vas[j];
+ const struct ctables_nest *a = &s0.nests[i];
+ const struct ctables_nest *b = &s1.nests[j];
size_t allocate = a->n + b->n;
struct variable **vars = xnmalloc (allocate, sizeof *vars);
vars[n++] = b->vars[k];
assert (n == allocate);
- const struct var_array *summary_src;
- if (!a->summary_var)
+ const struct ctables_nest *summary_src;
+ if (!a->specs[CSV_CELL].var)
summary_src = b;
- else if (!b->summary_var)
+ else if (!b->specs[CSV_CELL].var)
summary_src = a;
else
NOT_REACHED ();
- vaa.vas[vaa.n++] = (struct var_array) {
+
+ struct ctables_nest *new = &stack.nests[stack.n++];
+ *new = (struct ctables_nest) {
.vars = vars,
.scale_idx = (a->scale_idx != SIZE_MAX ? a->scale_idx
: b->scale_idx != SIZE_MAX ? a->n + b->scale_idx
: SIZE_MAX),
.n = n,
- .summaries = summary_src->summaries,
- .n_summaries = summary_src->n_summaries,
- .summary_var = summary_src->summary_var,
};
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ ctables_summary_spec_set_clone (&new->specs[sv], &summary_src->specs[sv]);
}
- var_array2_uninit (&va0);
- var_array2_uninit (&va1);
- return vaa;
+ ctables_stack_uninit (&s0);
+ ctables_stack_uninit (&s1);
+ return stack;
}
-static struct var_array2
-stack_fts (struct var_array2 va0, struct var_array2 va1)
+static struct ctables_stack
+stack_fts (struct ctables_stack s0, struct ctables_stack s1)
{
- struct var_array2 vaa = { .vas = xnmalloc (va0.n + va1.n, sizeof *vaa.vas) };
- for (size_t i = 0; i < va0.n; i++)
- vaa.vas[vaa.n++] = va0.vas[i];
- for (size_t i = 0; i < va1.n; i++)
- vaa.vas[vaa.n++] = va1.vas[i];
- assert (vaa.n == va0.n + va1.n);
- free (va0.vas);
- free (va1.vas);
- return vaa;
+ struct ctables_stack stack = { .nests = xnmalloc (s0.n + s1.n, sizeof *stack.nests) };
+ for (size_t i = 0; i < s0.n; i++)
+ stack.nests[stack.n++] = s0.nests[i];
+ for (size_t i = 0; i < s1.n; i++)
+ stack.nests[stack.n++] = s1.nests[i];
+ assert (stack.n == s0.n + s1.n);
+ free (s0.nests);
+ free (s1.nests);
+ return stack;
}
-static struct var_array2
+static struct ctables_stack
enumerate_fts (enum pivot_axis_type axis_type, const struct ctables_axis *a)
{
if (!a)
- return (struct var_array2) { .n = 0 };
+ return (struct ctables_stack) { .n = 0 };
switch (a->op)
{
struct variable **vars = xmalloc (sizeof *vars);
*vars = a->var.var;
- struct var_array *va = xmalloc (sizeof *va);
- *va = (struct var_array) {
+ struct ctables_nest *nest = xmalloc (sizeof *nest);
+ *nest = (struct ctables_nest) {
.vars = vars,
.n = 1,
.scale_idx = a->scale ? 0 : SIZE_MAX,
};
- if (a->n_summaries || a->scale)
- {
- va->summaries = a->summaries;
- va->n_summaries = a->n_summaries;
- va->summary_var = a->var.var;
- }
- return (struct var_array2) { .vas = va, .n = 1 };
+ if (a->specs[CSV_CELL].n || a->scale)
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ {
+ ctables_summary_spec_set_clone (&nest->specs[sv], &a->specs[sv]);
+ nest->specs[sv].var = a->var.var;
+ }
+ return (struct ctables_stack) { .nests = nest, .n = 1 };
case CTAO_STACK:
return stack_fts (enumerate_fts (axis_type, a->subs[0]),
/* MEAN, SEMEAN, STDDEV, SUM, VARIANCE, *.SUM. */
struct moments1 *moments;
- /* XXX percentiles, median, mode, multiple response */
+ /* MEDIAN, MODE, PTILE. */
+ struct
+ {
+ struct casewriter *writer;
+ double ovalid;
+ double ovalue;
+ };
+
+ /* XXX multiple response */
};
static void
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_MISSING:
case CSTF_TOTALN:
case CTSF_ETOTALN:
case CTSF_VALIDN:
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ {
+ struct caseproto *proto = caseproto_create ();
+ proto = caseproto_add_width (proto, 0);
+ proto = caseproto_add_width (proto, 0);
+
+ struct subcase ordering;
+ subcase_init (&ordering, 0, 0, SC_ASCEND);
+ s->writer = sort_create_writer (&ordering, proto);
+ subcase_uninit (&ordering);
+ caseproto_unref (proto);
+
+ s->ovalid = 0;
+ s->ovalue = SYSMIS;
+ }
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_MISSING:
case CSTF_TOTALN:
case CTSF_ETOTALN:
case CTSF_VALIDN:
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ casewriter_destroy (s->writer);
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
ctables_summary_add (union ctables_summary *s,
const struct ctables_summary_spec *ss,
const struct variable *var, const union value *value,
- double weight)
+ double d_weight, double e_weight)
{
switch (ss->function)
{
case CTSF_COUNT:
+ case CSTF_TOTALN:
+ case CTSF_VALIDN:
+ if (var_is_value_missing (var, value))
+ s->missing += d_weight;
+ else
+ s->valid += d_weight;
+ break;
+
case CTSF_ECOUNT:
case CTSF_ROWPCT_COUNT:
case CTSF_COLPCT_COUNT:
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
- case CSTF_TOTALN:
+ case CTSF_MISSING:
case CTSF_ETOTALN:
- case CTSF_VALIDN:
case CTSF_EVALIDN:
if (var_is_value_missing (var, value))
- s->missing += weight;
+ s->missing += e_weight;
else
- s->valid += weight;
+ s->valid += e_weight;
break;
case CTSF_MAXIMUM:
case CTSF_LAYERPCT_SUM:
case CTSF_LAYERROWPCT_SUM:
case CTSF_LAYERCOLPCT_SUM:
- moments1_add (s->moments, value->f, weight);
+ if (!var_is_value_missing (var, value))
+ moments1_add (s->moments, value->f, e_weight);
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ if (var_is_value_missing (var, value))
+ {
+ s->ovalid += e_weight;
+
+ struct ccase *c = case_create (casewriter_get_proto (s->writer));
+ *case_num_rw_idx (c, 0) = value->f;
+ *case_num_rw_idx (c, 1) = e_weight;
+ casewriter_write (s->writer, c);
+ }
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
}
}
+static enum ctables_domain_type
+ctables_function_domain (enum ctables_summary_function function)
+{
+ switch (function)
+ {
+ case CTSF_COUNT:
+ case CTSF_ECOUNT:
+ case CTSF_MISSING:
+ case CSTF_TOTALN:
+ case CTSF_ETOTALN:
+ case CTSF_VALIDN:
+ case CTSF_EVALIDN:
+ case CTSF_MAXIMUM:
+ case CTSF_MINIMUM:
+ case CTSF_RANGE:
+ case CTSF_MEAN:
+ case CTSF_SEMEAN:
+ case CTSF_STDDEV:
+ case CTSF_SUM:
+ case CTSF_VARIANCE:
+ case CTSF_MEDIAN:
+ case CTSF_PTILE:
+ case CTSF_MODE:
+ case CTSF_RESPONSES:
+ NOT_REACHED ();
+
+ case CTSF_COLPCT_COUNT:
+ case CTSF_COLPCT_COUNT_RESPONSES:
+ case CTSF_COLPCT_RESPONSES:
+ case CTSF_COLPCT_RESPONSES_COUNT:
+ case CTSF_COLPCT_SUM:
+ case CTSF_COLPCT_TOTALN:
+ case CTSF_COLPCT_VALIDN:
+ return CTDT_COL;
+
+ case CTSF_LAYERCOLPCT_COUNT:
+ case CTSF_LAYERCOLPCT_COUNT_RESPONSES:
+ case CTSF_LAYERCOLPCT_RESPONSES:
+ case CTSF_LAYERCOLPCT_RESPONSES_COUNT:
+ case CTSF_LAYERCOLPCT_SUM:
+ case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_LAYERCOLPCT_VALIDN:
+ return CTDT_LAYERCOL;
+
+ case CTSF_LAYERPCT_COUNT:
+ case CTSF_LAYERPCT_COUNT_RESPONSES:
+ case CTSF_LAYERPCT_RESPONSES:
+ case CTSF_LAYERPCT_RESPONSES_COUNT:
+ case CTSF_LAYERPCT_SUM:
+ case CTSF_LAYERPCT_TOTALN:
+ case CTSF_LAYERPCT_VALIDN:
+ return CTDT_LAYER;
+
+ case CTSF_LAYERROWPCT_COUNT:
+ case CTSF_LAYERROWPCT_COUNT_RESPONSES:
+ case CTSF_LAYERROWPCT_RESPONSES:
+ case CTSF_LAYERROWPCT_RESPONSES_COUNT:
+ case CTSF_LAYERROWPCT_SUM:
+ case CTSF_LAYERROWPCT_TOTALN:
+ case CTSF_LAYERROWPCT_VALIDN:
+ return CTDT_LAYERROW;
+
+ case CTSF_ROWPCT_COUNT:
+ case CTSF_ROWPCT_COUNT_RESPONSES:
+ case CTSF_ROWPCT_RESPONSES:
+ case CTSF_ROWPCT_RESPONSES_COUNT:
+ case CTSF_ROWPCT_SUM:
+ case CTSF_ROWPCT_TOTALN:
+ case CTSF_ROWPCT_VALIDN:
+ return CTDT_ROW;
+
+ case CTSF_SUBTABLEPCT_COUNT:
+ case CTSF_SUBTABLEPCT_COUNT_RESPONSES:
+ case CTSF_SUBTABLEPCT_RESPONSES:
+ case CTSF_SUBTABLEPCT_RESPONSES_COUNT:
+ case CTSF_SUBTABLEPCT_SUM:
+ case CTSF_SUBTABLEPCT_TOTALN:
+ case CTSF_SUBTABLEPCT_VALIDN:
+ return CTDT_SUBTABLE;
+
+ case CTSF_TABLEPCT_COUNT:
+ case CTSF_TABLEPCT_COUNT_RESPONSES:
+ case CTSF_TABLEPCT_RESPONSES:
+ case CTSF_TABLEPCT_RESPONSES_COUNT:
+ case CTSF_TABLEPCT_SUM:
+ case CTSF_TABLEPCT_TOTALN:
+ case CTSF_TABLEPCT_VALIDN:
+ return CTDT_TABLE;
+ }
+
+ NOT_REACHED ();
+}
+
static double
-ctables_summary_value (const struct ctables_cell *f,
+ctables_summary_value (const struct ctables_cell *cell,
union ctables_summary *s,
const struct ctables_summary_spec *ss)
{
case CTSF_ECOUNT:
return s->valid;
- case CTSF_SUBTABLEPCT_COUNT:
- return f->domains[CTDT_SUBTABLE]->valid ? s->valid / f->domains[CTDT_SUBTABLE]->valid * 100 : SYSMIS;
-
case CTSF_ROWPCT_COUNT:
- return f->domains[CTDT_ROW]->valid ? s->valid / f->domains[CTDT_ROW]->valid * 100 : SYSMIS;
-
case CTSF_COLPCT_COUNT:
- return f->domains[CTDT_COL]->valid ? s->valid / f->domains[CTDT_COL]->valid * 100 : SYSMIS;
-
case CTSF_TABLEPCT_COUNT:
- return f->domains[CTDT_TABLE]->valid ? s->valid / f->domains[CTDT_TABLE]->valid * 100 : SYSMIS;
-
+ case CTSF_SUBTABLEPCT_COUNT:
case CTSF_LAYERPCT_COUNT:
- return f->domains[CTDT_LAYER]->valid ? s->valid / f->domains[CTDT_LAYER]->valid * 100 : SYSMIS;
-
case CTSF_LAYERROWPCT_COUNT:
- return f->domains[CTDT_LAYERROW]->valid ? s->valid / f->domains[CTDT_LAYERROW]->valid * 100 : SYSMIS;
-
case CTSF_LAYERCOLPCT_COUNT:
- return f->domains[CTDT_LAYERCOL]->valid ? s->valid / f->domains[CTDT_LAYERCOL]->valid * 100 : SYSMIS;
+ {
