#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"
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 contain this cell. */
bool hide;
enum ctables_summary_variant sv;
- struct
+ struct ctables_cell_axis
{
struct ctables_cell_value
{
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 *e_weight; /* WEIGHT. */
size_t n_tables;
};
+static struct ctables_postcompute *ctables_find_postcompute (struct ctables *,
+ const char *name);
+
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. */
+ struct msg_location *location;
};
+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
{
struct ctables_summary_spec *specs;
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 ctables_stack stacks[PIVOT_N_AXES];
struct ctables_section *sections;
{
enum ctables_category_type
{
+ /* Explicit category lists. */
CCT_NUMBER,
CCT_STRING,
CCT_RANGE,
CCT_MISSING,
CCT_OTHERNM,
+ CCT_POSTCOMPUTE,
+ /* Totals and subtotals. */
CCT_SUBTOTAL,
- CCT_HSUBTOTAL,
CCT_TOTAL,
+ /* Implicit category lists. */
CCT_VALUE,
CCT_LABEL,
CCT_FUNCTION,
struct ctables_category *subtotal;
+ bool hide;
+
union
{
double number; /* CCT_NUMBER. */
char *string; /* CCT_STRING. */
double range[2]; /* CCT_RANGE. */
- char *total_label; /* CCT_SUBTOTAL, CCT_HSUBTOTAL, CCT_TOTAL. */
+
+ struct
+ {
+ char *total_label; /* CCT_SUBTOTAL, CCT_TOTAL. */
+ bool hide_subcategories; /* CCT_SUBTOTAL. */
+ };
+
+ const struct ctables_postcompute *pc; /* CCT_POSTCOMPUTE. */
/* CCT_VALUE, CCT_LABEL, CCT_FUNCTION. */
struct
double percentile;
};
};
+
+ /* Source location. This is null for CCT_TOTAL, CCT_VALUE, CCT_LABEL,
+ CCT_FUNCTION. */
+ struct msg_location *location;
};
static void
case CCT_RANGE:
case CCT_MISSING:
case CCT_OTHERNM:
+ case CCT_POSTCOMPUTE:
break;
case CCT_STRING:
break;
case CCT_SUBTOTAL:
- case CCT_HSUBTOTAL:
case CCT_TOTAL:
free (cat->total_label);
break;
case CCT_OTHERNM:
return true;
+ case CCT_POSTCOMPUTE:
+ return a->pc == b->pc;
+
case CCT_SUBTOTAL:
- case CCT_HSUBTOTAL:
case CCT_TOTAL:
return !strcmp (a->total_label, b->total_label);
};
}
+static bool
+ctables_table_parse_subtotal (struct lexer *lexer, bool hide_subcategories,
+ struct ctables_category *cat)
+{
+ char *total_label;
+ if (lex_match (lexer, T_EQUALS))
+ {
+ if (!lex_force_string (lexer))
+ return false;
+
+ total_label = ss_xstrdup (lex_tokss (lexer));
+ lex_get (lexer);
+ }
+ else
+ total_label = xstrdup (_("Subtotal"));
+
+ *cat = (struct ctables_category) {
+ .type = CCT_SUBTOTAL,
+ .hide_subcategories = hide_subcategories,
+ .total_label = total_label
+ };
+ return true;
+}
+
+static bool
+ctables_table_parse_explicit_category (struct lexer *lexer, struct ctables *ct,
+ struct ctables_category *cat)
+{
+ if (lex_match_id (lexer, "OTHERNM"))
+ *cat = (struct ctables_category) { .type = CCT_OTHERNM };
+ else if (lex_match_id (lexer, "MISSING"))
+ *cat = (struct ctables_category) { .type = CCT_MISSING };
+ else if (lex_match_id (lexer, "SUBTOTAL"))
+ return ctables_table_parse_subtotal (lexer, false, cat);
+ else if (lex_match_id (lexer, "HSUBTOTAL"))
+ return ctables_table_parse_subtotal (lexer, true, cat);
+ else if (lex_match_id (lexer, "LO"))
+ {
+ if (!lex_force_match_id (lexer, "THRU") || lex_force_num (lexer))
+ return false;
+ *cat = cct_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"))
+ *cat = cct_range (number, DBL_MAX);
+ else
+ {
+ if (!lex_force_num (lexer))
+ return false;
+ *cat = cct_range (number, lex_number (lexer));
+ lex_get (lexer);
+ }
