#include "data/case.h"
#include "data/casegrouper.h"
#include "data/casereader.h"
+#include "data/casewriter.h"
#include "data/dictionary.h"
#include "data/format.h"
#include "data/procedure.h"
#include "data/settings.h"
+#include "data/subcase.h"
#include "data/value-labels.h"
#include "data/variable.h"
#include "language/command.h"
#include "libpspp/str.h"
#include "math/histogram.h"
#include "math/moments.h"
+#include "math/sort.h"
#include "output/chart-item.h"
#include "output/charts/piechart.h"
#include "output/charts/plot-hist.h"
+#include "output/pivot-table.h"
#include "output/tab.h"
#include "gl/minmax.h"
{
double p; /* the %ile to be calculated */
double value; /* the %ile's value */
- double x1; /* The datum value <= the percentile */
- double x2; /* The datum value >= the percentile */
- int flag;
- int flag2; /* Set to 1 if this percentile value has been found */
bool show; /* True to show this percentile in the statistics box. */
};
| BI(FRQ_SKEW) | BI(FRQ_SESKEW) | BI(FRQ_RANGE) \
| BI(FRQ_RANGE) | BI(FRQ_MODE) | BI(FRQ_MEDIAN))
-/* Statistics; number of statistics. */
-static unsigned long stats;
-static int n_stats;
-
struct frq_chart
{
double x_min; /* X axis minimum value. */
bool include_missing; /* Whether to include missing values. */
};
-/* Histogram and pie chart settings. */
-static struct frq_chart hist, pie;
-
-/* Parsed command. */
-static struct cmd_frequencies cmd;
-
/* Frequency tables. */
/* Entire frequency table. */
struct freq_tab
{
- struct hmap data; /* Hash table for accumulating counts. */
+ struct casewriter *sorter;
+ struct casereader *data;
+
struct freq *valid; /* Valid freqs. */
- int n_valid; /* Number of total freqs. */
+ int n_valid; /* Number of valid freqs. */
const struct dictionary *dict; /* Source of entries in the table. */
struct freq *missing; /* Missing freqs. */
struct var_freqs *vars;
size_t n_vars;
+ /* Percentiles to calculate and possibly display. */
struct percentile *percentiles;
int n_percentiles, n_show_percentiles;
+
+ /* Frequency table display. */
+ int max_categories; /* Maximum categories to show. */
+ int sort; /* FRQ_AVALUE or FRQ_DVALUE
+ or FRQ_ACOUNT or FRQ_DCOUNT. */
+
+ /* Statistics; number of statistics. */
+ unsigned long stats;
+ int n_stats;
+
+ /* Histogram and pie chart settings. */
+ struct frq_chart *hist, *pie;
};
-static void determine_charts (void);
+static void determine_charts (struct frq_proc *,
+ const struct cmd_frequencies *);
-static void calc_stats (const struct frq_proc *, const struct var_freqs *,
- double d[FRQ_N_STATS]);
+static void calc_stats (const struct var_freqs *, double d[FRQ_N_STATS]);
+static void calc_percentiles (const struct frq_proc *,
+ const struct var_freqs *);
static void precalc (struct frq_proc *, struct casereader *, struct dataset *);
static void calc (struct frq_proc *, const struct ccase *,
const struct dataset *);
static void postcalc (struct frq_proc *, const struct dataset *);
-static void postprocess_freq_tab (struct var_freqs *);
+static void postprocess_freq_tab (const struct frq_proc *, struct var_freqs *);
static void dump_freq_table (const struct var_freqs *,
const struct variable *weight_var);
static void dump_statistics (const struct frq_proc *, const struct var_freqs *,
const struct variable *weight_var);
static void cleanup_freq_tab (struct var_freqs *);
-static algo_compare_func compare_value_numeric_a, compare_value_alpha_a;
-static algo_compare_func compare_value_numeric_d, compare_value_alpha_d;