+ enum ctables_domain_type d = ctables_function_domain (ss->function);
+ return (cell->domains[d]->e_valid
+ ? s->valid / cell->domains[d]->e_valid * 100
+ : SYSMIS);
+ }
case CTSF_ROWPCT_VALIDN:
case CTSF_COLPCT_VALIDN:
case CTSF_LAYERCOLPCT_TOTALN:
NOT_REACHED ();
+ case CTSF_MISSING:
+ return s->missing;
+
case CSTF_TOTALN:
case CTSF_ETOTALN:
return s->valid + s->missing;
NOT_REACHED ();
case CTSF_MEDIAN:
- case CTSF_MISSING:
- case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
-
+ if (s->writer)
+ {
+ struct casereader *reader = casewriter_make_reader (s->writer);
+ s->writer = NULL;
+
+ struct percentile *ptile = percentile_create (
+ ss->function == CTSF_PTILE ? ss->percentile : 0.5, s->ovalid);
+ struct order_stats *os = &ptile->parent;
+ order_stats_accumulate_idx (&os, 1, reader, 1, 0);
+ s->ovalue = percentile_calculate (ptile, PC_HAVERAGE);
+ statistic_destroy (&ptile->parent.parent);
+ }
+ return s->ovalue;
+
+ case CTSF_MODE:
+ if (s->writer)
+ {
+ struct casereader *reader = casewriter_make_reader (s->writer);
+ s->writer = NULL;
+
+ struct mode *mode = mode_create ();
+ struct order_stats *os = &mode->parent;
+ order_stats_accumulate_idx (&os, 1, reader, 1, 0);
+ s->ovalue = mode->mode;
+ statistic_destroy (&mode->parent.parent);
+ }
+ return s->ovalue;
+
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
case CTSF_COLPCT_RESPONSES:
struct ctables_cell_sort_aux
{
- const struct ctables_table *t;
+ const struct ctables_nest *nest;
enum pivot_axis_type a;
};
const struct ctables_cell *a = *ap;
const struct ctables_cell *b = *bp;
- size_t a_idx = a->axes[aux->a].vaa_idx;
- size_t b_idx = b->axes[aux->a].vaa_idx;
- if (a_idx != b_idx)
- return a_idx < b_idx ? -1 : 1;
-
- const struct var_array *va = &aux->t->vaas[aux->a].vas[a_idx];
- for (size_t i = 0; i < va->n; i++)
- if (i != va->scale_idx)
+ const struct ctables_nest *nest = aux->nest;
+ for (size_t i = 0; i < nest->n; i++)
+ if (i != nest->scale_idx)
{
- const struct variable *var = va->vars[i];
+ const struct variable *var = nest->vars[i];
const struct ctables_cell_value *a_cv = &a->axes[aux->a].cvs[i];
const struct ctables_cell_value *b_cv = &b->axes[aux->a].cvs[i];
if (a_cv->category != b_cv->category)
*/
static struct ctables_domain *
-ctables_domain_insert (struct ctables_table *t, struct ctables_cell *f,
+ctables_domain_insert (struct ctables_section *s, struct ctables_cell *cell,
enum ctables_domain_type domain)
{
size_t hash = 0;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
- size_t idx = f->axes[a].vaa_idx;
- const struct var_array *va = &t->vaas[a].vas[idx];
- hash = hash_int (idx, hash);
- for (size_t i = 0; i < va->n_domains[domain]; i++)
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n_domains[domain]; i++)
{
- size_t v_idx = va->domains[domain][i];
- hash = value_hash (&f->axes[a].cvs[v_idx].value,
- var_get_width (va->vars[v_idx]), hash);
+ size_t v_idx = nest->domains[domain][i];
+ hash = value_hash (&cell->axes[a].cvs[v_idx].value,
+ var_get_width (nest->vars[v_idx]), hash);
}
}
struct ctables_domain *d;
- HMAP_FOR_EACH_WITH_HASH (d, struct ctables_domain, node, hash, &t->domains[domain])
+ HMAP_FOR_EACH_WITH_HASH (d, struct ctables_domain, node, hash, &s->domains[domain])
{
const struct ctables_cell *df = d->example;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
- size_t idx = f->axes[a].vaa_idx;
- if (idx != df->axes[a].vaa_idx)
- goto not_equal;
-
- const struct var_array *va = &t->vaas[a].vas[idx];
- for (size_t i = 0; i < va->n_domains[domain]; i++)
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n_domains[domain]; i++)
{
- size_t v_idx = va->domains[domain][i];
+ size_t v_idx = nest->domains[domain][i];
if (!value_equal (&df->axes[a].cvs[v_idx].value,
- &f->axes[a].cvs[v_idx].value,
- var_get_width (va->vars[v_idx])))
+ &cell->axes[a].cvs[v_idx].value,
+ var_get_width (nest->vars[v_idx])))
goto not_equal;
}
}
}
d = xmalloc (sizeof *d);
- *d = (struct ctables_domain) { .example = f };
- hmap_insert (&t->domains[domain], &d->node, hash);
+ *d = (struct ctables_domain) { .example = cell };
+ hmap_insert (&s->domains[domain], &d->node, hash);
return d;
}
}
}
- return var_is_value_missing (var, v) ? NULL : othernm;
+ return var_is_value_missing (var, v) ? NULL : othernm;
+}
+
+static const struct ctables_category *
+ctables_categories_total (const struct ctables_categories *c)
+{
+ const struct ctables_category *first = &c->cats[0];
+ const struct ctables_category *last = &c->cats[c->n_cats - 1];
+ return (first->type == CCT_TOTAL ? first
+ : last->type == CCT_TOTAL ? last
+ : NULL);
+}
+
+static struct ctables_cell *
+ctables_cell_insert__ (struct ctables_section *s, const struct ccase *c,
+ const struct ctables_category *cats[PIVOT_N_AXES][10])
+{
+ const struct ctables_nest *ss = s->nests[s->table->summary_axis];
+
+ size_t hash = 0;
+ enum ctables_summary_variant sv = CSV_CELL;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n; i++)
+ if (i != nest->scale_idx)
+ {
+ hash = hash_pointer (cats[a][i], hash);
+ if (cats[a][i]->type != CCT_TOTAL
+ && cats[a][i]->type != CCT_SUBTOTAL
+ && cats[a][i]->type != CCT_HSUBTOTAL)
+ hash = value_hash (case_data (c, nest->vars[i]),
+ var_get_width (nest->vars[i]), hash);
+ else
+ sv = CSV_TOTAL;
+ }
+ }
+
+ struct ctables_cell *cell;
+ HMAP_FOR_EACH_WITH_HASH (cell, struct ctables_cell, node, hash, &s->cells)
+ {
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n; i++)
+ if (i != nest->scale_idx
+ && (cats[a][i] != cell->axes[a].cvs[i].category
+ || (cats[a][i]->type != CCT_TOTAL
+ && cats[a][i]->type != CCT_SUBTOTAL
+ && cats[a][i]->type != CCT_HSUBTOTAL
+ && !value_equal (case_data (c, nest->vars[i]),
+ &cell->axes[a].cvs[i].value,
+ var_get_width (nest->vars[i])))))
+ goto not_equal;
+ }
+
+ return cell;
+
+ not_equal: ;
+ }
+
+ cell = xmalloc (sizeof *cell);
+ cell->hide = false;
+ cell->sv = sv;
+ cell->contributes_to_domains = true;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ cell->axes[a].cvs = (nest->n
+ ? xnmalloc (nest->n, sizeof *cell->axes[a].cvs)
+ : NULL);
+ for (size_t i = 0; i < nest->n; i++)
+ {
+ const struct ctables_category *cat = cats[a][i];
+
+ if (i != nest->scale_idx)
+ {
+ const struct ctables_category *subtotal = cat->subtotal;
+ if (subtotal && subtotal->type == CCT_HSUBTOTAL)
+ cell->hide = true;
+
+ if (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
+ cell->contributes_to_domains = false;
+ }
+
+ cell->axes[a].cvs[i].category = cat;
+ value_clone (&cell->axes[a].cvs[i].value, case_data (c, nest->vars[i]),
+ var_get_width (nest->vars[i]));
+ }
+ }
+
+ const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
+ cell->summaries = xmalloc (specs->n * sizeof *cell->summaries);
+ for (size_t i = 0; i < specs->n; i++)
+ ctables_summary_init (&cell->summaries[i], &specs->specs[i]);
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ cell->domains[dt] = ctables_domain_insert (s, cell, dt);
+ hmap_insert (&s->cells, &cell->node, hash);
+ return cell;
+}
+
+static void
+ctables_cell_add__ (struct ctables_section *s, const struct ccase *c,
+ const struct ctables_category *cats[PIVOT_N_AXES][10],
+ double d_weight, double e_weight)
+{
+ struct ctables_cell *cell = ctables_cell_insert__ (s, c, cats);
+ const struct ctables_nest *ss = s->nests[s->table->summary_axis];
+
+ const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
+ for (size_t i = 0; i < specs->n; i++)
+ ctables_summary_add (&cell->summaries[i], &specs->specs[i], specs->var,
+ case_data (c, specs->var), d_weight, e_weight);
+ if (cell->contributes_to_domains)
+ {
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ {
+ cell->domains[dt]->d_valid += d_weight;
+ cell->domains[dt]->e_valid += e_weight;
+ }
+ }
+}
+
+static void
+recurse_totals (struct ctables_section *s, const struct ccase *c,
+ const struct ctables_category *cats[PIVOT_N_AXES][10],
+ double d_weight, double e_weight,
+ enum pivot_axis_type start_axis, size_t start_nest)
+{
+ for (enum pivot_axis_type a = start_axis; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = start_nest; i < nest->n; i++)
+ {
+ if (i == nest->scale_idx)
+ continue;
+
+ const struct variable *var = nest->vars[i];
+
+ const struct ctables_category *total = ctables_categories_total (
+ s->table->categories[var_get_dict_index (var)]);
+ if (total)
+ {
+ const struct ctables_category *save = cats[a][i];
+ cats[a][i] = total;
+ ctables_cell_add__ (s, c, cats, d_weight, e_weight);
+ recurse_totals (s, c, cats, d_weight, e_weight, a, i + 1);
+ cats[a][i] = save;
+ }
+ }
+ start_nest = 0;
+ }
+}
+
+static void
+recurse_subtotals (struct ctables_section *s, const struct ccase *c,
+ const struct ctables_category *cats[PIVOT_N_AXES][10],
+ double d_weight, double e_weight,
+ enum pivot_axis_type start_axis, size_t start_nest)
+{
+ for (enum pivot_axis_type a = start_axis; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = start_nest; i < nest->n; i++)
+ {
+ if (i == nest->scale_idx)