+ }
+ else
+ *cat = (struct ctables_category) {
+ .type = CCT_NUMBER,
+ .number = number
+ };
+ }
+ else if (lex_is_string (lexer))
+ {
+ *cat = (struct ctables_category) {
+ .type = CCT_STRING,
+ .string = ss_xstrdup (lex_tokss (lexer)),
+ };
+ lex_get (lexer);
+ }
+ else if (lex_match (lexer, T_AND))
+ {
+ if (!lex_force_id (lexer))
+ return false;
+ struct ctables_postcompute *pc = ctables_find_postcompute (
+ ct, lex_tokcstr (lexer));
+ if (!pc)
+ {
+ struct msg_location *loc = lex_get_location (lexer, -1, 0);
+ msg_at (SE, loc, _("Unknown postcompute &%s."),
+ lex_tokcstr (lexer));
+ msg_location_destroy (loc);
+ return false;
+ }
+ lex_get (lexer);
+
+ *cat = (struct ctables_category) { .type = CCT_POSTCOMPUTE, .pc = pc };
+ }
+ else
+ {
+ lex_error (lexer, NULL);
+ return false;
+ }
+
+ return true;
+}
+
+static struct ctables_category *
+ctables_find_category_for_postcompute (const struct ctables_categories *cats,
+ const struct ctables_pcexpr *e)
+{
+ struct ctables_category *best = NULL;
+ size_t n_subtotals = 0;
+ for (size_t i = 0; i < cats->n_cats; i++)
+ {
+ struct ctables_category *cat = &cats->cats[i];
+ switch (e->op)
+ {
+ case CTPO_CAT_NUMBER:
+ if (cat->type == CCT_NUMBER && cat->number == e->number)
+ best = cat;
+ break;
+
+ case CTPO_CAT_STRING:
+ if (cat->type == CCT_STRING && !strcmp (cat->string, e->string))
+ best = cat;
+ break;
+
+ case CTPO_CAT_RANGE:
+ if (cat->type == CCT_RANGE
+ && cat->range[0] == e->range[0]
+ && cat->range[1] == e->range[1])
+ best = cat;
+ break;
+
+ case CTPO_CAT_MISSING:
+ if (cat->type == CCT_MISSING)
+ best = cat;
+ break;
+
+ case CTPO_CAT_OTHERNM:
+ if (cat->type == CCT_OTHERNM)
+ best = cat;
+ break;
+
+ case CTPO_CAT_SUBTOTAL:
+ if (cat->type == CCT_SUBTOTAL)
+ {
+ n_subtotals++;
+ if (e->subtotal_index == n_subtotals)
+ return cat;
+ else if (e->subtotal_index == 0)
+ best = cat;
+ }
+ break;
+
+ case CTPO_CAT_TOTAL:
+ if (cat->type == CCT_TOTAL)
+ return cat;
+ break;
+
+ case CTPO_CONSTANT:
+ case CTPO_ADD:
+ case CTPO_SUB:
+ case CTPO_MUL:
+ case CTPO_DIV:
+ case CTPO_POW:
+ case CTPO_NEG:
+ NOT_REACHED ();
+ }
+ }
+ if (e->op == CTPO_CAT_SUBTOTAL && e->subtotal_index == 0 && n_subtotals > 1)
+ return NULL;
+ return best;
+}
+
+static bool
+ctables_recursive_check_postcompute (const struct ctables_pcexpr *e,
+ struct ctables_category *pc_cat,
+ const struct ctables_categories *cats,
+ const struct msg_location *cats_location)
+{
+ switch (e->op)
+ {
+ case CTPO_CAT_NUMBER:
+ case CTPO_CAT_STRING:
+ case CTPO_CAT_RANGE:
+ case CTPO_CAT_MISSING:
+ case CTPO_CAT_OTHERNM:
+ case CTPO_CAT_SUBTOTAL:
+ case CTPO_CAT_TOTAL:
+ {
+ struct ctables_category *cat = ctables_find_category_for_postcompute (
+ cats, e);
+ if (!cat)
+ {
+ if (e->op == CTPO_CAT_SUBTOTAL && e->subtotal_index == 0)
+ {
+ size_t n_subtotals = 0;
+ for (size_t i = 0; i < cats->n_cats; i++)
+ n_subtotals += cats->cats[i].type == CCT_SUBTOTAL;
+ if (n_subtotals > 1)
+ {
+ msg_at (SE, cats_location,
+ ngettext ("These categories include %zu instance "
+ "of SUBTOTAL or HSUBTOTAL, so references "
+ "from computed categories must refer to "
+ "subtotals by position.",
+ "These categories include %zu instances "
+ "of SUBTOTAL or HSUBTOTAL, so references "
+ "from computed categories must refer to "
+ "subtotals by position.",
+ n_subtotals),
+ n_subtotals);
+ msg_at (SN, e->location,
+ _("This is the reference that lacks a position."));
+ return NULL;
+ }
+ }
+
+ msg_at (SE, pc_cat->location,
