--- /dev/null
+/* PSPP - a program for statistical analysis.
+ Copyright (C) 1997-9, 2000, 2009, 2010, 2012, 2013, 2014
+ 2021, Free Software Foundation, Inc.
+
+ This program is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <config.h>
+
+#include <float.h>
+#include <stdlib.h>
+
+#include "data/case.h"
+#include "data/casereader.h"
+#include "data/dataset.h"
+#include "data/dictionary.h"
+#include "data/transformations.h"
+#include "data/variable.h"
+#include "language/command.h"
+#include "language/lexer/lexer.h"
+#include "language/lexer/variable-parser.h"
+#include "libpspp/array.h"
+#include "libpspp/compiler.h"
+#include "libpspp/hash-functions.h"
+#include "libpspp/hmap.h"
+#include "libpspp/i18n.h"
+#include "libpspp/message.h"
+#include "libpspp/pool.h"
+#include "libpspp/str.h"
+#include "output/pivot-table.h"
+
+#include "gl/xalloc.h"
+#include "gl/c-xvasprintf.h"
+#include "gl/mbiter.h"
+#include "gl/size_max.h"
+
+#include "gettext.h"
+#define _(msgid) gettext (msgid)
+#define N_(msgid) (msgid)
+
+/* Explains how to recode one value. */
+struct arc_item
+ {
+ struct hmap_node hmap_node; /* Element in "struct arc_spec" hash table. */
+ union value from; /* Original value. */
+ int width; /* Width of the original value */
+ bool missing; /* Is 'from' missing in its source varible? */
+ char *value_label; /* Value label in source variable, if any. */
+
+ double to; /* Recoded value. */
+ };
+
+/* Explains how to recode an AUTORECODE variable. */
+struct arc_spec
+ {
+ int width; /* Variable width. */
+ int src_idx; /* Case index of source variable. */
+ char *src_name; /* Name of source variable. */
+ struct fmt_spec format; /* Print format in source variable. */
+ struct variable *dst; /* Target variable. */
+ struct missing_values mv; /* Missing values of source variable. */
+ char *label; /* Variable label of source variable. */
+ struct rec_items *items;
+ };
+
+/* Descending or ascending sort order. */
+enum arc_direction
+ {
+ ASCENDING,
+ DESCENDING
+ };
+
+struct rec_items
+{
+ struct hmap ht; /* Hash table of "struct arc_item"s. */
+};
+
+
+
+/* AUTORECODE data. */
+struct autorecode_pgm
+{
+ struct arc_spec *specs;
+ size_t n_specs;
+
+ bool blank_valid;
+};
+
+static const struct trns_class autorecode_trns_class;
+
+static int compare_arc_items (const void *, const void *, const void *aux);
+static void arc_free (struct autorecode_pgm *);
+static struct arc_item *find_arc_item (
+ const struct rec_items *, const union value *, int width,
+ size_t hash);
+
+/* Returns WIDTH with any trailing spaces in VALUE trimmed off (except that a
+ minimum width of 1 is always returned because otherwise the width would
+ indicate a numeric type). */
+static int
+value_trim_spaces (const union value *value, int width)
+{
+ while (width > 1 && value->s[width - 1] == ' ')