+ continue;
+
+ const struct ctables_category *save = cats[a][i];
+ if (save->subtotal)
+ {
+ cats[a][i] = save->subtotal;
+ ctables_cell_add__ (s, c, cats, d_weight, e_weight);
+ recurse_subtotals (s, c, cats, d_weight, e_weight, a, i + 1);
+ cats[a][i] = save;
+ }
+ }
+ start_nest = 0;
+ }
+}
+
+static void
+ctables_add_occurrence (const struct variable *var,
+ const union value *value,
+ struct hmap *occurrences)
+{
+ int width = var_get_width (var);
+ unsigned int hash = value_hash (value, width, 0);
+
+ struct ctables_section_value *sv;
+ HMAP_FOR_EACH_WITH_HASH (sv, struct ctables_section_value, node, hash,
+ occurrences)
+ if (value_equal (value, &sv->value, width))
+ return;
+
+ sv = xmalloc (sizeof *sv);
+ value_clone (&sv->value, value, width);
+ hmap_insert (occurrences, &sv->node, hash);
+}
+
+static void
+ctables_cell_insert (struct ctables_section *s,
+ const struct ccase *c,
+ double d_weight, double e_weight)
+{
+ const struct ctables_category *cats[PIVOT_N_AXES][10]; /* XXX */
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n; i++)
+ {
+ if (i == nest->scale_idx)
+ continue;
+
+ const struct variable *var = nest->vars[i];
+ const union value *value = case_data (c, var);
+
+ if (var_is_numeric (var) && value->f == SYSMIS)
+ return;
+
+ cats[a][i] = ctables_categories_match (
+ s->table->categories[var_get_dict_index (var)], value, var);
+ if (!cats[a][i])
+ return;
+ }
+ }
+
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = s->nests[a];
+ for (size_t i = 0; i < nest->n; i++)
+ if (i != nest->scale_idx)
+ {
+ const struct variable *var = nest->vars[i];
+ const union value *value = case_data (c, var);
+ ctables_add_occurrence (var, value, &s->occurrences[a][i]);
+ }
+ }
+
+ ctables_cell_add__ (s, c, cats, d_weight, e_weight);
+
+ recurse_totals (s, c, cats, d_weight, e_weight, 0, 0);
+ recurse_subtotals (s, c, cats, d_weight, e_weight, 0, 0);
+}
+
+struct merge_item
+ {
+ const struct ctables_summary_spec_set *set;
+ size_t ofs;
+ };
+
+static int
+merge_item_compare_3way (const struct merge_item *a, const struct merge_item *b)
+{
+ const struct ctables_summary_spec *as = &a->set->specs[a->ofs];
+ const struct ctables_summary_spec *bs = &b->set->specs[b->ofs];
+ if (as->function != bs->function)
+ return as->function > bs->function ? 1 : -1;
+ else if (as->percentile != bs->percentile)
+ return as->percentile < bs->percentile ? 1 : -1;
+ return strcmp (as->label, bs->label);
+}
+
+static struct pivot_value *
+ctables_category_create_label (const struct ctables_category *cat,
+ const struct variable *var,
+ const union value *value)
+{
+ return (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL
+ ? pivot_value_new_user_text (cat->total_label, SIZE_MAX)
+ : pivot_value_new_var_value (var, value));
+}
+
+static struct ctables_value *
+ctables_value_find__ (struct ctables_table *t, const union value *value,
+ int width, unsigned int hash)
+{
+ struct ctables_value *clv;
+ HMAP_FOR_EACH_WITH_HASH (clv, struct ctables_value, node,
+ hash, &t->clabels_values_map)
+ if (value_equal (value, &clv->value, width))
+ return clv;
+ return NULL;
+}
+
+static struct ctables_value *
+ctables_value_find (struct ctables_table *t,
+ const union value *value, int width)
+{
+ return ctables_value_find__ (t, value, width,
+ value_hash (value, width, 0));
+}
+
+static void
+ctables_table_add_section (struct ctables_table *t, enum pivot_axis_type a,
+ size_t ix[PIVOT_N_AXES])
+{
+ if (a < PIVOT_N_AXES)
+ {
+ size_t limit = MAX (t->stacks[a].n, 1);
+ for (ix[a] = 0; ix[a] < limit; ix[a]++)
+ ctables_table_add_section (t, a + 1, ix);
+ }
+ else
+ {
+ struct ctables_section *s = &t->sections[t->n_sections++];
+ *s = (struct ctables_section) {
+ .table = t,
+ .cells = HMAP_INITIALIZER (s->cells),
+ };
+ for (a = 0; a < PIVOT_N_AXES; a++)
+ if (t->stacks[a].n)
+ {
+ struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
+ s->nests[a] = nest;
+ s->occurrences[a] = xnmalloc (nest->n, sizeof *s->occurrences[a]);
+ for (size_t i = 0; i < nest->n; i++)
+ hmap_init (&s->occurrences[a][i]);
+ }
+ for (size_t i = 0; i < N_CTDTS; i++)
+ hmap_init (&s->domains[i]);
+ }
+}
+
+static void
+ctables_table_output (struct ctables *ct, struct ctables_table *t)
+{
+ struct pivot_table *pt = pivot_table_create__ (
+ (t->title
+ ? pivot_value_new_user_text (t->title, SIZE_MAX)
+ : pivot_value_new_text (N_("Custom Tables"))),
+ "Custom Tables");
+ if (t->caption)
+ pivot_table_set_caption (
+ pt, pivot_value_new_user_text (t->caption, SIZE_MAX));
+ if (t->corner)
+ pivot_table_set_caption (
+ pt, pivot_value_new_user_text (t->corner, SIZE_MAX));
+
+ bool summary_dimension = (t->summary_axis != t->slabels_axis
+ || (!t->slabels_visible
+ && t->summary_specs.n > 1));
+ if (summary_dimension)
+ {
+ struct pivot_dimension *d = pivot_dimension_create (
+ pt, t->slabels_axis, N_("Statistics"));
+ const struct ctables_summary_spec_set *specs = &t->summary_specs;
+ if (!t->slabels_visible)
+ d->hide_all_labels = true;
+ for (size_t i = 0; i < specs->n; i++)
+ pivot_category_create_leaf (
+ d->root, pivot_value_new_text (specs->specs[i].label));
+ }
+
+ bool categories_dimension = t->clabels_example != NULL;
+ if (categories_dimension)
+ {
+ struct pivot_dimension *d = pivot_dimension_create (
+ pt, t->label_axis[t->clabels_from_axis],
+ t->clabels_from_axis == PIVOT_AXIS_ROW
+ ? N_("Row Categories")
+ : N_("Column Categories"));
+ const struct variable *var = t->clabels_example;
+ const struct ctables_categories *c = t->categories[var_get_dict_index (var)];
+ for (size_t i = 0; i < t->n_clabels_values; i++)
+ {
+ const struct ctables_value *value = t->clabels_values[i];
+ const struct ctables_category *cat = ctables_categories_match (c, &value->value, var);
+ assert (cat != NULL);
+ pivot_category_create_leaf (d->root, ctables_category_create_label (
+ cat, t->clabels_example, &value->value));
+ }
+ }
+
+ pivot_table_set_look (pt, ct->look);
+ struct pivot_dimension *d[PIVOT_N_AXES];
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ static const char *names[] = {
+ [PIVOT_AXIS_ROW] = N_("Rows"),
+ [PIVOT_AXIS_COLUMN] = N_("Columns"),
+ [PIVOT_AXIS_LAYER] = N_("Layers"),
+ };
+ d[a] = (t->axes[a] || a == t->summary_axis
+ ? pivot_dimension_create (pt, a, names[a])
+ : NULL);
+ if (!d[a])
+ continue;
+
+ assert (t->axes[a]);
+
+ for (size_t i = 0; i < t->stacks[a].n; i++)
+ {
+ struct ctables_nest *nest = &t->stacks[a].nests[i];
+ struct ctables_section **sections = xnmalloc (t->n_sections,
+ sizeof *sections);
+ size_t n_sections = 0;
+
+ size_t n_total_cells = 0;
+ size_t max_depth = 0;
+ for (size_t j = 0; j < t->n_sections; j++)
+ if (t->sections[j].nests[a] == nest)
+ {
+ struct ctables_section *s = &t->sections[j];
+ sections[n_sections++] = s;
+ n_total_cells += s->cells.count;
+
+ size_t depth = s->nests[a]->n;
+ max_depth = MAX (depth, max_depth);
+ }
+
+ struct ctables_cell **sorted = xnmalloc (n_total_cells,
+ sizeof *sorted);
+ size_t n_sorted = 0;
+
+ for (size_t j = 0; j < n_sections; j++)
+ {
+ struct ctables_section *s = sections[j];
+
+ struct ctables_cell *cell;
+ HMAP_FOR_EACH (cell, struct ctables_cell, node, &s->cells)
+ if (!cell->hide)
+ sorted[n_sorted++] = cell;
+ assert (n_sorted <= n_total_cells);
+ }
+
+ struct ctables_cell_sort_aux aux = { .nest = nest, .a = a };
+ sort (sorted, n_sorted, sizeof *sorted, ctables_cell_compare_3way, &aux);
+
+ struct ctables_level
+ {
+ enum ctables_level_type
+ {
+ CTL_VAR, /* Variable label for nest->vars[var_idx]. */
+ CTL_CATEGORY, /* Category for nest->vars[var_idx]. */
+ CTL_SUMMARY, /* Summary functions. */
+ }
+ type;
+
+ size_t var_idx;
+ };
+ struct ctables_level *levels = xnmalloc (1 + 2 * max_depth, sizeof *levels);
+ size_t n_levels = 0;
+ for (size_t k = 0; k < nest->n; k++)
+ {
+ enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[k])];
+ if (vlabel != CTVL_NONE)
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_VAR,
+ .var_idx = k,
+ };
+ }
+
+ if (nest->scale_idx != k
+ && (k != nest->n - 1 || t->label_axis[a] == a))
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_CATEGORY,
+ .var_idx = k,
+ };
+ }
+ }
+
+ if (!summary_dimension && a == t->slabels_axis)
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_SUMMARY,
+ .var_idx = SIZE_MAX,
+ };
+ }
+
+ /* Pivot categories:
+
+ - variable label for nest->vars[0], if vlabel != CTVL_NONE
+ - category for nest->vars[0], if nest->scale_idx != 0
+ - variable label for nest->vars[1], if vlabel != CTVL_NONE
+ - category for nest->vars[1], if nest->scale_idx != 1
+ ...
+ - variable label for nest->vars[n - 1], if vlabel != CTVL_NONE
+ - category for nest->vars[n - 1], if t->label_axis[a] == a && nest->scale_idx != n - 1.
+ - summary function, if 'a == t->slabels_axis && a ==
+ t->summary_axis'.
+
+ Additional dimensions:
+
+ - If 'a == t->slabels_axis && a != t->summary_axis', add a summary
+ dimension.
+ - If 't->label_axis[b] == a' for some 'b != a', add a category
+ dimension to 'a'.