+ _("Computed category &%s references a category not included "
+ "in the category list."),
+ cat->pc->name);
+ msg_at (SN, e->location, _("This is the missing category."));
+ msg_at (SN, cats_location,
+ _("To fix the problem, add the missing category to the "
+ "list of categories here."));
+ return false;
+ }
+ if (pc_cat->pc->hide_source_cats)
+ cat->hide = true;
+ return true;
+ }
+
+ case CTPO_CONSTANT:
+ return true;
+
+ 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++)
+ if (e->subs[i] && !ctables_recursive_check_postcompute (
+ e->subs[i], pc_cat, cats, cats_location))
+ return false;
+ return true;
+
+ default:
+ NOT_REACHED ();
+ }
+}
+
static bool
ctables_table_parse_categories (struct lexer *lexer, struct dictionary *dict,
- struct ctables_table *t)
+ struct ctables *ct, struct ctables_table *t)
{
if (!lex_match_id (lexer, "VARIABLES"))
return false;
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
size_t allocated_cats = 0;
if (lex_match (lexer, T_LBRACK))
{
+ int cats_start_ofs = lex_ofs (lexer);
do
{
if (c->n_cats >= allocated_cats)
c->cats = x2nrealloc (c->cats, &allocated_cats, sizeof *c->cats);
+ int start_ofs = lex_ofs (lexer);
struct ctables_category *cat = &c->cats[c->n_cats];
- if (lex_match_id (lexer, "OTHERNM"))
- cat->type = CCT_OTHERNM;
- else if (lex_match_id (lexer, "MISSING"))
- cat->type = CCT_MISSING;
- else if (lex_match_id (lexer, "SUBTOTAL"))
- *cat = (struct ctables_category)
- { .type = CCT_SUBTOTAL, .total_label = NULL };
- else if (lex_match_id (lexer, "HSUBTOTAL"))
- *cat = (struct ctables_category)
- { .type = CCT_HSUBTOTAL, .total_label = NULL };
- else if (lex_match_id (lexer, "LO"))
- {
- if (!lex_force_match_id (lexer, "THRU") || lex_force_num (lexer))
- return false;
- *cat = cct_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"))
- {
- cat->type = CCT_RANGE;
- cat->range[0] = number;
- if (lex_match_id (lexer, "HI"))
- *cat = cct_range (number, DBL_MAX);
- else
- {
- if (!lex_force_num (lexer))
- return false;
- *cat = cct_range (number, lex_number (lexer));
- lex_get (lexer);
- }
- }
- else
- *cat = (struct ctables_category) {
- .type = CCT_NUMBER,
- .number = number
- };
- }
- else if (lex_is_string (lexer))
- {
- *cat = (struct ctables_category) {
- .type = CCT_STRING,
- .string = ss_xstrdup (lex_tokss (lexer)),
- };
- lex_get (lexer);
- }
- else
- {
- lex_error (lexer, NULL);
- return false;
- }
-
- if (cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
- {
- if (lex_match (lexer, T_EQUALS))
- {
- if (!lex_force_string (lexer))
- return false;
-
- cat->total_label = ss_xstrdup (lex_tokss (lexer));
- lex_get (lexer);
- }
- else
- cat->total_label = xstrdup (_("Subtotal"));
- }
-
+ if (!ctables_table_parse_explicit_category (lexer, ct, cat))
+ return false;
+ cat->location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1);
c->n_cats++;
+
lex_match (lexer, T_COMMA);
}
while (!lex_match (lexer, T_RBRACK));
+
+ struct msg_location *cats_location
+ = lex_ofs_location (lexer, cats_start_ofs, lex_ofs (lexer) - 1);
+ for (size_t i = 0; i < c->n_cats; i++)
+ {
+ struct ctables_category *cat = &c->cats[i];
+ if (cat->type == CCT_POSTCOMPUTE
+ && !ctables_recursive_check_postcompute (cat->pc->expr, cat,
+ c, cats_location))
+ return false;
+ }
}
struct ctables_category cat = {
if (!c->n_cats)
{
if (c->n_cats >= allocated_cats)
- c->cats = x2nrealloc (c->cats, &allocated_cats,
- sizeof *c->cats);
+ c->cats = x2nrealloc (c->cats, &allocated_cats, sizeof *c->cats);