-static algo_compare_func compare_freq_numeric_a, compare_freq_alpha_a;
-static algo_compare_func compare_freq_numeric_d, compare_freq_alpha_d;
-
static void add_percentile (struct frq_proc *, double x, bool show,
size_t *allocated_percentiles);
-static void do_piechart(const struct variable *var,
- const struct freq_tab *frq_tab);
+static void do_piechart(const struct frq_chart *, const struct variable *,
+ const struct freq_tab *);
struct histogram *freq_tab_to_hist(const struct frq_proc *,
const struct freq_tab *,
int
cmd_frequencies (struct lexer *lexer, struct dataset *ds)
{
+ struct cmd_frequencies cmd;
struct frq_proc frq;
struct casegrouper *grouper;
struct casereader *input, *group;
frq.n_percentiles = 0;
frq.n_show_percentiles = 0;
+ frq.hist = NULL;
+ frq.pie = NULL;
+
allocated_percentiles = 0;
if (!parse_frequencies (lexer, ds, &cmd, &frq))
return CMD_FAILURE;
}
+ /* Figure out when to show frequency tables. */
+ frq.max_categories = (cmd.table == FRQ_NOTABLE ? -1
+ : cmd.table == FRQ_TABLE ? INT_MAX
+ : cmd.limit);
+ frq.sort = cmd.sort;
+
/* Figure out statistics to calculate. */
- stats = 0;
+ frq.stats = 0;
if (cmd.a_statistics[FRQ_ST_DEFAULT] || !cmd.sbc_statistics)
- stats |= FRQ_DEFAULT;
+ frq.stats |= FRQ_DEFAULT;
if (cmd.a_statistics[FRQ_ST_ALL])
- stats |= FRQ_ALL;
+ frq.stats |= FRQ_ALL;
if (cmd.sort != FRQ_AVALUE && cmd.sort != FRQ_DVALUE)
- stats &= ~BIT_INDEX (FRQ_MEDIAN);
+ frq.stats &= ~BIT_INDEX (FRQ_MEDIAN);
for (i = 0; i < FRQ_N_STATS; i++)
if (cmd.a_statistics[st_name[i].st_indx])
- stats |= BIT_INDEX (i);
- if (stats & FRQ_KURT)
- stats |= BIT_INDEX (FRQ_SEKURT);
- if (stats & FRQ_SKEW)
- stats |= BIT_INDEX (FRQ_SESKEW);
+ frq.stats |= BIT_INDEX (i);
+ if (frq.stats & FRQ_KURT)
+ frq.stats |= BIT_INDEX (FRQ_SEKURT);
+ if (frq.stats & FRQ_SKEW)
+ frq.stats |= BIT_INDEX (FRQ_SESKEW);
/* Calculate n_stats. */
- n_stats = 0;
+ frq.n_stats = 0;
for (i = 0; i < FRQ_N_STATS; i++)
- if ((stats & BIT_INDEX (i)))
- n_stats++;
+ if ((frq.stats & BIT_INDEX (i)))
+ frq.n_stats++;
/* Charting. */
- determine_charts ();
+ determine_charts (&frq, &cmd);
if (cmd.sbc_histogram || cmd.sbc_piechart || cmd.sbc_ntiles)
cmd.sort = FRQ_AVALUE;
&allocated_percentiles);
}
}
- if (stats & BIT_INDEX (FRQ_MEDIAN))
+ if (frq.stats & BIT_INDEX (FRQ_MEDIAN))
{
/* Treat the median as the 50% percentile.
We output it in the percentiles table as "50 (Median)." */
add_percentile (&frq, 0.5, true, &allocated_percentiles);
- stats &= ~BIT_INDEX (FRQ_MEDIAN);
- n_stats--;
+ frq.stats &= ~BIT_INDEX (FRQ_MEDIAN);
+ frq.n_stats--;
}
if (cmd.sbc_histogram)
{
pool_destroy (frq.pool);
free (frq.vars);
free (frq.percentiles);
+ free (frq.hist);
+ free (frq.pie);
return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
}
/* Figure out which charts the user requested. */
static void
-determine_charts (void)
+determine_charts (struct frq_proc *frq, const struct cmd_frequencies *cmd)
{
- if (cmd.sbc_barchart)
+ if (cmd->sbc_barchart)
msg (SW, _("Bar charts are not implemented."));
- if (cmd.sbc_histogram)
+ if (cmd->sbc_histogram)
{
- hist.x_min = cmd.hi_min;
- hist.x_max = cmd.hi_max;
- hist.y_scale = cmd.hi_scale;
- hist.y_max = cmd.hi_scale == FRQ_FREQ ? cmd.hi_freq : cmd.hi_pcnt;
- hist.draw_normal = cmd.hi_norm != FRQ_NONORMAL;
- hist.include_missing = false;
-
- if (hist.x_min != SYSMIS && hist.x_max != SYSMIS
- && hist.x_min >= hist.x_max)
+ struct frq_chart *hist;
+
+ hist = frq->hist = xmalloc (sizeof *frq->hist);
+ hist->x_min = cmd->hi_min;
+ hist->x_max = cmd->hi_max;