+ */
+
+
+ struct pivot_category **groups = xnmalloc (1 + 2 * max_depth, sizeof *groups);
+ int prev_leaf = 0;
+ for (size_t j = 0; j < n_sorted; j++)
+ {
+ struct ctables_cell *cell = sorted[j];
+ struct ctables_cell *prev = j > 0 ? sorted[j - 1] : NULL;
+
+ size_t n_common = 0;
+ if (j > 0)
+ {
+ for (; n_common < n_levels; n_common++)
+ {
+ const struct ctables_level *level = &levels[n_common];
+ if (level->type == CTL_CATEGORY)
+ {
+ size_t var_idx = level->var_idx;
+ const struct ctables_category *c = cell->axes[a].cvs[var_idx].category;
+ if (prev->axes[a].cvs[var_idx].category != c)
+ break;
+ else if (c->type != CCT_SUBTOTAL
+ && c->type != CCT_HSUBTOTAL
+ && c->type != CCT_TOTAL
+ && !value_equal (&prev->axes[a].cvs[var_idx].value,
+ &cell->axes[a].cvs[var_idx].value,
+ var_get_type (nest->vars[var_idx])))
+ break;
+ }
+ }
+ }
+
+ for (size_t k = n_common; k < n_levels; k++)
+ {
+ const struct ctables_level *level = &levels[k];
+ struct pivot_category *parent = k ? groups[k - 1] : d[a]->root;
+ if (level->type == CTL_SUMMARY)
+ {
+ assert (k == n_levels - 1);
+
+ const struct ctables_summary_spec_set *specs = &t->summary_specs;
+ for (size_t m = 0; m < specs->n; m++)
+ {
+ int leaf = pivot_category_create_leaf (
+ parent, pivot_value_new_text (specs->specs[m].label));
+ if (!m)
+ prev_leaf = leaf;
+ }
+ }
+ else
+ {
+ const struct variable *var = nest->vars[level->var_idx];
+ struct pivot_value *label;
+ if (level->type == CTL_VAR)
+ label = pivot_value_new_variable (var);
+ else if (level->type == CTL_CATEGORY)
+ {
+ const struct ctables_cell_value *cv = &cell->axes[a].cvs[level->var_idx];
+ label = ctables_category_create_label (cv->category,
+ var, &cv->value);
+ }
+ else
+ NOT_REACHED ();
+
+ if (k == n_levels - 1)
+ prev_leaf = pivot_category_create_leaf (parent, label);
+ else
+ groups[k] = pivot_category_create_group__ (parent, label);
+ }
+ }
+
+ cell->axes[a].leaf = prev_leaf;
+ }
+ free (sorted);
+ free (groups);
+ }
+ }
+
+ for (size_t i = 0; i < t->n_sections; i++)
+ {
+ struct ctables_section *s = &t->sections[i];
+
+ struct ctables_cell *cell;
+ HMAP_FOR_EACH (cell, struct ctables_cell, node, &s->cells)
+ {
+ if (cell->hide)
+ continue;
+
+ const struct ctables_nest *specs_nest = s->nests[t->summary_axis];
+ const struct ctables_summary_spec_set *specs = &specs_nest->specs[cell->sv];
+ for (size_t j = 0; j < specs->n; j++)
+ {
+ size_t dindexes[5];
+ size_t n_dindexes = 0;
+
+ if (summary_dimension)
+ dindexes[n_dindexes++] = specs->specs[j].axis_idx;
+
+ if (categories_dimension)
+ {
+ const struct ctables_nest *clabels_nest = s->nests[t->clabels_from_axis];
+ const struct variable *var = clabels_nest->vars[clabels_nest->n - 1];
+ const union value *value = &cell->axes[t->clabels_from_axis].cvs[clabels_nest->n - 1].value;
+ const struct ctables_value *ctv = ctables_value_find (t, value, var_get_width (var));
+ assert (ctv != NULL);
+ dindexes[n_dindexes++] = ctv->leaf;
+ }
+
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (d[a])
+ {
+ int leaf = cell->axes[a].leaf;
+ if (a == t->summary_axis && !summary_dimension)
+ leaf += j;
+ dindexes[n_dindexes++] = leaf;
+ }
+
+ double d = ctables_summary_value (cell, &cell->summaries[j], &specs->specs[j]);
+ struct pivot_value *value = pivot_value_new_number (d);
+ value->numeric.format = specs->specs[j].format;
+ pivot_table_put (pt, dindexes, n_dindexes, value);
+ }
+ }
+ }
+
+ pivot_table_submit (pt);
+}
+
+static bool
+ctables_check_label_position (struct ctables_table *t, enum pivot_axis_type a)
+{
+ enum pivot_axis_type label_pos = t->label_axis[a];
+ if (label_pos == a)
+ return true;
+
+ t->clabels_from_axis = a;
+
+ const char *subcommand_name = a == PIVOT_AXIS_ROW ? "ROWLABELS" : "COLLABELS";
+ const char *pos_name = label_pos == PIVOT_AXIS_LAYER ? "LAYER" : "OPPOSITE";
+
+ const struct ctables_stack *stack = &t->stacks[a];
+ if (!stack->n)
+ return true;
+
+ const struct ctables_nest *n0 = &stack->nests[0];
+ assert (n0->n > 0);
+ const struct variable *v0 = n0->vars[n0->n - 1];
+ struct ctables_categories *c0 = t->categories[var_get_dict_index (v0)];
+ t->clabels_example = v0;
+
+ for (size_t i = 0; i < c0->n_cats; i++)
+ if (c0->cats[i].type == CCT_FUNCTION)
+ {
+ msg (SE, _("%s=%s is not allowed with sorting based "
+ "on a summary function."),
+ subcommand_name, pos_name);
+ return false;
+ }
+ if (n0->n - 1 == n0->scale_idx)
+ {
+ msg (SE, _("%s=%s requires the variables to be moved to be categorical, "
+ "but %s is a scale variable."),
+ subcommand_name, pos_name, var_get_name (v0));
+ return false;
+ }
+
+ for (size_t i = 1; i < stack->n; i++)
+ {
+ const struct ctables_nest *ni = &stack->nests[i];
+ assert (ni->n > 0);
+ const struct variable *vi = ni->vars[ni->n - 1];
+ struct ctables_categories *ci = t->categories[var_get_dict_index (vi)];
+
+ if (ni->n - 1 == ni->scale_idx)
+ {
+ msg (SE, _("%s=%s requires the variables to be moved to be "
+ "categorical, but %s is a scale variable."),
+ subcommand_name, pos_name, var_get_name (vi));
+ return false;
+ }
+ if (var_get_width (v0) != var_get_width (vi))
+ {
+ msg (SE, _("%s=%s requires the variables to be "
+ "moved to have the same width, but %s has "
+ "width %d and %s has width %d."),
+ subcommand_name, pos_name,
+ var_get_name (v0), var_get_width (v0),
+ var_get_name (vi), var_get_width (vi));
+ return false;
+ }
+ if (!val_labs_equal (var_get_value_labels (v0),
+ var_get_value_labels (vi)))
+ {
+ msg (SE, _("%s=%s requires the variables to be "
+ "moved to have the same value labels, but %s "
+ "and %s have different value labels."),
+ subcommand_name, pos_name,
+ var_get_name (v0), var_get_name (vi));
+ return false;
+ }
+ if (!ctables_categories_equal (c0, ci))
+ {
+ msg (SE, _("%s=%s requires the variables to be "
+ "moved to have the same category "
+ "specifications, but %s and %s have different "
+ "category specifications."),
+ subcommand_name, pos_name,
+ var_get_name (v0), var_get_name (vi));
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static bool
+ctables_prepare_table (struct ctables_table *t)
+{
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (t->axes[a])
+ {
+ t->stacks[a] = enumerate_fts (a, t->axes[a]);
+
+ for (size_t j = 0; j < t->stacks[a].n; j++)
+ {
+ struct ctables_nest *nest = &t->stacks[a].nests[j];
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ {
+ nest->domains[dt] = xmalloc (nest->n * sizeof *nest->domains[dt]);
+ nest->n_domains[dt] = 0;
+
+ for (size_t k = 0; k < nest->n; k++)
+ {
+ if (k == nest->scale_idx)
+ continue;
+
+ switch (dt)
+ {
+ case CTDT_TABLE:
+ continue;
+
+ case CTDT_LAYER:
+ if (a != PIVOT_AXIS_LAYER)
+ continue;
+ break;
+
+ case CTDT_SUBTABLE:
+ case CTDT_ROW:
+ case CTDT_COL:
+ if (dt == CTDT_SUBTABLE ? a != PIVOT_AXIS_LAYER
+ : dt == CTDT_ROW ? a == PIVOT_AXIS_COLUMN
+ : a == PIVOT_AXIS_ROW)
+ {
+ if (k == nest->n - 1
+ || (nest->scale_idx == nest->n - 1
+ && k == nest->n - 2))
+ continue;
+ }
+ break;
+
+ case CTDT_LAYERROW:
+ if (a == PIVOT_AXIS_COLUMN)
+ continue;
+ break;
+
+ case CTDT_LAYERCOL:
+ if (a == PIVOT_AXIS_ROW)
+ continue;
+ break;
+ }
+
+ nest->domains[dt][nest->n_domains[dt]++] = k;
+ }
+ }
+ }
+ }
+ else
+ {
+ struct ctables_nest *nest = xmalloc (sizeof *nest);
+ *nest = (struct ctables_nest) { .n = 0 };
+ t->stacks[a] = (struct ctables_stack) { .nests = nest, .n = 1 };
+ }
+
+ struct ctables_stack *stack = &t->stacks[t->summary_axis];
+ for (size_t i = 0; i < stack->n; i++)
+ {
+ struct ctables_nest *nest = &stack->nests[i];
+ if (!nest->specs[CSV_CELL].n)
+ {
+ struct ctables_summary_spec_set *specs = &nest->specs[CSV_CELL];
+ specs->specs = xmalloc (sizeof *specs->specs);
+ specs->n = 1;
+
+ enum ctables_summary_function function
+ = specs->var ? CTSF_MEAN : CTSF_COUNT;
+ struct ctables_var var = { .is_mrset = false, .var = specs->var };
+
+ *specs->specs = (struct ctables_summary_spec) {
+ .function = function,
+ .format = ctables_summary_default_format (function, &var),
+ .label = ctables_summary_default_label (function, 0),
+ };
+ if (!specs->var)
+ specs->var = nest->vars[0];
+
+ ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL],
+ &nest->specs[CSV_CELL]);
+ }
+ else if (!nest->specs[CSV_TOTAL].n)
+ ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL],
+ &nest->specs[CSV_CELL]);
+ }
+
+ struct ctables_summary_spec_set *merged = &t->summary_specs;
+ struct merge_item *items = xnmalloc (2 * stack->n, sizeof *items);
+ size_t n_left = 0;
+ for (size_t j = 0; j < stack->n; j++)
+ {
+ const struct ctables_nest *nest = &stack->nests[j];
+ if (nest->n)
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ items[n_left++] = (struct merge_item) { .set = &nest->specs[sv] };
+ }
+
+ while (n_left > 0)
+ {
+ struct merge_item min = items[0];
+ for (size_t j = 1; j < n_left; j++)
+ if (merge_item_compare_3way (&items[j], &min) < 0)
+ min = items[j];
+
+ if (merged->n >= merged->allocated)
+ merged->specs = x2nrealloc (merged->specs, &merged->allocated,
+ sizeof *merged->specs);