c->cats[c->n_cats++] = cat;
}
cat->subtotal = subtotal;
break;
+ case CCT_POSTCOMPUTE:
+ break;
+
case CCT_SUBTOTAL:
- case CCT_HSUBTOTAL:
subtotal = cat;
break;
case CCT_NUMBER:
case CCT_STRING:
case CCT_SUBTOTAL:
- case CCT_HSUBTOTAL:
case CCT_TOTAL:
+ case CCT_POSTCOMPUTE:
/* Must be equal. */
continue;
return cat;
break;
+ case CCT_POSTCOMPUTE:
+ break;
+
case CCT_OTHERNM:
if (!othernm)
othernm = cat;
break;
case CCT_SUBTOTAL:
- case CCT_HSUBTOTAL:
case CCT_TOTAL:
break;
{
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)
+ && cats[a][i]->type != CCT_SUBTOTAL)
hash = value_hash (case_data (c, nest->vars[i]),
var_get_width (nest->vars[i]), hash);
else
&& (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])))))
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)
+ if (cat->hide || (subtotal && subtotal->hide_subcategories))
cell->hide = true;
- if (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
+ if (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL)
cell->contributes_to_domains = false;
}
}
}
+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];
+ 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 (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);
const struct variable *var,
const union value *value)
{
- return (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL
+ return (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL
? pivot_value_new_user_text (cat->total_label, SIZE_MAX)
: pivot_value_new_var_value (var, value));
}
.cells = HMAP_INITIALIZER (s->cells),
};
for (a = 0; a < PIVOT_N_AXES; a++)
- s->nests[a] = t->stacks[a].n ? &t->stacks[a].nests[ix[a]] : NULL;
+ 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]);
}
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,
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_POSTCOMPUTE:
+ break;
+
+ case CCT_SUBTOTAL:
+ 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)
{
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 },
+ .location = msg_location_merged (sub0->location, sub1->location),
+ };
+ return e;
+}
+
+/* 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, lhs->location, "%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.location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1);
+ return xmemdup (&e, sizeof e);
+}
+
+static struct ctables_pcexpr *
+ctables_pcexpr_allocate_neg (struct ctables_pcexpr *sub,
+ struct lexer *lexer, int start_ofs)
+{
+ struct ctables_pcexpr *e = xmalloc (sizeof *e);
+ *e = (struct ctables_pcexpr) {
+ .op = CTPO_NEG,
+ .subs = { sub },
+ .location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer) - 1),
+ };
+ return e;
+}
+
+static struct ctables_pcexpr *
+parse_exp (struct lexer *lexer)
+{
+ 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),
+ .location = lex_ofs_location (lexer, start_ofs, lex_ofs (lexer)),
+ };
+ lex_get (lexer);
+
+ struct ctables_pcexpr *node = parse_binary_operators__ (
+ lexer, &op, 1, parse_primary, chain_warning, lhs);
+ if (!node)
+ return NULL;
+
+ return ctables_pcexpr_allocate_neg (node, lexer, start_ofs);
+}
+
+/* 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);
+
+ struct ctables_pcexpr *inner = parse_neg (lexer);
+ if (!inner)
+ return NULL;
+
+ return ctables_pcexpr_allocate_neg (inner, lexer, start_ofs);
+}
+
+/* Parses the multiplication and division level. */
+static struct ctables_pcexpr *
+parse_mul (struct lexer *lexer)
+{
+ static const struct operator ops[] =
+ {
