-/* PSPP - RANK. -*-c-*-
+/* PSPP - a program for statistical analysis.
+ Copyright (C) 2005, 2006, 2007 Free Software Foundation, Inc.
-Copyright (C) 2005 Free Software Foundation, Inc.
-Author: John Darrington 2005
+ 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 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 2 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.
-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, write to the Free Software
-Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
-02110-1301, USA. */
+ 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 "command.h"
-#include "dictionary.h"
-#include "sort.h"
-#include "variable.h"
+
+#include <limits.h>
+#include <math.h>
+
+#include <data/case-ordering.h>
+#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/missing-values.h>
+#include <data/procedure.h>
+#include <data/short-names.h>
+#include <data/variable.h>
+#include <language/command.h>
+#include <language/stats/sort-criteria.h>
+#include <libpspp/compiler.h>
+#include <libpspp/taint.h>
+#include <math/sort.h>
+#include <output/manager.h>
+#include <output/table.h>
+
+#include <gsl/gsl_cdf.h>
#include "gettext.h"
#define _(msgid) gettext (msgid)
+n=custom;
+savage=custom;
+print=print:!yes/no;
- +missing=miss:!exclude/include.
+ +fraction=fraction:!blom/tukey/vw/rankit;
+ +ties=ties:!mean/low/high/condense;
+ missing=miss:!exclude/include.
*/
/* (declarations) */
/* (functions) */
+typedef double (*rank_function_t) (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_proportion (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_normal (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_percent (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_rfraction (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_rank (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_n (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_savage (double c, double cc, double cc_1,
+ int i, double w);
+
+static double rank_ntiles (double c, double cc, double cc_1,
+ int i, double w);
enum RANK_FUNC
N,
NTILES,
SAVAGE,
+ n_RANK_FUNCS
+ };
+
+static const struct fmt_spec dest_format[n_RANK_FUNCS] = {
+ {FMT_F, 9, 3}, /* rank */
+ {FMT_F, 6, 4}, /* normal */
+ {FMT_F, 6, 2}, /* percent */
+ {FMT_F, 6, 4}, /* rfraction */
+ {FMT_F, 6, 4}, /* proportion */
+ {FMT_F, 6, 0}, /* n */
+ {FMT_F, 3, 0}, /* ntiles */
+ {FMT_F, 8, 4} /* savage */
+};
+
+static const char * const function_name[n_RANK_FUNCS] = {
+ "RANK",
+ "NORMAL",
+ "PERCENT",
+ "RFRACTION",
+ "PROPORTION",
+ "N",
+ "NTILES",
+ "SAVAGE"
+};
+
+static const rank_function_t rank_func[n_RANK_FUNCS] = {
+ rank_rank,
+ rank_normal,
+ rank_percent,
+ rank_rfraction,
+ rank_proportion,
+ rank_n,
+ rank_ntiles,
+ rank_savage
};
{
enum RANK_FUNC rfunc;
struct variable **destvars;
- struct variable *srcvar;
};
+/* Categories of missing values to exclude. */
+static enum mv_class exclude_values;
+
static struct rank_spec *rank_specs;
static size_t n_rank_specs;
-static struct sort_criteria *sc;
+static struct case_ordering *sc;
-static struct variable **group_vars;
+static const struct variable **group_vars;
static size_t n_group_vars;
+static const struct variable **src_vars;
+static size_t n_src_vars;
+
+
+static int k_ntiles;
+
static struct cmd_rank cmd;
+static void rank_sorted_file (struct casereader *,
+ struct casewriter *,