+ merged->specs[merged->n++] = min.set->specs[min.ofs];
+
+ for (size_t j = 0; j < n_left; )
+ {
+ if (merge_item_compare_3way (&items[j], &min) == 0)
+ {
+ struct merge_item *item = &items[j];
+ item->set->specs[item->ofs].axis_idx = merged->n - 1;
+ if (++item->ofs >= item->set->n)
+ {
+ items[j] = items[--n_left];
+ continue;
+ }
+ }
+ j++;
+ }
+ }
+
+#if 0
+ for (size_t j = 0; j < merged->n; j++)
+ printf ("%s\n", ctables_summary_function_name (merged->specs[j].function));
+
+ for (size_t j = 0; j < stack->n; j++)
+ {
+ const struct ctables_nest *nest = &stack->nests[j];
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ {
+ const struct ctables_summary_spec_set *specs = &nest->specs[sv];
+ for (size_t k = 0; k < specs->n; k++)
+ printf ("(%s, %zu) ", ctables_summary_function_name (specs->specs[k].function),
+ specs->specs[k].axis_idx);
+ printf ("\n");
+ }
+ }
+#endif
+
+ return (ctables_check_label_position (t, PIVOT_AXIS_ROW)
+ && ctables_check_label_position (t, PIVOT_AXIS_COLUMN));
+}
+
+static void
+ctables_insert_clabels_values (struct ctables_table *t, const struct ccase *c,
+ enum pivot_axis_type a)
+{
+ struct ctables_stack *stack = &t->stacks[a];
+ for (size_t i = 0; i < stack->n; i++)
+ {
+ const struct ctables_nest *nest = &stack->nests[i];
+ const struct variable *var = nest->vars[nest->n - 1];
+ int width = var_get_width (var);
+ const union value *value = case_data (c, var);
+
+ if (var_is_numeric (var) && value->f == SYSMIS)
+ continue;
+
+ if (!ctables_categories_match (t->categories [var_get_dict_index (var)],
+ value, var))
+ continue;
+
+ unsigned int hash = value_hash (value, width, 0);
+
+ struct ctables_value *clv = ctables_value_find__ (t, value, width, hash);
+ if (!clv)
+ {
+ clv = xmalloc (sizeof *clv);
+ value_clone (&clv->value, value, width);
+ hmap_insert (&t->clabels_values_map, &clv->node, hash);
+ }
+ }
+}
+
+static int
+compare_clabels_values_3way (const void *a_, const void *b_, const void *width_)
+{
+ const struct ctables_value *const *ap = a_;
+ const struct ctables_value *const *bp = b_;
+ const struct ctables_value *a = *ap;
+ const struct ctables_value *b = *bp;
+ const int *width = width_;
+ return value_compare_3way (&a->value, &b->value, *width);
+}
+
+static void
+ctables_sort_clabels_values (struct ctables_table *t)
+{
+ int width = var_get_width (t->clabels_example);
+
+ size_t n = hmap_count (&t->clabels_values_map);
+ t->clabels_values = xnmalloc (n, sizeof *t->clabels_values);
+
+ struct ctables_value *clv;
+ size_t i = 0;
+ HMAP_FOR_EACH (clv, struct ctables_value, node, &t->clabels_values_map)
+ t->clabels_values[i++] = clv;
+ t->n_clabels_values = n;
+ assert (i == n);
+
+ sort (t->clabels_values, n, sizeof *t->clabels_values,
+ compare_clabels_values_3way, &width);
+
+ for (size_t i = 0; i < n; i++)
+ t->clabels_values[i]->leaf = i;
+}
+
+static void
+ctables_add_category_occurrences (const struct variable *var,
+ struct hmap *occurrences,
+ const struct ctables_categories *cats)
+{
+ const struct val_labs *val_labs = var_get_value_labels (var);
+
+ for (size_t i = 0; i < cats->n_cats; i++)
+ {
+ const struct ctables_category *c = &cats->cats[i];
+ switch (c->type)
+ {
+ case CCT_NUMBER:
+ ctables_add_occurrence (var, &(const union value) { .f = c->number },
+ occurrences);
+ break;
+
+ case CCT_STRING:
+ abort (); /* XXX */
+
+ case CCT_RANGE:
+ assert (var_is_numeric (var));
+ for (const struct val_lab *vl = val_labs_first (val_labs); vl;
+ vl = val_labs_next (val_labs, vl))
+ if (vl->value.f >= c->range[0] && vl->value.f <= c->range[1])
+ ctables_add_occurrence (var, &vl->value, occurrences);
+ break;
+
+ case CCT_MISSING:
+ for (const struct val_lab *vl = val_labs_first (val_labs); vl;
+ vl = val_labs_next (val_labs, vl))
+ if (var_is_value_missing (var, &vl->value))
+ ctables_add_occurrence (var, &vl->value, occurrences);
+ break;
+
+ case CCT_OTHERNM:
+ for (const struct val_lab *vl = val_labs_first (val_labs); vl;
+ vl = val_labs_next (val_labs, vl))
+ ctables_add_occurrence (var, &vl->value, occurrences);
+ break;
+
+ case CCT_SUBTOTAL:
+ case CCT_HSUBTOTAL:
+ case CCT_TOTAL:
+ break;
+
+ case CCT_VALUE:
+ case CCT_LABEL:
+ case CCT_FUNCTION:
+ for (const struct val_lab *vl = val_labs_first (val_labs); vl;
+ vl = val_labs_next (val_labs, vl))
+ if (c->include_missing || !var_is_value_missing (var, &vl->value))
+ ctables_add_occurrence (var, &vl->value, occurrences);
+ break;
+ }
+ }
+}
+
+static void
+ctables_section_recurse_add_empty_categories (
+ struct ctables_section *s,
+ const struct ctables_category *cats[PIVOT_N_AXES][10], struct ccase *c,
+ enum pivot_axis_type a, size_t a_idx)
+{
+ if (a >= PIVOT_N_AXES)
+ ctables_cell_insert__ (s, c, cats);
+ else if (!s->nests[a] || a_idx >= s->nests[a]->n)
+ ctables_section_recurse_add_empty_categories (s, cats, c, a + 1, 0);
+ else
+ {
+ const struct variable *var = s->nests[a]->vars[a_idx];
+ int width = var_get_width (var);
+ const struct hmap *occurrences = &s->occurrences[a][a_idx];
+ const struct ctables_section_value *sv;
+ HMAP_FOR_EACH (sv, struct ctables_section_value, node, occurrences)
+ {
+ union value *value = case_data_rw (c, var);
+ value_destroy (value, width);
+ value_clone (value, &sv->value, width);
+ cats[a][a_idx] = ctables_categories_match (
+ s->table->categories[var_get_dict_index (var)], value, var);
+ assert (cats[a][a_idx] != NULL);
+ ctables_section_recurse_add_empty_categories (s, cats, c, a, a_idx + 1);
+ }
+ }
+}
+
+static void
+ctables_section_add_empty_categories (struct ctables_section *s)
+{
+ bool show_empty = false;
+ for (size_t a = 0; a < PIVOT_N_AXES; a++)
+ if (s->nests[a])
+ for (size_t k = 0; k < s->nests[a]->n; k++)
+ if (k != s->nests[a]->scale_idx)
+ {
+ const struct variable *var = s->nests[a]->vars[k];
+ const struct ctables_categories *cats = s->table->categories[
+ var_get_dict_index (var)];
+ if (cats->show_empty)
+ {
+ show_empty = true;
+ ctables_add_category_occurrences (var, &s->occurrences[a][k], cats);
+ }
+ }
+ if (!show_empty)
+ return;
+
+ const struct ctables_category *cats[PIVOT_N_AXES][10]; /* XXX */
+ struct ccase *c = case_create (dict_get_proto (s->table->ctables->dict));
+ ctables_section_recurse_add_empty_categories (s, cats, c, 0, 0);
+ case_unref (c);
+}
+
+static bool
+ctables_execute (struct dataset *ds, struct ctables *ct)
+{
+ for (size_t i = 0; i < ct->n_tables; i++)
+ {
+ struct ctables_table *t = ct->tables[i];
+ t->sections = xnmalloc (MAX (1, t->stacks[PIVOT_AXIS_ROW].n) *
+ MAX (1, t->stacks[PIVOT_AXIS_COLUMN].n) *
+ MAX (1, t->stacks[PIVOT_AXIS_LAYER].n),
+ sizeof *t->sections);
+ size_t ix[PIVOT_N_AXES];
+ ctables_table_add_section (t, 0, ix);
+ }
+
+ struct casereader *input = proc_open (ds);
+ bool warn_on_invalid = true;
+ for (struct ccase *c = casereader_read (input); c;
+ case_unref (c), c = casereader_read (input))
+ {
+ double d_weight = dict_get_case_weight (dataset_dict (ds), c,
+ &warn_on_invalid);
+ double e_weight = (ct->e_weight
+ ? var_force_valid_weight (ct->e_weight,
+ case_num (c, ct->e_weight),
+ &warn_on_invalid)
+ : d_weight);
+
+ for (size_t i = 0; i < ct->n_tables; i++)
+ {
+ struct ctables_table *t = ct->tables[i];
+
+ for (size_t j = 0; j < t->n_sections; j++)
+ ctables_cell_insert (&t->sections[j], c, d_weight, e_weight);
+
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (t->label_axis[a] != a)
+ ctables_insert_clabels_values (t, c, a);
+ }
+ }
+ casereader_destroy (input);
+
+ for (size_t i = 0; i < ct->n_tables; i++)
+ {
+ struct ctables_table *t = ct->tables[i];
+
+ if (t->clabels_example)
+ ctables_sort_clabels_values (t);
+
+ for (size_t j = 0; j < t->n_sections; j++)
+ ctables_section_add_empty_categories (&t->sections[j]);
+
+ ctables_table_output (ct, ct->tables[i]);
+ }
+ return proc_commit (ds);
+}
+\f
+/* Postcomputes. */
+
+typedef struct ctables_pcexpr *parse_recursively_func (struct lexer *);
+
+static void
+ctables_pcexpr_destroy (struct ctables_pcexpr *e)
+{
+ if (e)
+ {
+ switch (e->op)
+ {
+ case CTPO_CAT_STRING:
+ free (e->string);
+ break;
+
+ case CTPO_ADD:
+ case CTPO_SUB:
+ case CTPO_MUL:
+ case CTPO_DIV:
+ case CTPO_POW:
+ case CTPO_NEG:
+ for (size_t i = 0; i < 2; i++)
+ ctables_pcexpr_destroy (e->subs[i]);
+ break;
+
+ case CTPO_CONSTANT:
+ case CTPO_CAT_NUMBER:
+ case CTPO_CAT_RANGE:
+ case CTPO_CAT_MISSING:
+ case CTPO_CAT_OTHERNM:
+ case CTPO_CAT_SUBTOTAL:
+ case CTPO_CAT_TOTAL:
+ break;
+ }
+
+ msg_location_destroy (e->location);
+ free (e);
+ }
+}
+
+static struct ctables_pcexpr *
+ctables_pcexpr_allocate_binary (enum ctables_postcompute_op op,
+ struct ctables_pcexpr *sub0,
+ struct ctables_pcexpr *sub1)
+{
+ struct ctables_pcexpr *e = xmalloc (sizeof *e);
+ *e = (struct ctables_pcexpr) {
+ .op = op,
+ .subs = { sub0, sub1 },
+ .ofs = { sub0->ofs[0], sub1->ofs[1] }
+ };
+ return e;
+}
+
+static struct msg_location *
+ctables_pcexpr_location (struct lexer *lexer, const struct ctables_pcexpr *e_)
+{
+ if (!e_->location)
+ {
+ struct ctables_pcexpr *e = CONST_CAST (struct ctables_pcexpr *, e_);