+ width--;
+ return width;
+}
+
+/* Performs the AUTORECODE procedure. */
+int
+cmd_autorecode (struct lexer *lexer, struct dataset *ds)
+{
+ struct dictionary *dict = dataset_dict (ds);
+
+ const struct variable **src_vars = NULL;
+ size_t n_srcs = 0;
+
+ char **dst_names = NULL;
+ size_t n_dsts = 0;
+
+ enum arc_direction direction = ASCENDING;
+ bool print = false;
+
+ /* Create procedure. */
+ struct autorecode_pgm *arc = XZALLOC (struct autorecode_pgm);
+ arc->blank_valid = true;
+
+ /* Parse variable lists. */
+ lex_match_id (lexer, "VARIABLES");
+ lex_match (lexer, T_EQUALS);
+ if (!parse_variables_const (lexer, dict, &src_vars, &n_srcs,
+ PV_NO_DUPLICATE | PV_NO_SCRATCH))
+ goto error;
+ lex_match (lexer, T_SLASH);
+ if (!lex_force_match_id (lexer, "INTO"))
+ goto error;
+ lex_match (lexer, T_EQUALS);
+ if (!parse_DATA_LIST_vars (lexer, dict, &dst_names, &n_dsts,
+ PV_NO_DUPLICATE))
+ goto error;
+ if (n_dsts != n_srcs)
+ {
+ msg (SE, _("Source variable count (%zu) does not match "
+ "target variable count (%zu)."),
+ n_srcs, n_dsts);
+
+ goto error;
+ }
+ for (size_t i = 0; i < n_dsts; i++)
+ {
+ const char *name = dst_names[i];
+
+ if (dict_lookup_var (dict, name) != NULL)
+ {
+ msg (SE, _("Target variable %s duplicates existing variable %s."),
+ name, name);
+ goto error;
+ }
+ }
+
+ /* Parse options. */
+ bool group = false;
+ while (lex_match (lexer, T_SLASH))
+ {
+ if (lex_match_id (lexer, "DESCENDING"))
+ direction = DESCENDING;
+ else if (lex_match_id (lexer, "PRINT"))
+ print = true;
+ else if (lex_match_id (lexer, "GROUP"))
+ group = true;
+ else if (lex_match_id (lexer, "BLANK"))
+ {
+ lex_match (lexer, T_EQUALS);
+ if (lex_match_id (lexer, "VALID"))
+ {
+ arc->blank_valid = true;
+ }
+ else if (lex_match_id (lexer, "MISSING"))
+ {
+ arc->blank_valid = false;
+ }
+ else
+ {
+ lex_error_expecting (lexer, "VALID", "MISSING");
+ goto error;
+ }
+ }
+ else
+ {
+ lex_error_expecting (lexer, "DESCENDING", "PRINT", "GROUP", "BLANK");
+ goto error;
+ }
+ }
+
+ if (lex_token (lexer) != T_ENDCMD)
+ {
+ lex_error (lexer, _("Syntax error expecting end of command."));
+ goto error;
+ }
+
+ /* If GROUP is specified, verify that the variables are all string or all
+ numeric. */
+ if (group)
+ {
+ enum val_type type = var_get_type (src_vars[0]);
+ for (size_t i = 1; i < n_dsts; i++)
+ {
+ if (var_get_type (src_vars[i]) != type)
+ {
+ size_t string_idx = type == VAL_STRING ? 0 : i;
+ size_t numeric_idx = type == VAL_STRING ? i : 0;
+ lex_error (lexer, _("With GROUP, variables may not mix string "
+ "variables (such as %s) and numeric "
+ "variables (such as %s)."),
+ var_get_name (src_vars[string_idx]),
+ var_get_name (src_vars[numeric_idx]));
+ goto error;
+ }
+ }
+ }
+
+ /* Allocate all the specs and the rec_items that they point to.