+ const struct dictionary *,
+ const struct rank_spec *rs,
+ int n_rank_specs,
+ int idx,
+ const struct variable *rank_var);
+static const char *
+fraction_name(void)
+{
+ static char name[10];
+ switch ( cmd.fraction )
+ {
+ case RANK_BLOM:
+ strcpy (name, "BLOM");
+ break;
+ case RANK_RANKIT:
+ strcpy (name, "RANKIT");
+ break;
+ case RANK_TUKEY:
+ strcpy (name, "TUKEY");
+ break;
+ case RANK_VW:
+ strcpy (name, "VW");
+ break;
+ default:
+ NOT_REACHED ();
+ }
+ return name;
+}
-int cmd_rank(void);
+/* Create a label on DEST_VAR, describing its derivation from SRC_VAR and F */
+static void
+create_var_label (struct variable *dest_var,
+ const struct variable *src_var, enum RANK_FUNC f)
+{
+ struct string label;
+ ds_init_empty (&label);
-int
-cmd_rank(void)
+ if ( n_group_vars > 0 )
+ {
+ struct string group_var_str;
+ int g;
+
+ ds_init_empty (&group_var_str);
+
+ for (g = 0 ; g < n_group_vars ; ++g )
+ {
+ if ( g > 0 ) ds_put_cstr (&group_var_str, " ");
+ ds_put_cstr (&group_var_str, var_get_name (group_vars[g]));
+ }
+
+ ds_put_format (&label, _("%s of %s by %s"), function_name[f],
+ var_get_name (src_var), ds_cstr (&group_var_str));
+ ds_destroy (&group_var_str);
+ }
+ else
+ ds_put_format (&label, _("%s of %s"),
+ function_name[f], var_get_name (src_var));
+
+ var_set_label (dest_var, ds_cstr (&label));
+
+ ds_destroy (&label);
+}
+
+
+static bool
+rank_cmd (struct dataset *ds, const struct case_ordering *sc,
+ const struct rank_spec *rank_specs, int n_rank_specs)
{
- size_t i;
- n_rank_specs = 0;
+ struct dictionary *d = dataset_dict (ds);
+ bool ok = true;
+ int i;
- if ( !parse_rank(&cmd) )
- return CMD_FAILURE;
+ for (i = 0 ; i < case_ordering_get_var_cnt (sc) ; ++i )
+ {
+ /* Rank variable at index I in SC. */
+ struct casegrouper *split_grouper;
+ struct casereader *split_group;
+ struct casewriter *output;
+
+ proc_discard_output (ds);
+ split_grouper = casegrouper_create_splits (proc_open (ds), d);
+ output = autopaging_writer_create (dict_get_next_value_idx (d));
+
+ while (casegrouper_get_next_group (split_grouper, &split_group))
+ {
+ struct case_ordering *ordering;
+ struct casereader *ordered;
+ struct casegrouper *by_grouper;
+ struct casereader *by_group;
+ int j;
+
+ /* Sort this split group by the BY variables as primary
+ keys and the rank variable as secondary key. */
+ ordering = case_ordering_create ();
+ for (j = 0; j < n_group_vars; j++)
+ case_ordering_add_var (ordering, group_vars[j], SRT_ASCEND);
+ case_ordering_add_var (ordering,
+ case_ordering_get_var (sc, i),
+ case_ordering_get_direction (sc, i));
+ ordered = sort_execute (split_group, ordering);
+
+ /* Rank the rank variable within this split group. */
+ by_grouper = casegrouper_create_vars (ordered,
+ group_vars, n_group_vars);
+ while (casegrouper_get_next_group (by_grouper, &by_group))
+ {
+ /* Rank the rank variable within this BY group
+ within the split group. */
+
+ rank_sorted_file (by_group, output, d, rank_specs, n_rank_specs,
+ i, src_vars[i]);
+ }
+ ok = casegrouper_destroy (by_grouper) && ok;
+ }
+ ok = casegrouper_destroy (split_grouper);
+ ok = proc_commit (ds) && ok;
+ ok = (proc_set_active_file_data (ds, casewriter_make_reader (output))
+ && ok);
+ if (!ok)
+ break;
+ }
+
+ return ok;
+}
-#if 1
- for (i = 0 ; i < sc->crit_cnt ; ++i )
+/* Hardly a rank function !! */
+static double