+ e->location = lex_ofs_location (lexer, e->ofs[0], e->ofs[1]);
+ }
+ return e_->location;
+}
+
+/* How to parse an operator. */
+struct operator
+ {
+ enum token_type token;
+ enum ctables_postcompute_op op;
+ };
+
+static const struct operator *
+match_operator (struct lexer *lexer, const struct operator ops[], size_t n_ops)
+{
+ for (const struct operator *op = ops; op < ops + n_ops; op++)
+ if (lex_token (lexer) == op->token)
+ {
+ if (op->token != T_NEG_NUM)
+ lex_get (lexer);
+
+ return op;
+ }
+
+ return NULL;
+}
+
+static struct ctables_pcexpr *
+parse_binary_operators__ (struct lexer *lexer,
+ const struct operator ops[], size_t n_ops,
+ parse_recursively_func *parse_next_level,
+ const char *chain_warning,
+ struct ctables_pcexpr *lhs)
+{
+ for (int op_count = 0; ; op_count++)
+ {
+ const struct operator *op = match_operator (lexer, ops, n_ops);
+ if (!op)
+ {
+ if (op_count > 1 && chain_warning)
+ msg_at (SW, ctables_pcexpr_location (lexer, lhs),
+ "%s", chain_warning);
+
+ return lhs;
+ }
+
+ struct ctables_pcexpr *rhs = parse_next_level (lexer);
+ if (!rhs)
+ {
+ ctables_pcexpr_destroy (lhs);
+ return NULL;
+ }
+
+ lhs = ctables_pcexpr_allocate_binary (op->op, lhs, rhs);
+ }
+}
+
+static struct ctables_pcexpr *
+parse_binary_operators (struct lexer *lexer,
+ const struct operator ops[], size_t n_ops,
+ parse_recursively_func *parse_next_level,
+ const char *chain_warning)
+{
+ struct ctables_pcexpr *lhs = parse_next_level (lexer);
+ if (!lhs)
+ return NULL;
+
+ return parse_binary_operators__ (lexer, ops, n_ops, parse_next_level,
+ chain_warning, lhs);
+}
+
+static struct ctables_pcexpr *parse_add (struct lexer *);
+
+static struct ctables_pcexpr
+ctpo_cat_range (double low, double high)
+{
+ return (struct ctables_pcexpr) {
+ .op = CTPO_CAT_RANGE,
+ .range = { low, high },
+ };
+}
+
+static struct ctables_pcexpr *
+parse_primary (struct lexer *lexer)
+{
+ int start_ofs = lex_ofs (lexer);
+ struct ctables_pcexpr e;
+ if (lex_is_number (lexer))
+ {
+ e = (struct ctables_pcexpr) { .op = CTPO_CONSTANT,
+ .number = lex_number (lexer) };
+ lex_get (lexer);
+ }
+ else if (lex_match_id (lexer, "MISSING"))
+ e = (struct ctables_pcexpr) { .op = CTPO_CAT_MISSING };
+ else if (lex_match_id (lexer, "OTHERNM"))
+ e = (struct ctables_pcexpr) { .op = CTPO_CAT_OTHERNM };
+ else if (lex_match_id (lexer, "TOTAL"))
+ e = (struct ctables_pcexpr) { .op = CTPO_CAT_TOTAL };
+ else if (lex_match_id (lexer, "SUBTOTAL"))
+ {
+ size_t subtotal_index = 0;
+ if (lex_match (lexer, T_LBRACK))
+ {
+ if (!lex_force_int_range (lexer, "SUBTOTAL", 1, LONG_MAX))
+ return NULL;
+ subtotal_index = lex_integer (lexer);
+ lex_get (lexer);
+ if (!lex_force_match (lexer, T_RBRACK))
+ return NULL;
+ }
+ e = (struct ctables_pcexpr) { .op = CTPO_CAT_SUBTOTAL,
+ .subtotal_index = subtotal_index };
+ }
+ else if (lex_match (lexer, T_LBRACK))
+ {
+ if (lex_match_id (lexer, "LO"))
+ {
+ if (!lex_force_match_id (lexer, "THRU") || lex_force_num (lexer))
+ return false;
+ e = ctpo_cat_range (-DBL_MAX, lex_number (lexer));
+ lex_get (lexer);
+ }
+ else if (lex_is_number (lexer))
+ {
+ double number = lex_number (lexer);
+ lex_get (lexer);
+ if (lex_match_id (lexer, "THRU"))
+ {
+ if (lex_match_id (lexer, "HI"))
+ e = ctpo_cat_range (number, DBL_MAX);
+ else
+ {
+ if (!lex_force_num (lexer))
+ return false;
+ e = ctpo_cat_range (number, lex_number (lexer));
+ lex_get (lexer);
+ }
+ }
+ else
+ e = (struct ctables_pcexpr) { .op = CTPO_CAT_NUMBER,
+ .number = number };
+ }
+ else if (lex_is_string (lexer))
+ {
+ e = (struct ctables_pcexpr) {
+ .op = CTPO_CAT_STRING,
+ .string = ss_xstrdup (lex_tokss (lexer)),
+ };
+ lex_get (lexer);
+ }
+ else
+ {
+ lex_error (lexer, NULL);
+ return NULL;
+ }
+
+ if (!lex_force_match (lexer, T_RBRACK))
+ {
+ if (e.op == CTPO_CAT_STRING)
+ free (e.string);
+ return NULL;
+ }
+ }
+ else if (lex_match (lexer, T_LPAREN))
+ {
+ struct ctables_pcexpr *ep = parse_add (lexer);
+ if (!ep)
+ return NULL;
+ if (!lex_force_match (lexer, T_RPAREN))
+ {
+ ctables_pcexpr_destroy (ep);
+ return NULL;
+ }
+ return ep;
+ }
+ else
+ {
+ lex_error (lexer, NULL);
+ return NULL;
+ }
+
+ e.ofs[0] = start_ofs;
+ e.ofs[1] = lex_ofs (lexer) - 1;
+ return xmemdup (&e, sizeof e);
}
-static const struct ctables_category *
-ctables_categories_total (const struct ctables_categories *c)
+static struct ctables_pcexpr *
+ctables_pcexpr_allocate_neg (struct ctables_pcexpr *sub,
+ struct lexer *lexer, int start_ofs)
{
- const struct ctables_category *first = &c->cats[0];
- const struct ctables_category *last = &c->cats[c->n_cats - 1];
- return (first->type == CCT_TOTAL ? first
- : last->type == CCT_TOTAL ? last
- : NULL);
+ struct ctables_pcexpr *e = xmalloc (sizeof *e);
+ *e = (struct ctables_pcexpr) {
+ .op = CTPO_NEG,
+ .subs = { sub },
+ .ofs = { start_ofs, lex_ofs (lexer) - 1 },
+ };
+ return e;
}
-static void
-ctables_cell_insert__ (struct ctables_table *t, const struct ccase *c,
- size_t ix[PIVOT_N_AXES],
- const struct ctables_category *cats[PIVOT_N_AXES][10],
- double weight)
+static struct ctables_pcexpr *
+parse_exp (struct lexer *lexer)
{
- const struct var_array *ss = &t->vaas[t->summary_axis].vas[ix[t->summary_axis]];
-
- size_t hash = 0;
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- {
- const struct var_array *va = &t->vaas[a].vas[ix[a]];
- hash = hash_int (ix[a], hash);
- for (size_t i = 0; i < va->n; i++)
- if (i != va->scale_idx)
- {
- hash = hash_pointer (cats[a][i], hash);
- if (cats[a][i]->type != CCT_TOTAL)
- hash = value_hash (case_data (c, va->vars[i]),
- var_get_width (va->vars[i]), hash);
- }
- }
+ static const struct operator op = { T_EXP, CTPO_POW };
+
+ const char *chain_warning =
+ _("The exponentiation operator (`**') is left-associative: "
+ "`a**b**c' equals `(a**b)**c', not `a**(b**c)'. "
+ "To disable this warning, insert parentheses.");
+
+ if (lex_token (lexer) != T_NEG_NUM || lex_next_token (lexer, 1) != T_EXP)
+ return parse_binary_operators (lexer, &op, 1,
+ parse_primary, chain_warning);
+
+ /* Special case for situations like "-5**6", which must be parsed as
+ -(5**6). */
+
+ int start_ofs = lex_ofs (lexer);
+ struct ctables_pcexpr *lhs = xmalloc (sizeof *lhs);
+ *lhs = (struct ctables_pcexpr) {
+ .op = CTPO_CONSTANT,
+ .number = -lex_tokval (lexer),
+ .ofs = { start_ofs, lex_ofs (lexer) },
+ };
+ lex_get (lexer);
- struct ctables_cell *f;
- HMAP_FOR_EACH_WITH_HASH (f, struct ctables_cell, node, hash, &t->cells)
- {
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- {
- const struct var_array *va = &t->vaas[a].vas[ix[a]];
- if (f->axes[a].vaa_idx != ix[a])
- goto not_equal;
- for (size_t i = 0; i < va->n; i++)
- if (i != va->scale_idx
- && (cats[a][i] != f->axes[a].cvs[i].category
- || (cats[a][i]->type != CCT_TOTAL
- && !value_equal (case_data (c, va->vars[i]),
- &f->axes[a].cvs[i].value,
- var_get_width (va->vars[i])))))
- goto not_equal;
- }
+ struct ctables_pcexpr *node = parse_binary_operators__ (
+ lexer, &op, 1, parse_primary, chain_warning, lhs);
+ if (!node)
+ return NULL;
- goto summarize;
+ return ctables_pcexpr_allocate_neg (node, lexer, start_ofs);
+}
- not_equal: ;
- }
+/* Parses the unary minus level. */
+static struct ctables_pcexpr *
+parse_neg (struct lexer *lexer)
+{
+ int start_ofs = lex_ofs (lexer);
+ if (!lex_match (lexer, T_DASH))
+ return parse_exp (lexer);
- f = xmalloc (sizeof *f);
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- {
- const struct var_array *va = &t->vaas[a].vas[ix[a]];
- f->axes[a].vaa_idx = ix[a];
- f->axes[a].cvs = (va->n
- ? xnmalloc (va->n, sizeof *f->axes[a].cvs)
- : NULL);
- for (size_t i = 0; i < va->n; i++)
- {
- f->axes[a].cvs[i].category = cats[a][i];
- value_clone (&f->axes[a].cvs[i].value, case_data (c, va->vars[i]),
- var_get_width (va->vars[i]));
- }
- }
- f->summaries = xmalloc (ss->n_summaries * sizeof *f->summaries);
- for (size_t i = 0; i < ss->n_summaries; i++)
- ctables_summary_init (&f->summaries[i], &ss->summaries[i]);
- for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
- f->domains[dt] = ctables_domain_insert (t, f, dt);
- hmap_insert (&t->cells, &f->node, hash);
+ struct ctables_pcexpr *inner = parse_neg (lexer);
+ if (!inner)
+ return NULL;
-summarize:
- for (size_t i = 0; i < ss->n_summaries; i++)
- ctables_summary_add (&f->summaries[i], &ss->summaries[i], ss->summary_var,
- case_data (c, ss->summary_var), weight);
- for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
- f->domains[dt]->valid += weight;
+ return ctables_pcexpr_allocate_neg (inner, lexer, start_ofs);
}
-static void
-recurse_totals (struct ctables_table *t, const struct ccase *c,
- size_t ix[PIVOT_N_AXES],
- const struct ctables_category *cats[PIVOT_N_AXES][10],
- double weight,
- enum pivot_axis_type start_a, size_t start_va)
+/* Parses the multiplication and division level. */
+static struct ctables_pcexpr *
+parse_mul (struct lexer *lexer)
{
- for (enum pivot_axis_type a = start_a; a < PIVOT_N_AXES; a++)
+ static const struct operator ops[] =
{
- const struct var_array *va = &t->vaas[a].vas[ix[a]];