+ hist->y_scale = cmd->hi_scale;
+ hist->y_max = cmd->hi_scale == FRQ_FREQ ? cmd->hi_freq : cmd->hi_pcnt;
+ hist->draw_normal = cmd->hi_norm != FRQ_NONORMAL;
+ hist->include_missing = false;
+
+ if (hist->x_min != SYSMIS && hist->x_max != SYSMIS
+ && hist->x_min >= hist->x_max)
{
msg (SE, _("MAX for histogram must be greater than or equal to MIN, "
"but MIN was specified as %.15g and MAX as %.15g. "
- "MIN and MAX will be ignored."), hist.x_min, hist.x_max);
- hist.x_min = hist.x_max = SYSMIS;
+ "MIN and MAX will be ignored."),
+ hist->x_min, hist->x_max);
+ hist->x_min = hist->x_max = SYSMIS;
}
}
- if (cmd.sbc_piechart)
+ if (cmd->sbc_piechart)
{
- pie.x_min = cmd.pie_min;
- pie.x_max = cmd.pie_max;
- pie.y_scale = cmd.pie_scale;
- pie.include_missing = cmd.pie_missing == FRQ_MISSING;
+ struct frq_chart *pie;
- if (pie.x_min != SYSMIS && pie.x_max != SYSMIS
- && pie.x_min >= pie.x_max)
+ pie = frq->pie = xmalloc (sizeof *frq->pie);
+ pie->x_min = cmd->pie_min;
+ pie->x_max = cmd->pie_max;
+ pie->y_scale = cmd->pie_scale;
+ pie->include_missing = cmd->pie_missing == FRQ_MISSING;
+
+ if (pie->x_min != SYSMIS && pie->x_max != SYSMIS
+ && pie->x_min >= pie->x_max)
{
msg (SE, _("MAX for pie chart must be greater than or equal to MIN, "
"but MIN was specified as %.15g and MAX as %.15g. "
- "MIN and MAX will be ignored."), pie.x_min, pie.x_max);
- pie.x_min = pie.x_max = SYSMIS;
+ "MIN and MAX will be ignored."), pie->x_min, pie->x_max);
+ pie->x_min = pie->x_max = SYSMIS;
}
}
-
}
/* Add data from case C to the frequency table. */
for (i = 0; i < frq->n_vars; i++)
{
struct var_freqs *vf = &frq->vars[i];
- const union value *value = case_data (c, vf->var);
- size_t hash = value_hash (value, vf->width, 0);
- struct freq *f;
+ struct casewriter *writer = vf->tab.sorter;
+ struct ccase *f;
- f = freq_hmap_search (&vf->tab.data, value, vf->width, hash);
- if (f == NULL)
- f = freq_hmap_insert (&vf->tab.data, value, vf->width, hash);
+ f = case_create (casewriter_get_proto (writer));
+ value_copy (case_data_rw_idx (f, 0), case_data (c, vf->var), vf->width);
+ case_data_rw_idx (f, 1)->f = weight;
- f->count += weight;
+ casewriter_write (writer, f);
}
}
+static struct ccase *
+combine_freq_cases (struct ccase *a, struct ccase *b, void *aux UNUSED)
+{
+ a = case_unshare (a);
+ case_data_rw_idx (a, 1)->f += case_data_idx (b, 1)->f;
+ case_unref (b);
+
+ return a;
+}
+
/* Prepares each variable that is the target of FREQUENCIES by setting
up its hash table. */
static void
}
for (i = 0; i < frq->n_vars; i++)
- hmap_init (&frq->vars[i].tab.data);
+ {
+ int width = var_get_width (frq->vars[i].var);
+ struct caseproto *proto;
+ struct subcase ordering;
+
+ proto = caseproto_create ();
+ proto = caseproto_add_width (proto, width);
+ proto = caseproto_add_width (proto, 0);
+
+ subcase_init (&ordering, 0, width, SC_ASCEND);
+ frq->vars[i].tab.sorter = sort_distinct_create_writer (
+ &ordering, proto, combine_freq_cases, NULL, NULL);
+
+ caseproto_unref (proto);
+ subcase_destroy (&ordering);
+ }
}
/* Finishes up with the variables after frequencies have been
for (i = 0; i < frq->n_vars; i++)
{
struct var_freqs *vf = &frq->vars[i];
- int n_categories;
- postprocess_freq_tab (vf);
+ postprocess_freq_tab (frq, vf);
/* Frequencies tables. */
- n_categories = vf->tab.n_valid + vf->tab.n_missing;
- if (cmd.table == FRQ_TABLE
- || (cmd.table == FRQ_LIMIT && n_categories <= cmd.limit))
+ if (vf->tab.n_valid + vf->tab.n_missing <= frq->max_categories)
dump_freq_table (vf, wv);