+
+ If GROUP is specified, there is only a single global rec_items, with the
+ maximum width 'width', and all of the specs point to it; otherwise each
+ spec has its own rec_items. */
+ arc->specs = xmalloc (n_dsts * sizeof *arc->specs);
+ arc->n_specs = n_dsts;
+ for (size_t i = 0; i < n_dsts; i++)
+ {
+ struct arc_spec *spec = &arc->specs[i];
+
+ spec->width = var_get_width (src_vars[i]);
+ spec->src_idx = var_get_case_index (src_vars[i]);
+ spec->src_name = xstrdup (var_get_name (src_vars[i]));
+ spec->format = *var_get_print_format (src_vars[i]);
+
+ const char *label = var_get_label (src_vars[i]);
+ spec->label = xstrdup_if_nonnull (label);
+
+ if (group && i > 0)
+ spec->items = arc->specs[0].items;
+ else
+ {
+ spec->items = xzalloc (sizeof (*spec->items));
+ hmap_init (&spec->items->ht);
+ }
+ }
+
+ /* Initialize specs[*]->mv to the user-missing values for each
+ source variable. */
+ if (group)
+ {
+ /* Use the first source variable that has any user-missing values. */
+ size_t mv_idx = 0;
+ for (size_t i = 0; i < n_dsts; i++)
+ if (var_has_missing_values (src_vars[i]))
+ {
+ mv_idx = i;
+ break;
+ }
+
+ for (size_t i = 0; i < n_dsts; i++)
+ mv_copy (&arc->specs[i].mv, var_get_missing_values (src_vars[mv_idx]));
+ }
+ else
+ {
+ /* Each variable uses its own user-missing values. */
+ for (size_t i = 0; i < n_dsts; i++)
+ mv_copy (&arc->specs[i].mv, var_get_missing_values (src_vars[i]));
+ }
+
+ /* Execute procedure. */
+ struct casereader *input = proc_open (ds);
+ struct ccase *c;
+ for (; (c = casereader_read (input)) != NULL; case_unref (c))
+ for (size_t i = 0; i < arc->n_specs; i++)
+ {
+ struct arc_spec *spec = &arc->specs[i];
+ const union value *value = case_data_idx (c, spec->src_idx);
+ if (spec->width == 0 && value->f == SYSMIS)
+ {
+ /* AUTORECODE never changes the system-missing value.
+ (Leaving it out of the translation table has this
+ effect automatically because values not found in the
+ translation table get translated to system-missing.) */
+ continue;
+ }
+
+ int width = value_trim_spaces (value, spec->width);
+ if (width == 1 && value->s[0] == ' ' && !arc->blank_valid)
+ continue;
+
+ size_t hash = value_hash (value, width, 0);
+ if (find_arc_item (spec->items, value, width, hash))
+ continue;
+
+ struct string value_label = DS_EMPTY_INITIALIZER;
+ var_append_value_name__ (src_vars[i], value,
+ SETTINGS_VALUE_SHOW_LABEL, &value_label);
+
+ struct arc_item *item = xmalloc (sizeof *item);
+ item->width = width;
+ value_clone (&item->from, value, width);
+ item->missing = mv_is_value_missing_varwidth (&spec->mv, value,
+ spec->width);
+ item->value_label = ds_steal_cstr (&value_label);
+ hmap_insert (&spec->items->ht, &item->hmap_node, hash);
+
+ ds_destroy (&value_label);
+ }
+ bool ok = casereader_destroy (input);
+ ok = proc_commit (ds) && ok;
+
+ /* Re-fetch dictionary because it might have changed (if TEMPORARY was in
+ use). */
+ dict = dataset_dict (ds);
+
+ /* Create transformation. */
+ for (size_t i = 0; i < arc->n_specs; i++)
+ {
+ struct arc_spec *spec = &arc->specs[i];
+
+ /* Create destination variable. */
+ spec->dst = dict_create_var_assert (dict, dst_names[i], 0);
+ var_set_label (spec->dst, spec->label);
+
+ /* Set print format. */
+ size_t n_items = hmap_count (&spec->items->ht);
+ char *longest_value = xasprintf ("%zu", n_items);
+ struct fmt_spec format = { .type = FMT_F, .w = strlen (longest_value) };
+ var_set_both_formats (spec->dst, &format);
+ free (longest_value);
+
+ /* Create array of pointers to items. */
+ struct arc_item **items = xmalloc (n_items * sizeof *items);
+ struct arc_item *item;
+ size_t j = 0;
+ HMAP_FOR_EACH (item, struct arc_item, hmap_node, &spec->items->ht)
+ items[j++] = item;
+ assert (j == n_items);
+
+ /* Sort array by value. */
+ sort (items, n_items, sizeof *items, compare_arc_items, &direction);
+
+ /* Assign recoded values in sorted order. */
+ for (j = 0; j < n_items; j++)
+ items[j]->to = j + 1;
+
+ if (print && (!group || i == 0))
+ {
+ struct pivot_value *title
+ = (group
+ ? pivot_value_new_text (N_("Recoding grouped variables."))