- for (size_t i = start_va; i < va->n; i++)
- {
- if (i == va->scale_idx)
- continue;
-
- const struct variable *var = va->vars[i];
+ { T_ASTERISK, CTPO_MUL },
+ { T_SLASH, CTPO_DIV },
+ };
- const struct ctables_category *total = ctables_categories_total (
- t->categories[var_get_dict_index (var)]);
- if (total)
- {
- const struct ctables_category *save = cats[a][i];
- cats[a][i] = total;
- ctables_cell_insert__ (t, c, ix, cats, weight);
- recurse_totals (t, c, ix, cats, weight, a, i + 1);
- cats[a][i] = save;
- }
- }
- start_va = 0;
- }
+ return parse_binary_operators (lexer, ops, sizeof ops / sizeof *ops,
+ parse_neg, NULL);
}
-static void
-ctables_cell_insert (struct ctables_table *t,
- const struct ccase *c,
- size_t ir, size_t ic, size_t il,
- double weight)
+/* Parses the addition and subtraction level. */
+static struct ctables_pcexpr *
+parse_add (struct lexer *lexer)
{
- size_t ix[PIVOT_N_AXES] = {
- [PIVOT_AXIS_ROW] = ir,
- [PIVOT_AXIS_COLUMN] = ic,
- [PIVOT_AXIS_LAYER] = il,
- };
-
- const struct ctables_category *cats[PIVOT_N_AXES][10];
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ static const struct operator ops[] =
{
- const struct var_array *va = &t->vaas[a].vas[ix[a]];
- for (size_t i = 0; i < va->n; i++)
- {
- if (i == va->scale_idx)
- continue;
-
- const struct variable *var = va->vars[i];
- const union value *value = case_data (c, var);
-
- if (var_is_numeric (var) && value->f == SYSMIS)
- return;
-
- cats[a][i] = ctables_categories_match (
- t->categories[var_get_dict_index (var)], value, var);
- if (!cats[a][i])
- return;
- }
- }
+ { T_PLUS, CTPO_ADD },
+ { T_DASH, CTPO_SUB },
+ { T_NEG_NUM, CTPO_ADD },
+ };
- ctables_cell_insert__ (t, c, ix, cats, weight);
+ return parse_binary_operators (lexer, ops, sizeof ops / sizeof *ops,
+ parse_mul, NULL);
+}
- recurse_totals (t, c, ix, cats, weight, 0, 0);
+static struct ctables_postcompute *
+ctables_find_postcompute (struct ctables *ct, const char *name)
+{
+ struct ctables_postcompute *pc;
+ HMAP_FOR_EACH_WITH_HASH (pc, struct ctables_postcompute, hmap_node,
+ utf8_hash_case_string (name, 0), &ct->postcomputes)
+ if (!utf8_strcasecmp (pc->name, name))
+ return pc;
+ return NULL;
}
static bool
-ctables_execute (struct dataset *ds, struct ctables *ct)
+ctables_parse_pcompute (struct lexer *lexer, struct ctables *ct)
{
- for (size_t i = 0; i < ct->n_tables; i++)
- {
- struct ctables_table *t = ct->tables[i];
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- if (t->axes[a])
- {
- t->vaas[a] = enumerate_fts (a, t->axes[a]);
+ int start_ofs = lex_ofs (lexer) - 1;
- for (size_t j = 0; j < t->vaas[a].n; j++)
- {
- struct var_array *va = &t->vaas[a].vas[j];
- for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
- {
- va->domains[dt] = xmalloc (va->n * sizeof *va->domains[dt]);
- va->n_domains[dt] = 0;
+ if (!lex_force_match (lexer, T_AND) || !lex_force_id (lexer))
+ return false;
- for (size_t k = 0; k < va->n; k++)
- {
- if (k == va->scale_idx)
- continue;
+ char *name = ss_xstrdup (lex_tokss (lexer));
- switch (dt)
- {
- case CTDT_TABLE:
- continue;
+ lex_get (lexer);
+ if (!lex_force_match (lexer, T_EQUALS)
+ || !lex_force_match_id (lexer, "EXPR")
+ || !lex_force_match (lexer, T_LPAREN))
+ {
+ free (name);
+ return false;
+ }
- case CTDT_LAYER:
- if (a != PIVOT_AXIS_LAYER)
- continue;
- break;
+ struct ctables_pcexpr *expr = parse_add (lexer);
+ if (!expr || !lex_force_match (lexer, T_RPAREN))
+ {
+ free (name);
+ return false;
+ }
- case CTDT_SUBTABLE:
- case CTDT_ROW:
- case CTDT_COL:
- if (dt == CTDT_SUBTABLE ? a != PIVOT_AXIS_LAYER
- : dt == CTDT_ROW ? a == PIVOT_AXIS_COLUMN
- : a == PIVOT_AXIS_ROW)
- {
- if (k == va->n - 1
- || (va->scale_idx == va->n - 1
- && k == va->n - 2))
- continue;
- }
- break;
+ struct msg_location *location = lex_ofs_location (lexer, start_ofs,
+ lex_ofs (lexer) - 1);
- case CTDT_LAYERROW:
- if (a == PIVOT_AXIS_COLUMN)
- continue;
- break;
+ struct ctables_postcompute *pc = ctables_find_postcompute (ct, name);
+ if (pc)
+ {
+ msg_at (SW, location, _("New definition of &%s will override the "
+ "previous definition."),
+ pc->name);
+ msg_at (SN, pc->location, _("This is the previous definition."));
+
+ ctables_pcexpr_destroy (pc->expr);
+ msg_location_destroy (pc->location);
+ free (name);
+ }
+ else
+ {
+ pc = xmalloc (sizeof *pc);
+ *pc = (struct ctables_postcompute) { .name = name };
+ hmap_insert (&ct->postcomputes, &pc->hmap_node,
+ utf8_hash_case_string (pc->name, 0));
+ }
+ pc->expr = expr;
+ pc->location = location;
+ return true;
+}
- case CTDT_LAYERCOL:
- if (a == PIVOT_AXIS_ROW)
- continue;
- break;
- }
+static bool
+ctables_parse_pproperties_format (struct lexer *lexer,
+ struct ctables_summary_spec_set *sss)
+{
+ *sss = (struct ctables_summary_spec_set) { .n = 0 };
- va->domains[dt][va->n_domains[dt]++] = k;
- }
- }
- }
- }
- else
- {
- struct var_array *va = xmalloc (sizeof *va);
- *va = (struct var_array) { .n = 0 };
- t->vaas[a] = (struct var_array2) { .vas = va, .n = 1 };
- }
+ while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH
+ && !(lex_token (lexer) == T_ID
+ && (lex_id_match (ss_cstr ("LABEL"), lex_tokss (lexer))
+ || lex_id_match (ss_cstr ("HIDESOURCECATS"),
+ lex_tokss (lexer)))))
+ {
+ /* Parse function. */
+ enum ctables_summary_function function;
+ if (!parse_ctables_summary_function (lexer, &function))
+ goto error;
- for (size_t i = 0; i < t->vaas[t->summary_axis].n; i++)
+ /* Parse percentile. */
+ double percentile = 0;
+ if (function == CTSF_PTILE)
{
- struct var_array *va = &t->vaas[t->summary_axis].vas[i];
- if (!va->n_summaries)
- {
- va->summaries = xmalloc (sizeof *va->summaries);
- va->n_summaries = 1;
+ if (!lex_force_num_range_closed (lexer, "PTILE", 0, 100))
+ goto error;
+ percentile = lex_number (lexer);
+ lex_get (lexer);
+ }
- enum ctables_summary_function function
- = va->summary_var ? CTSF_MEAN : CTSF_COUNT;
- struct ctables_var var = { .is_mrset = false, .var = va->summary_var };
+ /* Parse format. */
+ struct fmt_spec format;
+ if (!parse_format_specifier (lexer, &format)
+ || !fmt_check_output (&format)
+ || !fmt_check_type_compat (&format, VAL_NUMERIC))
+ goto error;
- *va->summaries = (struct ctables_summary_spec) {
- .function = function,
- .format = ctables_summary_default_format (function, &var),
- .label = ctables_summary_default_label (function, 0),
- };
- if (!va->summary_var)
- va->summary_var = va->vars[0];
- }
- }
+ if (sss->n >= sss->allocated)
+ sss->specs = x2nrealloc (sss->specs, &sss->allocated,
+ sizeof *sss->specs);
+ sss->specs[sss->n++] = (struct ctables_summary_spec) {
+ .function = function,
+ .percentile = percentile,
+ .format = format,
+ };
}
+ return true;
- struct casereader *input = casereader_create_filter_weight (proc_open (ds),
- dataset_dict (ds),
- NULL, NULL);
- bool warn_on_invalid = true;
- double total_weight = 0;
- for (struct ccase *c = casereader_read (input); c;
- case_unref (c), c = casereader_read (input))
- {
- double weight = dict_get_case_weight (dataset_dict (ds), c,
- &warn_on_invalid);
- total_weight += weight;
+error:
+ ctables_summary_spec_set_uninit (sss);
+ return false;
+}
- for (size_t i = 0; i < ct->n_tables; i++)
- {
- struct ctables_table *t = ct->tables[i];
+static bool
+ctables_parse_pproperties (struct lexer *lexer, struct ctables *ct)
+{
+ struct ctables_postcompute **pcs = NULL;
+ size_t n_pcs = 0;
+ size_t allocated_pcs = 0;
- for (size_t ir = 0; ir < t->vaas[PIVOT_AXIS_ROW].n; ir++)
- for (size_t ic = 0; ic < t->vaas[PIVOT_AXIS_COLUMN].n; ic++)
- for (size_t il = 0; il < t->vaas[PIVOT_AXIS_LAYER].n; il++)
- ctables_cell_insert (t, c, ir, ic, il, weight);
+ while (lex_match (lexer, T_AND))
+ {
+ if (!lex_force_id (lexer))
+ goto error;
+ struct ctables_postcompute *pc
+ = ctables_find_postcompute (ct, lex_tokcstr (lexer));
+ if (!pc)
+ {
+ msg (SE, _("Unknown computed category &%s."), lex_tokcstr (lexer));
+ goto error;
}
+ lex_get (lexer);
+
+ if (n_pcs >= allocated_pcs)
+ pcs = x2nrealloc (pcs, &allocated_pcs, sizeof *pcs);
+ pcs[n_pcs++] = pc;
}
- casereader_destroy (input);
- for (size_t i = 0; i < ct->n_tables; i++)
+ while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
{
- struct ctables_table *t = ct->tables[i];
-
- struct pivot_table *pt = pivot_table_create__ (
- (t->title
- ? pivot_value_new_user_text (t->title, SIZE_MAX)
- : pivot_value_new_text (N_("Custom Tables"))),
- "Custom Tables");
- if (t->caption)
- pivot_table_set_caption (
- pt, pivot_value_new_user_text (t->caption, SIZE_MAX));
- if (t->corner)
- pivot_table_set_caption (
- pt, pivot_value_new_user_text (t->corner, SIZE_MAX));
-
- pivot_table_set_look (pt, ct->look);
- struct pivot_dimension *d[PIVOT_N_AXES];
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (lex_match_id (lexer, "LABEL"))
{
- static const char *names[] = {
- [PIVOT_AXIS_ROW] = N_("Rows"),
- [PIVOT_AXIS_COLUMN] = N_("Columns"),
- [PIVOT_AXIS_LAYER] = N_("Layers"),
- };
- d[a] = (t->axes[a] || a == t->summary_axis
- ? pivot_dimension_create (pt, a, names[a])