/* Statistics. */
- if (n_stats)
+ if (frq->n_stats)
dump_statistics (frq, vf, wv);
- if (cmd.sbc_histogram && var_is_numeric (vf->var) && vf->tab.n_valid > 0)
+ if (frq->hist && var_is_numeric (vf->var) && vf->tab.n_valid > 0)
{
double d[FRQ_N_STATS];
struct histogram *histogram;
- calc_stats (frq, vf, d);
+ calc_stats (vf, d);
histogram = freq_tab_to_hist (frq, &vf->tab, vf->var);
vf->tab.valid_cases,
d[FRQ_MEAN],
d[FRQ_STDDEV],
- hist.draw_normal));
+ frq->hist->draw_normal));
statistic_destroy (&histogram->parent);
}
- if (cmd.sbc_piechart)
- do_piechart(vf->var, &vf->tab);
+ if (frq->pie)
+ do_piechart(frq->pie, vf->var, &vf->tab);
cleanup_freq_tab (vf);
}
}
-/* Returns the comparison function that should be used for
- sorting a frequency table by FRQ_SORT using VAL_TYPE
- values. */
-static algo_compare_func *
-get_freq_comparator (int frq_sort, enum val_type val_type)
-{
- bool is_numeric = val_type == VAL_NUMERIC;
- switch (frq_sort)
- {
- case FRQ_AVALUE:
- return is_numeric ? compare_value_numeric_a : compare_value_alpha_a;
- case FRQ_DVALUE:
- return is_numeric ? compare_value_numeric_d : compare_value_alpha_d;
- case FRQ_AFREQ:
- return is_numeric ? compare_freq_numeric_a : compare_freq_alpha_a;
- case FRQ_DFREQ:
- return is_numeric ? compare_freq_numeric_d : compare_freq_alpha_d;
- default:
- NOT_REACHED ();
- }
-}
-
/* Returns true iff the value in struct freq F is non-missing
for variable V. */
static bool
return !var_is_value_missing (v, &f->value, MV_ANY);
}
+struct freq_compare_aux
+ {
+ bool by_freq;
+ bool ascending_freq;
+
+ int width;
+ bool ascending_value;
+ };
+
+static int
+compare_freq (const void *a_, const void *b_, const void *aux_)
+{
+ const struct freq_compare_aux *aux = aux_;
+ const struct freq *a = a_;
+ const struct freq *b = b_;
+
+ if (aux->by_freq && a->count != b->count)
+ {
+ int cmp = a->count > b->count ? 1 : -1;
+ return aux->ascending_freq ? cmp : -cmp;
+ }
+ else
+ {
+ int cmp = value_compare_3way (&a->value, &b->value, aux->width);
+ return aux->ascending_value ? cmp : -cmp;
+ }
+}
/* Summarizes the frequency table data for variable V. */
static void
-postprocess_freq_tab (struct var_freqs *vf)
+postprocess_freq_tab (const struct frq_proc *frq, struct var_freqs *vf)
{
struct freq_tab *ft = &vf->tab;
- algo_compare_func *compare;
+ struct freq_compare_aux aux;
+ struct casereader *reader;
+ struct ccase *c;
size_t count;
struct freq *freqs, *f;
size_t i;
/* Extract data from hash table. */
- count = hmap_count (&ft->data);
- freqs = freq_hmap_extract (&ft->data);
+ reader = casewriter_make_reader (ft->sorter);
+ ft->data = casereader_clone (reader);
+ freqs = xnmalloc (casereader_count_cases (reader), sizeof *freqs);
+ for (count = 0; (c = casereader_read (reader)) != NULL; count++)
+ {
+ struct freq *f = &freqs[count];
+ value_clone (&f->value, case_data_idx (c, 0), vf->width);
+ f->count = case_num_idx (c, 1);
+ case_unref (c);
+ }
+ casereader_destroy (reader);
/* Put data into ft. */
ft->valid = freqs;
ft->n_missing = count - ft->n_valid;
/* Sort data. */
- compare = get_freq_comparator (cmd.sort, var_get_type (vf->var));
- sort (ft->valid, ft->n_valid, sizeof *ft->valid, compare, vf);
- sort (ft->missing, ft->n_missing, sizeof *ft->missing, compare, vf);
+ aux.by_freq = frq->sort == FRQ_AFREQ || frq->sort == FRQ_DFREQ;
+ aux.ascending_freq = frq->sort != FRQ_DFREQ;
+ aux.width = vf->width;
+ aux.ascending_value = frq->sort != FRQ_DVALUE;
+ sort (ft->valid, ft->n_valid, sizeof *ft->valid, compare_freq, &aux);