+ : spec->label && spec->label[0]
+ ? pivot_value_new_text_format (N_("Recoding %s into %s (%s)."),
+ spec->src_name,
+ var_get_name (spec->dst),
+ spec->label)
+ : pivot_value_new_text_format (N_("Recoding %s into %s."),
+ spec->src_name,
+ var_get_name (spec->dst)));
+ struct pivot_table *table = pivot_table_create__ (title, "Recoding");
+
+ pivot_dimension_create (
+ table, PIVOT_AXIS_COLUMN, N_("Attributes"),
+ N_("New Value"), N_("Value Label"));
+
+ struct pivot_dimension *old_values = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Old Value"));
+ old_values->root->show_label = true;
+
+ for (size_t k = 0; k < n_items; k++)
+ {
+ const struct arc_item *item = items[k];
+ int old_value_idx = pivot_category_create_leaf (
+ old_values->root, pivot_value_new_value (
+ &item->from, item->width,
+ (item->width
+ ? &(struct fmt_spec) { .type = FMT_F, .w = item->width }
+ : &spec->format),
+ dict_get_encoding (dict)));
+ pivot_table_put2 (table, 0, old_value_idx,
+ pivot_value_new_integer (item->to));
+
+ const char *value_label = item->value_label;
+ if (value_label && value_label[0])
+ pivot_table_put2 (table, 1, old_value_idx,
+ pivot_value_new_user_text (value_label, -1));
+ }
+
+ pivot_table_submit (table);
+ }
+
+ /* Assign user-missing values.
+
+ User-missing values in the source variable(s) must be marked
+ as user-missing values in the destination variable. There
+ might be an arbitrary number of missing values, since the
+ source variable might have a range. Our sort function always
+ puts missing values together at the top of the range, so that
+ means that we can use a missing value range to cover all of
+ the user-missing values in any case (but we avoid it unless
+ necessary because user-missing value ranges are an obscure
+ feature). */
+ size_t n_missing = n_items;
+ for (size_t k = 0; k < n_items; k++)
+ if (!items[n_items - k - 1]->missing)
+ {
+ n_missing = k;
+ break;
+ }
+ if (n_missing > 0)
+ {
+ size_t lo = n_items - (n_missing - 1);
+ size_t hi = n_items;
+
+ struct missing_values mv;
+ mv_init (&mv, 0);
+ if (n_missing > 3)
+ mv_add_range (&mv, lo, hi);
+ else
+ for (size_t k = 0; k < n_missing; k++)
+ mv_add_num (&mv, lo + k);
+ var_set_missing_values (spec->dst, &mv);
+ mv_destroy (&mv);
+ }
+
+ /* Add value labels to the destination variable. */
+ for (j = 0; j < n_items; j++)
+ {
+ const char *value_label = items[j]->value_label;
+ if (value_label && value_label[0])
+ {
+ union value to_val = { .f = items[j]->to };
+ var_add_value_label (spec->dst, &to_val, value_label);
+ }
+ }
+
+ /* Free array. */
+ free (items);
+ }
+ add_transformation (ds, &autorecode_trns_class, arc);
+
+ for (size_t i = 0; i < n_dsts; i++)
+ free (dst_names[i]);
+ free (dst_names);
+ free (src_vars);
+
+ return ok ? CMD_SUCCESS : CMD_CASCADING_FAILURE;
+
+error:
+ for (size_t i = 0; i < n_dsts; i++)
+ free (dst_names[i]);
+ free (dst_names);
+ free (src_vars);
+ arc_free (arc);
+ return CMD_CASCADING_FAILURE;
+}
+
+static void