+rank_n (double c UNUSED, double cc UNUSED, double cc_1 UNUSED,
+ int i UNUSED, double w)
+{
+ return w;
+}
+
+
+static double
+rank_rank (double c, double cc, double cc_1,
+ int i, double w UNUSED)
+{
+ double rank;
+
+ if ( c >= 1.0 )
{
- struct sort_criterion *crit = &sc->crits[i];
-
- printf("Dir: %d; Index: %d\n", crit->dir, crit->fv);
+ switch (cmd.ties)
+ {
+ case RANK_LOW:
+ rank = cc_1 + 1;
+ break;
+ case RANK_HIGH:
+ rank = cc;
+ break;
+ case RANK_MEAN:
+ rank = cc_1 + (c + 1.0)/ 2.0;
+ break;
+ case RANK_CONDENSE:
+ rank = i;
+ break;
+ default:
+ NOT_REACHED ();
+ }
+ }
+ else
+ {
+ switch (cmd.ties)
+ {
+ case RANK_LOW:
+ rank = cc_1;
+ break;
+ case RANK_HIGH:
+ rank = cc;
+ break;
+ case RANK_MEAN:
+ rank = cc_1 + c / 2.0 ;
+ break;
+ case RANK_CONDENSE:
+ rank = i;
+ break;
+ default:
+ NOT_REACHED ();
+ }
}
- for (i = 0 ; i < n_group_vars ; ++i )
- printf("Group var: %s\n",group_vars[0]->name);
+ return rank;
+}
- for (i = 0 ; i < n_rank_specs ; ++i )
+
+static double
+rank_rfraction (double c, double cc, double cc_1,
+ int i, double w)
+{
+ return rank_rank (c, cc, cc_1, i, w) / w ;
+}
+
+
+static double
+rank_percent (double c, double cc, double cc_1,
+ int i, double w)
+{
+ return rank_rank (c, cc, cc_1, i, w) * 100.0 / w ;
+}
+
+
+static double
+rank_proportion (double c, double cc, double cc_1,
+ int i, double w)
+{
+ const double r = rank_rank (c, cc, cc_1, i, w) ;
+
+ double f;
+
+ switch ( cmd.fraction )
+ {
+ case RANK_BLOM:
+ f = (r - 3.0/8.0) / (w + 0.25);
+ break;
+ case RANK_RANKIT:
+ f = (r - 0.5) / w ;
+ break;
+ case RANK_TUKEY:
+ f = (r - 1.0/3.0) / (w + 1.0/3.0);
+ break;
+ case RANK_VW:
+ f = r / ( w + 1.0);
+ break;
+ default:
+ NOT_REACHED ();
+ }
+
+
+ return (f > 0) ? f : SYSMIS;
+}
+
+static double
+rank_normal (double c, double cc, double cc_1,
+ int i, double w)
+{
+ double f = rank_proportion (c, cc, cc_1, i, w);
+
+ return gsl_cdf_ugaussian_Pinv (f);
+}
+
+static double
+rank_ntiles (double c, double cc, double cc_1,
+ int i, double w)
+{
+ double r = rank_rank (c, cc, cc_1, i, w);
+
+
+ return ( floor (( r * k_ntiles) / ( w + 1) ) + 1);
+}
+
+/* Expected value of the order statistics from an exponential distribution */
+static double
+ee (int j, double w_star)
+{
+ int k;
+ double sum = 0.0;
+
+ for (k = 1 ; k <= j; k++)
+ sum += 1.0 / ( w_star + 1 - k );
+
+ return sum;
+}
+
+
+static double
+rank_savage (double c, double cc, double cc_1,
+ int i UNUSED, double w)
+{
+ double int_part;
+ const int i_1 = floor (cc_1);
+ const int i_2 = floor (cc);
+
+ const double w_star = (modf (w, &int_part) == 0 ) ? w : floor (w) + 1;
+
+ const double g_1 = cc_1 - i_1;
+ const double g_2 = cc - i_2;
+
+ /* The second factor is infinite, when the first is zero.
+ Therefore, evaluate the second, only when the first is non-zero */
+ const double expr1 = (1 - g_1) ? (1 - g_1) * ee(i_1+1, w_star) : ( 1 - g_1);
+ const double expr2 = g_2 ? g_2 * ee (i_2+1, w_star) : g_2 ;
+
+ if ( i_1 == i_2 )
+ return ee (i_1 + 1, w_star) - 1;
+
+ if ( i_1 + 1 == i_2 )
+ return ( ( expr1 + expr2 )/c ) - 1;
+
+ if ( i_1 + 2 <= i_2 )
{
int j;
- printf("Ranks spec %d; Func: %d\n",i, rank_specs[i].rfunc);
-
- for (j=0; j < sc->crit_cnt ; ++j )
- printf("Dest var is \"%s\"\n", rank_specs[i].destvars[j]->name);
+ double sigma = 0.0;
+ for (j = i_1 + 2 ; j <= i_2; ++j )
+ sigma += ee (j, w_star);
+ return ( (expr1 + expr2 + sigma) / c) -1;