+ sort (ft->missing, ft->n_missing, sizeof *ft->missing, compare_freq, &aux);
/* Summary statistics. */
ft->valid_cases = 0.0;
static void
cleanup_freq_tab (struct var_freqs *vf)
{
+ if (value_needs_init (vf->width))
+ {
+ int i;
+
+ for (i = 0; i < vf->tab.n_valid + vf->tab.n_missing; i++)
+ value_destroy (&vf->tab.valid[i].value, vf->width);
+ }
+ casereader_destroy (vf->tab.data);
free (vf->tab.valid);
- freq_hmap_destroy (&vf->tab.data, vf->width);
}
/* Parses the VARIABLES subcommand. */
/* Comparison functions. */
-/* Ascending numeric compare of values. */
-static int
-compare_value_numeric_a (const void *a_, const void *b_,
- const void *vf_ UNUSED)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
-
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
- else
- return 0;
-}
-
-/* Ascending string compare of values. */
-static int
-compare_value_alpha_a (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
-
- return value_compare_3way (&a->value, &b->value, vf->width);
-}
+\f
+/* Frequency table display. */
-/* Descending numeric compare of values. */
-static int
-compare_value_numeric_d (const void *a, const void *b, const void *vf_ UNUSED)
+/* Displays a full frequency table for variable V. */
+static void
+dump_freq_table (const struct var_freqs *vf, const struct variable *wv)
{
- return -compare_value_numeric_a (a, b, vf_);
-}
+ struct variable *value_var, *freq_var;
+ struct pivot_table *pt;
-/* Descending string compare of values. */
-static int
-compare_value_alpha_d (const void *a, const void *b, const void *vf_)
-{
- return -compare_value_alpha_a (a, b, vf_);
-}
+ pt = xmalloc (sizeof *pt);
-/* Ascending numeric compare of frequency;
- secondary key on ascending numeric value. */
-static int
-compare_freq_numeric_a (const void *a_, const void *b_, const void *vf_ UNUSED)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
+ pt->data = casereader_clone (vf->tab.data);
- if (a->count > b->count)
- return 1;
- else if (a->count < b->count)
- return -1;
-
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
+ pt->dict = dict_create ();
+ value_var = dict_clone_var_assert (pt->dict, vf->var);
+ if (wv != NULL)
+ freq_var = dict_clone_var_assert (pt->dict, wv);
else
- return 0;
-}
-
-/* Ascending numeric compare of frequency;
- secondary key on ascending string value. */
-static int
-compare_freq_alpha_a (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
+ {
+ freq_var = dict_create_var (pt->dict, "$Frequency", 0);
+ var_set_both_formats (freq_var, &F_8_0);
+ var_set_label (freq_var, _("Frequency"));
+ }
- if (a->count > b->count)
- return 1;
- else if (a->count < b->count)
- return -1;
- else
- return value_compare_3way (&a->value, &b->value, vf->width);
+ subcase_init_empty (&pt->split);
+ subcase_init_empty (&pt->dimensions[TABLE_HORZ]);
+ subcase_init_empty (&pt->dimensions[TABLE_VERT]);
+ subcase_add_var (&pt->dimensions[TABLE_VERT], value_var, SC_ASCEND);
+
+ pt->pane.n[TABLE_HORZ] = 4;
+ pt->pane.n[TABLE_VERT] = 1;
+ pt->pane.cells = xmalloc (sizeof *pt->pane.cells);
+ pt->pane.cells[0] = xmalloc (4 * sizeof **pt->pane.cells);
+
+ /* Frequency. */
+ pivot_cell_init (
+ &pt->pane.cells[0][0],
+ pivot_value_create (freq_var, PIVOT_SUM, 0, 0, PIVOT_INCLUDE_ALL));
+
+ /* Percent. */
+ pivot_cell_init (
+ &pt->pane.cells[0][1],
+ pivot_value_create (freq_var, PIVOT_SUM, 0, 0, PIVOT_INCLUDE_ALL));
+ pt->pane.cells[0][1].cmp = PIVOT_PERCENT;
+ pt->pane.cells[0][1].cmp_args[0] = pivot_value_create (
