double 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
struct ctables_domain *domains[N_CTDTS];
bool hide;
- bool total;
+ enum ctables_summary_variant sv;
struct
{
struct variable *var;
};
+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. */
size_t *domains[N_CTDTS];
size_t n_domains[N_CTDTS];
- struct ctables_summary_spec_set cell_sss;
- struct ctables_summary_spec_set total_sss;
+ struct ctables_summary_spec_set specs[N_CSVS];
};
/* A stack of nestings, e.g. nest1 + nest2 + ... + nestN. */
{
struct ctables_var var;
bool scale;
- struct ctables_summary_spec_set cell_sss;
- struct ctables_summary_spec_set total_sss;
+ 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)
{
switch (axis->op)
{
case CTAO_VAR:
- ctables_summary_spec_set_uninit (&axis->cell_sss);
- ctables_summary_spec_set_uninit (&axis->total_sss);
+ 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, bool totals)
+ const struct msg_location *loc, enum ctables_summary_variant sv)
{
if (axis->op == CTAO_VAR)
{
break;
}
- struct ctables_summary_spec_set *set = (totals ? &axis->total_sss
- : &axis->cell_sss);
+ struct ctables_summary_spec_set *set = &axis->specs[sv];
if (set->n >= set->allocated)
set->specs = x2nrealloc (set->specs, &set->allocated,
sizeof *set->specs);
{
for (size_t i = 0; i < 2; i++)
if (!add_summary_spec (axis->subs[i], function, percentile, label,
- format, loc, totals))
+ 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;
- bool totals = false;
+ 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,
- totals);
+ printf ("add %s\n", ctables_summary_function_name (function));
+ add_summary_spec (sub, function, percentile, label, formatp, loc, sv);
free (label);
msg_location_destroy (loc);
- if (lex_match (ctx->lexer, T_COMMA))
+ lex_match (ctx->lexer, T_COMMA);
+ if (sv == CSV_CELL && lex_match_id (ctx->lexer, "TOTALS"))
{
- if (!totals && lex_match_id (ctx->lexer, "TOTALS"))
- {
- if (!lex_force_match (ctx->lexer, T_LBRACK))
- goto error;
- totals = true;
- }
+ if (!lex_force_match (ctx->lexer, T_LBRACK))
+ goto error;
+ sv = CSV_TOTAL;
}
- else if (lex_force_match (ctx->lexer, T_RBRACK))
+ else if (lex_match (ctx->lexer, T_RBRACK))
{
- if (totals && !lex_force_match (ctx->lexer, T_RBRACK))
+ if (sv == CSV_TOTAL && !lex_force_match (ctx->lexer, T_RBRACK))
goto error;
return sub;
}
- else
- goto error;
}
error:
if (!axis)
return NULL;
else if (axis->op == CTAO_VAR)
- return !axis->scale && axis->cell_sss.n ? axis : NULL;
+ return !axis->scale && axis->specs[CSV_CELL].n ? axis : NULL;
else
{
for (size_t i = 0; i < 2; i++)
assert (n == allocate);
const struct ctables_nest *summary_src;
- if (!a->cell_sss.var)
+ if (!a->specs[CSV_CELL].var)
summary_src = b;
- else if (!b->cell_sss.var)
+ else if (!b->specs[CSV_CELL].var)
summary_src = a;
else
NOT_REACHED ();
- stack.nests[stack.n++] = (struct ctables_nest) {
+
+ 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,
- .cell_sss = summary_src->cell_sss,
- .total_sss = summary_src->total_sss,
};
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ ctables_summary_spec_set_clone (&new->specs[sv], &summary_src->specs[sv]);
}
ctables_stack_uninit (&s0);
ctables_stack_uninit (&s1);
.n = 1,
.scale_idx = a->scale ? 0 : SIZE_MAX,
};
- if (a->cell_sss.n || a->scale)
- {
- nest->cell_sss = a->cell_sss;
- nest->total_sss = a->total_sss;
- nest->cell_sss.var = a->var.var;
- nest->total_sss.var = a->var.var;
- }
+ 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:
const struct ctables_nest *ss = &t->stacks[t->summary_axis].nests[ix[t->summary_axis]];
size_t hash = 0;
- bool total = false;
+ enum ctables_summary_variant sv = CSV_CELL;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
hash = value_hash (case_data (c, nest->vars[i]),
var_get_width (nest->vars[i]), hash);
else
- total = true;
+ sv = CSV_TOTAL;
}
}
cell = xmalloc (sizeof *cell);
cell->hide = false;
- cell->total = total;
+ cell->sv = sv;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
}
}
- {
- const struct ctables_summary_spec_set *sss
- = (cell->total ? &ss->total_sss : &ss->cell_sss);
- cell->summaries = xmalloc (sss->n * sizeof *cell->summaries);