}
-#endif
+ NOT_REACHED();
+}
- free(group_vars);
-
- for (i = 0 ; i < n_rank_specs ; ++i )
+static void
+rank_sorted_file (struct casereader *input,
+ struct casewriter *output,
+ const struct dictionary *dict,
+ const struct rank_spec *rs,
+ int n_rank_specs,
+ int dest_idx,
+ const struct variable *rank_var)
+{
+ struct casereader *pass1, *pass2, *pass2_1;
+ struct casegrouper *tie_grouper;
+ struct ccase c;
+ double w = 0.0;
+ double cc = 0.0;
+ int tie_group = 1;
+
+
+ input = casereader_create_filter_missing (input, &rank_var, 1,
+ exclude_values, output);
+ input = casereader_create_filter_weight (input, dict, NULL, output);
+
+ casereader_split (input, &pass1, &pass2);
+
+ /* Pass 1: Get total group weight. */
+ for (; casereader_read (pass1, &c); case_destroy (&c))
+ w += dict_get_case_weight (dict, &c, NULL);
+ casereader_destroy (pass1);
+
+ /* Pass 2: Do ranking. */
+ tie_grouper = casegrouper_create_vars (pass2, &rank_var, 1);
+ while (casegrouper_get_next_group (tie_grouper, &pass2_1))
+ {
+ struct casereader *pass2_2;
+ double cc_1 = cc;
+ double tw = 0.0;
+ int i;
+
+ pass2_2 = casereader_clone (pass2_1);
+ taint_propagate (casereader_get_taint (pass2_2),
+ casewriter_get_taint (output));
+
+ /* Pass 2.1: Sum up weight for tied cases. */
+ for (; casereader_read (pass2_1, &c); case_destroy (&c))
+ tw += dict_get_case_weight (dict, &c, NULL);
+ cc += tw;
+ casereader_destroy (pass2_1);
+
+ /* Pass 2.2: Rank tied cases. */
+ while (casereader_read (pass2_2, &c))
+ {
+ for (i = 0; i < n_rank_specs; ++i)
+ {
+ const struct variable *dst_var = rs[i].destvars[dest_idx];
+ double *dst_value = &case_data_rw (&c, dst_var)->f;
+ *dst_value = rank_func[rs[i].rfunc] (tw, cc, cc_1, tie_group, w);
+ }
+ casewriter_write (output, &c);
+ }
+ casereader_destroy (pass2_2);
+
+ tie_group++;
+ }
+ casegrouper_destroy (tie_grouper);
+}
+
+/* Transformation function to enumerate all the cases */
+static int
+create_resort_key (void *key_var_, struct ccase *cc, casenumber case_num)
+{
+ struct variable *key_var = key_var_;
+
+ case_data_rw(cc, key_var)->f = case_num;
+
+ return TRNS_CONTINUE;
+}
+
+
+/* Create and return a new variable in which to store the ranks of SRC_VAR
+ accoring to the rank function F.
+ VNAME is the name of the variable to be created.
+ If VNAME is NULL, then a name will be automatically chosen.
+ */
+static struct variable *
+create_rank_variable (struct dictionary *dict, enum RANK_FUNC f,
+ const struct variable *src_var,
+ const char *vname)
+{
+ int i;
+ struct variable *var = NULL;
+ char name[SHORT_NAME_LEN + 1];
+
+ if ( vname )
+ var = dict_create_var(dict, vname, 0);
+
+ if ( NULL == var )
{
- free(rank_specs[i].destvars);
+ snprintf (name, SHORT_NAME_LEN + 1, "%c%s",
+ function_name[f][0], var_get_name (src_var));
+
+ var = dict_create_var(dict, name, 0);
}
-
- free(rank_specs);
- sort_destroy_criteria(sc);
+ i = 1;
+ while( NULL == var )
+ {
+ char func_abb[4];
+ snprintf(func_abb, 4, "%s", function_name[f]);
+ snprintf(name, SHORT_NAME_LEN + 1, "%s%03d", func_abb,
+ i);
+
+ var = dict_create_var(dict, name, 0);
+ if (i++ >= 999)
+ break;
+ }
- return CMD_SUCCESS;
+ i = 1;
+ while ( NULL == var )
+ {
+ char func_abb[3];
+ snprintf(func_abb, 3, "%s", function_name[f]);
+
+ snprintf(name, SHORT_NAME_LEN + 1,
+ "RNK%s%02d", func_abb, i);
+
+ var = dict_create_var(dict, name, 0);