+ freq_var, PIVOT_SUM, 0, 1, PIVOT_INCLUDE_ALL);
+ pt->pane.cells[0][1].label = xstrdup ("Percent");
+
+ /* Valid Percent. */
+ pivot_cell_init (
+ &pt->pane.cells[0][2],
+ pivot_value_create (freq_var, PIVOT_SUM, 0, 0, PIVOT_INCLUDE_VALID));
+ pt->pane.cells[0][2].cmp = PIVOT_PERCENT;
+ pt->pane.cells[0][2].cmp_args[0] = pivot_value_create (
+ freq_var, PIVOT_SUM, 0, 1, PIVOT_INCLUDE_VALID);
+ pt->pane.cells[0][2].label = xstrdup ("Valid Percent");
+
+ /* Cumulative (Valid) Percent. */
+ pivot_cell_init (
+ &pt->pane.cells[0][3],
+ pivot_value_create (freq_var, PIVOT_SUM, 0, 0, PIVOT_INCLUDE_VALID));
+ pt->pane.cells[0][3].base->n_cum_vars[TABLE_VERT] = 1;
+ pt->pane.cells[0][3].cmp = PIVOT_PERCENT;
+ pt->pane.cells[0][3].cmp_args[0] = pivot_value_create (
+ freq_var, PIVOT_SUM, 0, 1, PIVOT_INCLUDE_VALID);
+ pt->pane.cells[0][3].label = xstrdup ("Cum Percent");
+
+ pivot_table_dump (pt);
}
+\f
+/* Statistical display. */
-/* Descending numeric compare of frequency;
- secondary key on ascending numeric value. */
-static int
-compare_freq_numeric_d (const void *a_, const void *b_, const void *vf_ UNUSED)
+static double
+calc_percentile (double p, double valid_cases, double x1, double x2)
{
- const struct freq *a = a_;
- const struct freq *b = b_;
-
- if (a->count > b->count)
- return -1;
- else if (a->count < b->count)
- return 1;
+ double s, dummy;
- if (a->value.f > b->value.f)
- return 1;
- else if (a->value.f < b->value.f)
- return -1;
- else
- return 0;
-}
+ s = (settings_get_algorithm () != COMPATIBLE
+ ? modf ((valid_cases - 1) * p, &dummy)
+ : modf ((valid_cases + 1) * p - 1, &dummy));
-/* Descending numeric compare of frequency;
- secondary key on ascending string value. */
-static int
-compare_freq_alpha_d (const void *a_, const void *b_, const void *vf_)
-{
- const struct freq *a = a_;
- const struct freq *b = b_;
- const struct var_freqs *vf = vf_;
-
- if (a->count > b->count)
- return -1;
- else if (a->count < b->count)
- return 1;
- else
- return value_compare_3way (&a->value, &b->value, vf->width);
+ return x1 + (x2 - x1) * s;
}
-\f
-/* Frequency table display. */
-/* Displays a full frequency table for variable V. */
+/* Calculates all of the percentiles for VF within FRQ. */
static void
-dump_freq_table (const struct var_freqs *vf, const struct variable *wv)
+calc_percentiles (const struct frq_proc *frq, const struct var_freqs *vf)
{
- const struct fmt_spec *wfmt = wv ? var_get_print_format (wv) : &F_8_0;
const struct freq_tab *ft = &vf->tab;
- int n_categories;
- struct freq *f;
- struct tab_table *t;
- int r, x;
- double cum_total = 0.0;
- double cum_freq = 0.0;
-
- static const char *headings[] = {
- N_("Value Label"),
- N_("Value"),
- N_("Frequency"),
- N_("Percent"),
- N_("Valid Percent"),
- N_("Cum Percent")
- };
-
- n_categories = ft->n_valid + ft->n_missing;
- t = tab_create (6, n_categories + 2);
- tab_headers (t, 0, 0, 1, 0);
+ double W = ft->valid_cases;
+ const struct freq *f;
+ int percentile_idx;
+ double rank;
- for (x = 0; x < 6; x++)
- tab_text (t, x, 0, TAB_CENTER | TAT_TITLE, gettext (headings[x]));
+ assert (ft->n_valid > 0);
- r = 1;
+ rank = 0;
+ percentile_idx = 0;
for (f = ft->valid; f < ft->missing; f++)
{
- const char *label;
- double percent, valid_percent;
-
- cum_freq += f->count;
+ rank += f->count;
+ for (; percentile_idx < frq->n_percentiles; percentile_idx++)
+ {
+ struct percentile *pc = &frq->percentiles[percentile_idx];
+ double tp;
- percent = f->count / ft->total_cases * 100.0;
- valid_percent = f->count / ft->valid_cases * 100.0;