+ { T_ASTERISK, CTPO_MUL },
+ { T_SLASH, CTPO_DIV },
+ };
+
+ return parse_binary_operators (lexer, ops, sizeof ops / sizeof *ops,
+ parse_neg, NULL);
+}
+
+/* Parses the addition and subtraction level. */
+static struct ctables_pcexpr *
+parse_add (struct lexer *lexer)
+{
+ static const struct operator ops[] =
+ {
+ { T_PLUS, CTPO_ADD },
+ { T_DASH, CTPO_SUB },
+ { T_NEG_NUM, CTPO_ADD },
+ };
+
+ return parse_binary_operators (lexer, ops, sizeof ops / sizeof *ops,
+ parse_mul, NULL);
+}
+
+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_parse_pcompute (struct lexer *lexer, struct ctables *ct)
+{
+ int start_ofs = lex_ofs (lexer) - 1;
+
+ if (!lex_force_match (lexer, T_AND) || !lex_force_id (lexer))
+ return false;
+
+ char *name = ss_xstrdup (lex_tokss (lexer));
+
+ 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;
+ }
+
+ struct ctables_pcexpr *expr = parse_add (lexer);
+ if (!expr || !lex_force_match (lexer, T_RPAREN))
+ {
+ free (name);
+ return false;
+ }
+
+ struct msg_location *location = lex_ofs_location (lexer, start_ofs,
+ lex_ofs (lexer) - 1);
+
+ 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;
+}
+
+static bool
+ctables_parse_pproperties_format (struct lexer *lexer,
+ struct ctables_summary_spec_set *sss)
+{
+ *sss = (struct ctables_summary_spec_set) { .n = 0 };
+
+ 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;
+
+ /* Parse percentile. */
+ double percentile = 0;
+ if (function == CTSF_PTILE)
+ {
+ if (!lex_force_num_range_closed (lexer, "PTILE", 0, 100))
+ goto error;
+ percentile = lex_number (lexer);
+ lex_get (lexer);
+ }
+
+ /* 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;
+
+ 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;
+
+error:
+ ctables_summary_spec_set_uninit (sss);
+ return false;
+}
+
+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;
+
+ 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;
+ }
+
+ while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
+ {
+ if (lex_match_id (lexer, "LABEL"))
+ {
+ lex_match (lexer, T_EQUALS);
+ if (!lex_force_string (lexer))
+ goto error;
+
+ for (size_t i = 0; i < n_pcs; i++)
+ {
+ free (pcs[i]->label);
+ pcs[i]->label = ss_xstrdup (lex_tokss (lexer));
+ }
+
+ lex_get (lexer);
+ }
+ else if (lex_match_id (lexer, "FORMAT"))
+ {
+ lex_match (lexer, T_EQUALS);
+
+ struct ctables_summary_spec_set sss;
+ if (!ctables_parse_pproperties_format (lexer, &sss))
+ goto error;
+
+ for (size_t i = 0; i < n_pcs; i++)
+ {
+ if (pcs[i]->specs)
+ ctables_summary_spec_set_uninit (pcs[i]->specs);
+ else
+ pcs[i]->specs = xmalloc (sizeof *pcs[i]->specs);
+ ctables_summary_spec_set_clone (pcs[i]->specs, &sss);
+ }
+ ctables_summary_spec_set_uninit (&sss);
+ }
+ else if (lex_match_id (lexer, "HIDESOURCECATS"))
+ {
+ 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;
+ }
+ }
+ free (pcs);
+ return true;
+
+error:
+ free (pcs);
+ return false;
+}
int
cmd_ctables (struct lexer *lexer, struct dataset *ds)
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"))
.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) {
+ .ctables = ct,
.slabels_axis = PIVOT_AXIS_COLUMN,
.slabels_visible = true,
.clabels_values_map = HMAP_INITIALIZER (t->clabels_values_map),
}
else if (lex_match_id (lexer, "CATEGORIES"))
{
- if (!ctables_table_parse_categories (lexer, dataset_dict (ds), t))
+ if (!ctables_table_parse_categories (lexer, dataset_dict (ds),
+ ct, t))
goto error;
}
else if (lex_match_id (lexer, "TITLES"))
projects / pspp / blobdiff