+arc_free (struct autorecode_pgm *arc)
+{
+ if (arc != NULL)
+ {
+ for (size_t i = 0; i < arc->n_specs; i++)
+ {
+ struct arc_spec *spec = &arc->specs[i];
+ struct arc_item *item, *next;
+
+ HMAP_FOR_EACH_SAFE (item, next, struct arc_item, hmap_node,
+ &spec->items->ht)
+ {
+ value_destroy (&item->from, item->width);
+ free (item->value_label);
+ hmap_delete (&spec->items->ht, &item->hmap_node);
+ free (item);
+ }
+ free (spec->label);
+ free (spec->src_name);
+ mv_destroy (&spec->mv);
+ }
+
+ size_t n_rec_items =
+ (arc->n_specs >= 2 && arc->specs[0].items == arc->specs[1].items
+ ? 1
+ : arc->n_specs);
+
+ for (size_t i = 0; i < n_rec_items; i++)
+ {
+ struct arc_spec *spec = &arc->specs[i];
+ hmap_destroy (&spec->items->ht);
+ free (spec->items);
+ }
+
+ free (arc->specs);
+ free (arc);
+ }
+}
+
+static struct arc_item *
+find_arc_item (const struct rec_items *items,
+ const union value *value, int width,
+ size_t hash)
+{
+ struct arc_item *item;
+
+ HMAP_FOR_EACH_WITH_HASH (item, struct arc_item, hmap_node, hash, &items->ht)
+ if (item->width == width && value_equal (value, &item->from, width))
+ return item;
+ return NULL;
+}
+
+static int
+compare_arc_items (const void *a_, const void *b_, const void *direction_)
+{
+ const struct arc_item *const *ap = a_;
+ const struct arc_item *const *bp = b_;
+ const struct arc_item *a = *ap;
+ const struct arc_item *b = *bp;
+
+ /* User-missing values always sort to the highest target values
+ (regardless of sort direction). */
+ if (a->missing != b->missing)
+ return a->missing < b->missing ? -1 : 1;
+
+ /* Otherwise, compare the data. */
+ int aw = a->width;
+ int bw = b->width;
+ int cmp;
+ if (aw == bw)
+ cmp = value_compare_3way (&a->from, &b->from, aw);
+ else
+ {
+ assert (aw && bw);
+ cmp = buf_compare_rpad (CHAR_CAST_BUG (const char *, a->from.s), aw,
+ CHAR_CAST_BUG (const char *, b->from.s), bw);
+ }
+
+ /* Then apply sort direction. */
+ const enum arc_direction *directionp = direction_;
+ enum arc_direction direction = *directionp;
+ return direction == ASCENDING ? cmp : -cmp;
+}
+
+static enum trns_result
+autorecode_trns_proc (void *arc_, struct ccase **c,
+ casenumber case_idx UNUSED)
+{
+ struct autorecode_pgm *arc = arc_;
+
+ *c = case_unshare (*c);
+ for (size_t i = 0; i < arc->n_specs; i++)
+ {
+ const struct arc_spec *spec = &arc->specs[i];
+ const union value *value = case_data_idx (*c, spec->src_idx);
+ int width = value_trim_spaces (value, spec->width);
+ size_t hash = value_hash (value, width, 0);
+ const struct arc_item *item = find_arc_item (spec->items, value, width,
+ hash);
+ *case_num_rw (*c, spec->dst) = item ? item->to : SYSMIS;
+ }
+
+ return TRNS_CONTINUE;
+}
+
+static bool
+autorecode_trns_free (void *arc_)
+{
+ struct autorecode_pgm *arc = arc_;
+
+ arc_free (arc);
+ return true;
+}
+
+static const struct trns_class autorecode_trns_class = {
+ .name = "AUTORECODE",
+ .execute = autorecode_trns_proc,
+ .destroy = autorecode_trns_free,
+};
projects / pspp / blobdiff