+ if ( i++ >= 99 )
+ break;
+ }
+
+ if ( NULL == var )
+ {
+ msg(ME, _("Cannot create new rank variable. All candidates in use."));
+ return NULL;
+ }
+
+ var_set_both_formats (var, &dest_format[f]);
+
+ return var;
}
+static void
+rank_cleanup(void)
+{
+ int i;
-/* Parser for the variables sub command
- Returns 1 on success */
-static int
-rank_custom_variables(struct cmd_rank *cmd UNUSED)
+ free (group_vars);
+ group_vars = NULL;
+ n_group_vars = 0;
+
+ for (i = 0 ; i < n_rank_specs ; ++i )
+ free (rank_specs[i].destvars);
+
+ free (rank_specs);
+ rank_specs = NULL;
+ n_rank_specs = 0;
+
+ case_ordering_destroy (sc);
+ sc = NULL;
+
+ free (src_vars);
+ src_vars = NULL;
+ n_src_vars = 0;
+}
+
+int
+cmd_rank (struct lexer *lexer, struct dataset *ds)
{
- static const int terminators[2] = {T_BY, 0};
+ bool result;
+ struct variable *order;
+ size_t i;
+ n_rank_specs = 0;
- lex_match('=');
+ if ( !parse_rank (lexer, ds, &cmd, NULL) )
+ {
+ rank_cleanup ();
+ return CMD_FAILURE;
+ }
- if ((token != T_ID || dict_lookup_var (default_dict, tokid) == NULL)
- && token != T_ALL)
- return 2;
+ /* If /MISSING=INCLUDE is set, then user missing values are ignored */
+ exclude_values = cmd.miss == RANK_INCLUDE ? MV_SYSTEM : MV_ANY;
+
+ /* Default to /RANK if no function subcommands are given */
+ if ( !( cmd.sbc_normal || cmd.sbc_ntiles || cmd.sbc_proportion ||
+ cmd.sbc_rfraction || cmd.sbc_savage || cmd.sbc_n ||
+ cmd.sbc_percent || cmd.sbc_rank ) )
+ {
+ assert ( n_rank_specs == 0 );
+
+ rank_specs = xmalloc (sizeof (*rank_specs));
+ rank_specs[0].rfunc = RANK;
+ rank_specs[0].destvars =
+ xcalloc (case_ordering_get_var_cnt (sc), sizeof (struct variable *));
+
+ n_rank_specs = 1;
+ }
- sc = sort_parse_criteria (default_dict, 0, 0, 0, terminators);
+ assert ( case_ordering_get_var_cnt (sc) == n_src_vars);
- if ( lex_match(T_BY) )
+ /* Create variables for all rank destinations which haven't
+ already been created with INTO.
+ Add labels to all the destination variables.
+ */
+ for (i = 0 ; i < n_rank_specs ; ++i )
{
- if ((token != T_ID || dict_lookup_var (default_dict, tokid) == NULL))
+ int v;
+ for ( v = 0 ; v < n_src_vars ; v ++ )
{
- return 2;
+ if ( rank_specs[i].destvars[v] == NULL )
+ {
+ rank_specs[i].destvars[v] =
+ create_rank_variable (dataset_dict(ds), rank_specs[i].rfunc, src_vars[v], NULL);
+ }
+
+ create_var_label ( rank_specs[i].destvars[v],
+ src_vars[v],
+ rank_specs[i].rfunc);
}
+ }
- if (!parse_variables (default_dict, &group_vars, &n_group_vars,
- PV_NO_DUPLICATE | PV_NUMERIC | PV_NO_SCRATCH) )
+ if ( cmd.print == RANK_YES )
+ {
+ int v;
+
+ tab_output_text (0, _("Variables Created By RANK"));
+ tab_output_text (0, "\n");
+
+ for (i = 0 ; i < n_rank_specs ; ++i )
{
- free (group_vars);
- return 0;
+ for ( v = 0 ; v < n_src_vars ; v ++ )
+ {
+ if ( n_group_vars > 0 )
+ {
+ struct string varlist;
+ int g;
+
+ ds_init_empty (&varlist);
+ for ( g = 0 ; g < n_group_vars ; ++g )
+ {
+ ds_put_cstr (&varlist, var_get_name (group_vars[g]));
+
+ if ( g < n_group_vars - 1)
+ ds_put_cstr (&varlist, " ");
+ }
+
+ if ( rank_specs[i].rfunc == NORMAL ||
+ rank_specs[i].rfunc == PROPORTION )
+ tab_output_text (TAT_PRINTF,
+ _("%s into %s(%s of %s using %s BY %s)"),
+ var_get_name (src_vars[v]),
+ var_get_name (rank_specs[i].destvars[v]),
+ function_name[rank_specs[i].rfunc],
+ var_get_name (src_vars[v]),
+ fraction_name(),