- cum_total += valid_percent;
+ tp = (settings_get_algorithm () == ENHANCED
+ ? (W - 1) * pc->p
+ : (W + 1) * pc->p - 1);
- label = var_lookup_value_label (vf->var, &f->value);
- if (label != NULL)
- tab_text (t, 0, r, TAB_LEFT, label);
+ if (rank <= tp)
+ break;
- tab_value (t, 1, r, TAB_NONE, &f->value, ft->dict, &vf->print);
- tab_double (t, 2, r, TAB_NONE, f->count, wfmt);
- tab_double (t, 3, r, TAB_NONE, percent, NULL);
- tab_double (t, 4, r, TAB_NONE, valid_percent, NULL);
- tab_double (t, 5, r, TAB_NONE, cum_total, NULL);
- r++;
+ if (f->count > 1
+ && (rank - (f->count - 1) > tp || f + 1 >= ft->missing))
+ pc->value = f->value.f;
+ else
+ pc->value = calc_percentile (pc->p, W, f->value.f, f[1].value.f);
+ }
}
- for (; f < &ft->valid[n_categories]; f++)
+ for (; percentile_idx < frq->n_percentiles; percentile_idx++)
{
- const char *label;
-
- cum_freq += f->count;
-
- label = var_lookup_value_label (vf->var, &f->value);
- if (label != NULL)
- tab_text (t, 0, r, TAB_LEFT, label);
-
- tab_value (t, 1, r, TAB_NONE, &f->value, ft->dict, &vf->print);
- tab_double (t, 2, r, TAB_NONE, f->count, wfmt);
- tab_double (t, 3, r, TAB_NONE,
- f->count / ft->total_cases * 100.0, NULL);
- tab_text (t, 4, r, TAB_NONE, _("Missing"));
- r++;
+ struct percentile *pc = &frq->percentiles[percentile_idx];
+ pc->value = ft->valid[ft->n_valid - 1].value.f;
}
-
- tab_box (t, TAL_1, TAL_1, -1, TAL_1, 0, 0, 5, r);
- tab_hline (t, TAL_2, 0, 5, 1);
- tab_hline (t, TAL_2, 0, 5, r);
- tab_joint_text (t, 0, r, 1, r, TAB_RIGHT | TAT_TITLE, _("Total"));
- tab_vline (t, TAL_0, 1, r, r);
- tab_double (t, 2, r, TAB_NONE, cum_freq, wfmt);
- tab_fixed (t, 3, r, TAB_NONE, 100.0, 5, 1);
- tab_fixed (t, 4, r, TAB_NONE, 100.0, 5, 1);
-
- tab_title (t, "%s", var_to_string (vf->var));
- tab_submit (t);
}
-\f
-/* Statistical display. */
-/* Calculates all the pertinent statistics for variable V, putting them in
- array D[]. */
+/* Calculates all the pertinent statistics for VF, putting them in array
+ D[]. */
static void
-calc_stats (const struct frq_proc *frq,
- const struct var_freqs *vf, double d[FRQ_N_STATS])
+calc_stats (const struct var_freqs *vf, double d[FRQ_N_STATS])
{
const struct freq_tab *ft = &vf->tab;
double W = ft->valid_cases;
+ const struct freq *f;
struct moments *m;
- struct freq *f=0;
- double prev_value;
int most_often;
double X_mode;
- double rank;
- int i = 0;
- int idx;
-
- /* Calculate percentiles. */
-
assert (ft->n_valid > 0);
- for (i = 0; i < frq->n_percentiles; i++)
- {
- struct percentile *pc = &frq->percentiles[i];
-
- pc->flag = 0;
- pc->flag2 = 0;
- }
-
- rank = 0;
- prev_value = SYSMIS;
- for (idx = 0; idx < ft->n_valid; ++idx)
- {
- f = &ft->valid[idx];
- rank += f->count ;
- for (i = 0; i < frq->n_percentiles; i++)
- {
- struct percentile *pc = &frq->percentiles[i];
- double tp;
-
- if ( pc->flag2 ) continue ;
-
- if ( settings_get_algorithm () != COMPATIBLE )
- tp =
- (ft->valid_cases - 1) * pc->p;
- else
- tp =
- (ft->valid_cases + 1) * pc->p - 1;
-
- if ( pc->flag )
- {
- pc->x2 = f->value.f;
- pc->x1 = prev_value;
- pc->flag2 = 1;
- continue;
- }
-
- if (rank > tp )
- {
- if ( f->count > 1 && rank - (f->count - 1) > tp )
- {
- pc->x2 = pc->x1 = f->value.f;
- pc->flag2 = 1;
- }
- else
- {
- pc->flag=1;
- }
-
- continue;
- }
- }
- prev_value = f->value.f;
- }
-
- for (i = 0; i < frq->n_percentiles; i++)
- {
- struct percentile *pc = &frq->percentiles[i];
-
- /* Catches the case when p == 100% */
- if ( ! pc->flag2 )