- : NULL);
- if (!d[a])
- continue;
-
- assert (t->axes[a]);
-
- struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
+ lex_match (lexer, T_EQUALS);
+ if (!lex_force_string (lexer))
+ goto error;
- struct ctables_cell *f;
- size_t n = 0;
- HMAP_FOR_EACH (f, struct ctables_cell, node, &t->cells)
- sorted[n++] = f;
- assert (n == t->cells.count);
+ for (size_t i = 0; i < n_pcs; i++)
+ {
+ free (pcs[i]->label);
+ pcs[i]->label = ss_xstrdup (lex_tokss (lexer));
+ }
- struct ctables_cell_sort_aux aux = { .t = t, .a = a };
- sort (sorted, n, sizeof *sorted, ctables_cell_compare_3way, &aux);
+ lex_get (lexer);
+ }
+ else if (lex_match_id (lexer, "FORMAT"))
+ {
+ lex_match (lexer, T_EQUALS);
- size_t max_depth = 0;
- for (size_t j = 0; j < t->vaas[a].n; j++)
- if (t->vaas[a].vas[j].n > max_depth)
- max_depth = t->vaas[a].vas[j].n;
+ struct ctables_summary_spec_set sss;
+ if (!ctables_parse_pproperties_format (lexer, &sss))
+ goto error;
- struct pivot_category **groups = xnmalloc (max_depth, sizeof *groups);
- struct pivot_category *top = NULL;
- int prev_leaf = 0;
- for (size_t j = 0; j < n; j++)
+ for (size_t i = 0; i < n_pcs; i++)
{
- struct ctables_cell *f = sorted[j];
- const struct var_array *va = &t->vaas[a].vas[f->axes[a].vaa_idx];
-
- size_t n_common = 0;
- bool new_subtable = false;
- if (j > 0)
- {
- struct ctables_cell *prev = sorted[j - 1];
- if (prev->axes[a].vaa_idx == f->axes[a].vaa_idx)
- {
- for (; n_common < va->n; n_common++)
- if (n_common != va->scale_idx
- && (prev->axes[a].cvs[n_common].category
- != f->axes[a].cvs[n_common].category
- || !value_equal (&prev->axes[a].cvs[n_common].value,
- &f->axes[a].cvs[n_common].value,
- var_get_type (va->vars[n_common]))))
- break;
- }
- else
- new_subtable = true;
- }
+ if (pcs[i]->specs)
+ ctables_summary_spec_set_uninit (pcs[i]->specs);
else
- new_subtable = true;
-
- if (new_subtable)
- {
- enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (va->vars[0])];
- top = d[a]->root;
- if (vlabel != CTVL_NONE)
- top = pivot_category_create_group__ (
- top, pivot_value_new_variable (va->vars[0]));
- }
- if (n_common == va->n)
- {
- f->axes[a].leaf = prev_leaf;
- continue;
- }
-
- for (size_t k = n_common; k < va->n; k++)
- {
- struct pivot_category *parent = k > 0 ? groups[k - 1] : top;
-
- struct pivot_value *label
- = (k == va->scale_idx ? NULL
- : f->axes[a].cvs[k].category->type == CCT_TOTAL
- ? pivot_value_new_user_text (f->axes[a].cvs[k].category->total_label,
- SIZE_MAX)
- : pivot_value_new_var_value (va->vars[k],
- &f->axes[a].cvs[k].value));
- if (k == va->n - 1)
- {
- if (a == t->summary_axis)
- {
- if (label)
- parent = pivot_category_create_group__ (parent, label);
- for (size_t m = 0; m < va->n_summaries; m++)
- {
- int leaf = pivot_category_create_leaf (
- parent, pivot_value_new_text (va->summaries[m].label));
- if (m == 0)
- prev_leaf = leaf;
- }
- }
- else
- {
- /* This assertion is true as long as the summary axis
- is the axis where the summaries are displayed. */
- assert (label);
-
- prev_leaf = pivot_category_create_leaf (parent, label);
- }
- break;
- }
-
- if (label)
- parent = pivot_category_create_group__ (parent, label);
-
- enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (va->vars[k + 1])];
- if (vlabel != CTVL_NONE)
- parent = pivot_category_create_group__ (
- parent, pivot_value_new_variable (va->vars[k + 1]));
- groups[k] = parent;
- }
-
- f->axes[a].leaf = prev_leaf;
+ pcs[i]->specs = xmalloc (sizeof *pcs[i]->specs);
+ ctables_summary_spec_set_clone (pcs[i]->specs, &sss);
}
- free (sorted);
- free (groups);
+ ctables_summary_spec_set_uninit (&sss);
}
- struct ctables_cell *f;
- HMAP_FOR_EACH (f, struct ctables_cell, node, &t->cells)
+ else if (lex_match_id (lexer, "HIDESOURCECATS"))
{
- const struct var_array *ss = &t->vaas[t->summary_axis].vas[f->axes[t->summary_axis].vaa_idx];
- for (size_t j = 0; j < ss->n_summaries; j++)
- {
- size_t dindexes[3];
- size_t n_dindexes = 0;
-
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- if (d[a])
- {
- int leaf = f->axes[a].leaf;
- if (a == t->summary_axis)
- leaf += j;
- dindexes[n_dindexes++] = leaf;
- }
-
- double d = ctables_summary_value (f, &f->summaries[j], &ss->summaries[j]);
- struct pivot_value *value = pivot_value_new_number (d);
- value->numeric.format = ss->summaries[j].format;
- pivot_table_put (pt, dindexes, n_dindexes, value);
- }
+ lex_match (lexer, T_EQUALS);
+ bool hide_source_cats;
+ if (!parse_bool (lexer, &hide_source_cats))
+ goto error;
+ for (size_t i = 0; i < n_pcs; i++)
+ pcs[i]->hide_source_cats = hide_source_cats;
+ }
+ else
+ {
+ lex_error_expecting (lexer, "LABEL", "FORMAT", "HIDESOURCECATS");
+ goto error;
}
-
- pivot_table_submit (pt);
}
+ free (pcs);
+ return true;
- return proc_commit (ds);
+error:
+ free (pcs);
+ return false;
}
int
struct ctables *ct = xmalloc (sizeof *ct);
*ct = (struct ctables) {
+ .dict = dataset_dict (ds),
.look = pivot_table_look_unshare (pivot_table_look_ref (
pivot_table_look_get_default ())),
.vlabels = vlabels,
+ .postcomputes = HMAP_INITIALIZER (ct->postcomputes),
.hide_threshold = 5,
};
ct->look->omit_empty = false;
goto error;
}
}
- /* XXX PCOMPUTE */
+ else if (lex_match_id (lexer, "PCOMPUTE"))
+ {
+ if (!ctables_parse_pcompute (lexer, ct))
+ goto error;
+ }
+ else if (lex_match_id (lexer, "PPROPERTIES"))
+ {
+ if (!ctables_parse_pproperties (lexer, ct))
+ goto error;
+ }
else if (lex_match_id (lexer, "WEIGHT"))
{
if (!lex_force_match_id (lexer, "VARIABLE"))
goto error;
lex_match (lexer, T_EQUALS);
- ct->base_weight = parse_variable (lexer, dataset_dict (ds));
- if (!ct->base_weight)
+ ct->e_weight = parse_variable (lexer, dataset_dict (ds));
+ if (!ct->e_weight)
goto error;
}
else if (lex_match_id (lexer, "HIDESMALLCOUNTS"))
.n_refs = n_vars,
.cats = cat,
.n_cats = 1,
+ .show_empty = true,
};
struct ctables_categories **categories = xnmalloc (n_vars,
struct ctables_table *t = xmalloc (sizeof *t);
*t = (struct ctables_table) {
- .cells = HMAP_INITIALIZER (t->cells),
- .slabels_position = PIVOT_AXIS_COLUMN,
+ .ctables = ct,
+ .slabels_axis = PIVOT_AXIS_COLUMN,
.slabels_visible = true,
- .row_labels = CTLP_NORMAL,
- .col_labels = CTLP_NORMAL,
+ .clabels_values_map = HMAP_INITIALIZER (t->clabels_values_map),
+ .label_axis = {
+ [PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW,
+ [PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN,
+ [PIVOT_AXIS_LAYER] = PIVOT_AXIS_LAYER,
+ },
+ .clabels_from_axis = PIVOT_AXIS_LAYER,
.categories = categories,
.n_categories = n_vars,
.cilevel = 95,
};
- for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
- hmap_init (&t->domains[dt]);
ct->tables[ct->n_tables++] = t;
lex_match (lexer, T_EQUALS);
}
if (lex_token (lexer) == T_ENDCMD)
- break;
+ {
+ if (!ctables_prepare_table (t))
+ goto error;
+ break;
+ }
if (!lex_force_match (lexer, T_SLASH))
break;
{
if (lex_match_id (lexer, "SLABELS"))
{
- while (lex_token (lexer) != T_SLASH)
+ while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
{
if (lex_match_id (lexer, "POSITION"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "COLUMN"))
- t->slabels_position = PIVOT_AXIS_COLUMN;
+ t->slabels_axis = PIVOT_AXIS_COLUMN;
else if (lex_match_id (lexer, "ROW"))
- t->slabels_position = PIVOT_AXIS_ROW;
+ t->slabels_axis = PIVOT_AXIS_ROW;
else if (lex_match_id (lexer, "LAYER"))
- t->slabels_position = PIVOT_AXIS_LAYER;
+ t->slabels_axis = PIVOT_AXIS_LAYER;
else
{
lex_error_expecting (lexer, "COLUMN", "ROW", "LAYER");
}
else if (lex_match_id (lexer, "CLABELS"))
{
- while (lex_token (lexer) != T_SLASH)
+ while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
{
if (lex_match_id (lexer, "AUTO"))
- t->row_labels = t->col_labels = CTLP_NORMAL;
+ {
+ t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW;
+ t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN;
+ }
else if (lex_match_id (lexer, "ROWLABELS"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "OPPOSITE"))
- t->row_labels = CTLP_OPPOSITE;
+ t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_COLUMN;
else if (lex_match_id (lexer, "LAYER"))
- t->row_labels = CTLP_LAYER;
+ t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_LAYER;
else
{
lex_error_expecting (lexer, "OPPOSITE", "LAYER");
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "OPPOSITE"))
- t->col_labels = CTLP_OPPOSITE;
+ t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_ROW;
else if (lex_match_id (lexer, "LAYER"))
- t->col_labels = CTLP_LAYER;
+ t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_LAYER;
else
{
lex_error_expecting (lexer, "OPPOSITE", "LAYER");
"SIGTEST", "COMPARETEST");
goto error;
}
+
+ if (!lex_match (lexer, T_SLASH))
+ break;
}
- if (t->row_labels != CTLP_NORMAL && t->col_labels != CTLP_NORMAL)
+ if (t->label_axis[PIVOT_AXIS_ROW] != PIVOT_AXIS_ROW
+ && t->label_axis[PIVOT_AXIS_COLUMN] != PIVOT_AXIS_COLUMN)
{
msg (SE, _("ROWLABELS and COLLABELS may not both be specified."));
goto error;
}
+ if (!ctables_prepare_table (t))
+ goto error;
}
while (lex_token (lexer) != T_ENDCMD);