- for (size_t i = 0; i < sss->n; i++)
- ctables_summary_init (&cell->summaries[i], &sss->specs[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 (t, cell, dt);
hmap_insert (&t->cells, &cell->node, hash);
struct ctables_cell *cell = ctables_cell_insert__ (t, c, ix, cats);
const struct ctables_nest *ss = &t->stacks[t->summary_axis].nests[ix[t->summary_axis]];
- const struct ctables_summary_spec_set *sss
- = (cell->total ? &ss->total_sss : &ss->cell_sss);
- for (size_t i = 0; i < sss->n; i++)
- ctables_summary_add (&cell->summaries[i], &sss->specs[i], sss->var,
- case_data (c, sss->var), weight);
+ 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), weight);
for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
cell->domains[dt]->valid += weight;
}
}
}
+struct merge_item
+ {
+ size_t tiebreaker;
+ 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 bool
ctables_execute (struct dataset *ds, struct ctables *ct)
{
t->stacks[a] = (struct ctables_stack) { .nests = nest, .n = 1 };
}
- for (size_t i = 0; i < t->stacks[t->summary_axis].n; i++)
+ struct ctables_stack *stack = &t->stacks[t->summary_axis];
+ for (size_t i = 0; i < stack->n; i++)
{
- struct ctables_nest *nest = &t->stacks[t->summary_axis].nests[i];
- if (!nest->cell_sss.n)
+ struct ctables_nest *nest = &stack->nests[i];
+ if (!nest->specs[CSV_CELL].n)
{
- struct ctables_summary_spec_set *sss = &nest->cell_sss;
- sss->specs = xmalloc (sizeof *sss->specs);
- sss->n = 1;
+ struct ctables_summary_spec_set *specs = &nest->specs[CSV_CELL];
+ specs->specs = xmalloc (sizeof *specs->specs);
+ specs->n = 1;
enum ctables_summary_function function
- = sss->var ? CTSF_MEAN : CTSF_COUNT;
- struct ctables_var var = { .is_mrset = false, .var = sss->var };
+ = specs->var ? CTSF_MEAN : CTSF_COUNT;
+ struct ctables_var var = { .is_mrset = false, .var = specs->var };
- *sss->specs = (struct ctables_summary_spec) {
+ *specs->specs = (struct ctables_summary_spec) {
.function = function,
.format = ctables_summary_default_format (function, &var),
.label = ctables_summary_default_label (function, 0),
};
- if (!sss->var)
- sss->var = nest->vars[0];
+ if (!specs->var)
+ specs->var = nest->vars[0];
- nest->total_sss = nest->cell_sss;
+ 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 = { .n = 0 };
+ 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)
+ continue;
+
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ {
+ items[n_left] = (struct merge_item) {
+ .tiebreaker = n_left,
+ .set = &nest->specs[sv]
+ };
+ n_left++;
+ }
+ }
+
+ 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];
+
+ /* XXX Add to 'merged' */
+ 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++;
+ }
+ }
+
+ 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");
}
- else if (!nest->total_sss.n)
- nest->total_sss = nest->cell_sss;
}
}
{
if (label)
parent = pivot_category_create_group__ (parent, label);
- const struct ctables_summary_spec_set *sss
- = cell->total ? &nest->total_sss : &nest->cell_sss;
- for (size_t m = 0; m < sss->n; m++)
+ const struct ctables_summary_spec_set *specs = &nest->specs[cell->sv];
+ for (size_t m = 0; m < specs->n; m++)
{
int leaf = pivot_category_create_leaf (
- parent, pivot_value_new_text (sss->specs[m].label));
+ parent, pivot_value_new_text (specs->specs[m].label));
if (m == 0)
prev_leaf = leaf;
}
continue;
const struct ctables_nest *nest = &t->stacks[t->summary_axis].nests[cell->axes[t->summary_axis].stack_idx];
- const struct ctables_summary_spec_set *sss = cell->total ? &nest->total_sss : &nest->cell_sss;
- for (size_t j = 0; j < sss->n; j++)
+ const struct ctables_summary_spec_set *specs = &nest->specs[cell->sv];
+ for (size_t j = 0; j < specs->n; j++)
{
size_t dindexes[3];
size_t n_dindexes = 0;
dindexes[n_dindexes++] = leaf;
}
- double d = ctables_summary_value (cell, &cell->summaries[j], &sss->specs[j]);
+ double d = ctables_summary_value (cell, &cell->summaries[j], &specs->specs[j]);
struct pivot_value *value = pivot_value_new_number (d);
- value->numeric.format = sss->specs[j].format;
+ value->numeric.format = specs->specs[j].format;
pivot_table_put (pt, dindexes, n_dindexes, value);
}
}
projects / pspp / blobdiff