- pc->x1 = pc->x2 = f->value.f;
-
- /*
- printf("percentile %d (p==%.2f); X1 = %g; X2 = %g\n",
- i,pc->p,pc->x1,pc->x2);
- */
- }
-
- for (i = 0; i < frq->n_percentiles; i++)
- {
- struct percentile *pc = &frq->percentiles[i];
- double s;
-
- double dummy;
- if ( settings_get_algorithm () != COMPATIBLE )
- {
- s = modf((ft->valid_cases - 1) * pc->p , &dummy);
- }
- else
- {
- s = modf((ft->valid_cases + 1) * pc->p -1, &dummy);
- }
-
- pc->value = pc->x1 + (pc->x2 - pc->x1) * s ;
- }
-
-
/* Calculate the mode. */
most_often = -1;
X_mode = SYSMIS;
var_get_name (vf->var));
return;
}
- calc_stats (frq, vf, stat_value);
+ calc_stats (vf, stat_value);
+ calc_percentiles (frq, vf);
- t = tab_create (3, n_stats + frq->n_show_percentiles + 2);
+ t = tab_create (3, frq->n_stats + frq->n_show_percentiles + 2);
tab_box (t, TAL_1, TAL_1, -1, -1 , 0 , 0 , 2, tab_nr(t) - 1) ;
r=2; /* N missing and N valid are always dumped */
for (i = 0; i < FRQ_N_STATS; i++)
- if (stats & BIT_INDEX (i))
+ if (frq->stats & BIT_INDEX (i))
{
tab_text (t, 0, r, TAB_LEFT | TAT_TITLE,
gettext (st_name[i].s10));
valid_freq = 0;
for (i = 0; i < ft->n_valid; i++)
{
- const struct freq *frq = &ft->valid[i];
- if (chart_includes_value (&hist, var, &frq->value))
+ const struct freq *f = &ft->valid[i];
+ if (chart_includes_value (frq->hist, var, &f->value))
{
- x_min = MIN (x_min, frq->value.f);
- x_max = MAX (x_max, frq->value.f);
- valid_freq += frq->count;
+ x_min = MIN (x_min, f->value.f);
+ x_max = MAX (x_max, f->value.f);
+ valid_freq += f->count;
}
}
histogram = histogram_create (bins, x_min, x_max);
for (i = 0; i < ft->n_valid; i++)
{
- const struct freq *frq = &ft->valid[i];
- if (chart_includes_value (&hist, var, &frq->value))
- histogram_add (histogram, frq->value.f, frq->count);
+ const struct freq *f = &ft->valid[i];
+ if (chart_includes_value (frq->hist, var, &f->value))
+ histogram_add (histogram, f->value.f, f->count);
}
return histogram;
}
static int
-add_slice (const struct freq *freq, const struct variable *var,
- struct slice *slice)
+add_slice (const struct frq_chart *pie, const struct freq *freq,
+ const struct variable *var, struct slice *slice)
{
- if (chart_includes_value (&pie, var, &freq->value))
+ if (chart_includes_value (pie, var, &freq->value))
{
ds_init_empty (&slice->label);
var_append_value_name (var, &freq->value, &slice->label);
The caller is responsible for freeing slices
*/
static struct slice *
-freq_tab_to_slice_array(const struct freq_tab *frq_tab,
+freq_tab_to_slice_array(const struct frq_chart *pie,
+ const struct freq_tab *frq_tab,
const struct variable *var,
int *n_slicesp)
{
n_slices = 0;
for (i = 0; i < frq_tab->n_valid; i++)
- n_slices += add_slice (&frq_tab->valid[i], var, &slices[n_slices]);
+ n_slices += add_slice (pie, &frq_tab->valid[i], var, &slices[n_slices]);
for (i = 0; i < frq_tab->n_missing; i++)
- n_slices += add_slice (&frq_tab->missing[i], var, &slices[n_slices]);
+ n_slices += add_slice (pie, &frq_tab->missing[i], var, &slices[n_slices]);
*n_slicesp = n_slices;
return slices;
static void
-do_piechart(const struct variable *var, const struct freq_tab *frq_tab)
+do_piechart(const struct frq_chart *pie, const struct variable *var,
+ const struct freq_tab *frq_tab)
{
struct slice *slices;
int n_slices, i;
- slices = freq_tab_to_slice_array (frq_tab, var, &n_slices);
+ slices = freq_tab_to_slice_array (pie, frq_tab, var, &n_slices);
if (n_slices < 2)
msg (SW, _("Omitting pie chart for %s, which has only %d unique values."),
projects / pspp / blobdiff