+ ds_cstr (&varlist)
+ );
+
+ else
+ tab_output_text (TAT_PRINTF,
+ _("%s into %s(%s of %s BY %s)"),
+ var_get_name (src_vars[v]),
+ var_get_name (rank_specs[i].destvars[v]),
+ function_name[rank_specs[i].rfunc],
+ var_get_name (src_vars[v]),
+ ds_cstr (&varlist)
+ );
+ ds_destroy (&varlist);
+ }
+ else
+ {
+ if ( rank_specs[i].rfunc == NORMAL ||
+ rank_specs[i].rfunc == PROPORTION )
+ tab_output_text (TAT_PRINTF,
+ _("%s into %s(%s of %s using %s)"),
+ var_get_name (src_vars[v]),
+ var_get_name (rank_specs[i].destvars[v]),
+ function_name[rank_specs[i].rfunc],
+ var_get_name (src_vars[v]),
+ fraction_name()
+ );
+
+ else
+ tab_output_text (TAT_PRINTF,
+ _("%s into %s(%s of %s)"),
+ var_get_name (src_vars[v]),
+ var_get_name (rank_specs[i].destvars[v]),
+ function_name[rank_specs[i].rfunc],
+ var_get_name (src_vars[v])
+ );
+ }
+ }
}
}
- return 1;
+ if ( cmd.sbc_fraction &&
+ ( ! cmd.sbc_normal && ! cmd.sbc_proportion) )
+ msg(MW, _("FRACTION has been specified, but NORMAL and PROPORTION rank functions have not been requested. The FRACTION subcommand will be ignored.") );
+
+ /* Add a variable which we can sort by to get back the original
+ order */
+ order = dict_create_var_assert (dataset_dict (ds), "$ORDER_", 0);
+
+ add_transformation (ds, create_resort_key, 0, order);
+
+ /* Do the ranking */
+ result = rank_cmd (ds, sc, rank_specs, n_rank_specs);
+
+ /* Put the active file back in its original order. Delete
+ our sort key, which we don't need anymore. */
+ {
+ struct case_ordering *ordering = case_ordering_create ();
+ struct casereader *sorted;
+ case_ordering_add_var (ordering, order, SRT_ASCEND);
+ /* FIXME: loses error conditions. */
+ proc_discard_output (ds);
+ sorted = sort_execute (proc_open (ds), ordering);
+ result = proc_commit (ds) && result;
+
+ dict_delete_var (dataset_dict (ds), order);
+ result = proc_set_active_file_data (ds, sorted) && result;
+ }
+
+ rank_cleanup();
+
+
+ return (result ? CMD_SUCCESS : CMD_CASCADING_FAILURE);
}
-/* Return a name for a new variable which ranks the variable VAR_NAME,
- according to the ranking function F.
- If IDX is non zero, then IDX is used as a disambiguating number.
- FIXME: This is not very robust.
-*/
-static char *
-new_variable_name(const char *ranked_var_name, enum RANK_FUNC f, int idx)
+/* Parser for the variables sub command
+ Returns 1 on success */
+static int
+rank_custom_variables (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd UNUSED, void *aux UNUSED)
{
- static char new_name[SHORT_NAME_LEN + 1];
- char temp[SHORT_NAME_LEN + 1];
-
- if ( idx == 0 )
- {
- switch (f)
- {
- case RANK:
- case RFRACTION:
- strcpy(new_name,"R");
- break;
+ lex_match (lexer, '=');
- case NORMAL:
- case N:
- case NTILES:
- strcpy(new_name,"N");
- break;
-
- case PERCENT:
- case PROPORTION:
- strcpy(new_name,"P");
- break;
+ if ((lex_token (lexer) != T_ID || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) == NULL)
+ && lex_token (lexer) != T_ALL)
+ return 2;
- case SAVAGE:
- strcpy(new_name,"S");
- break;
+ sc = parse_case_ordering (lexer, dataset_dict (ds), NULL);
+ if (sc == NULL)
+ return 0;
+ case_ordering_get_vars (sc, &src_vars, &n_src_vars);
- default:
- assert(false);
- break;
- }
-
- strncat(new_name, ranked_var_name, 7);
- }
- else
+ if ( lex_match (lexer, T_BY) )
{
- strncpy(temp, ranked_var_name, 3);
- snprintf(new_name, SHORT_NAME_LEN, "%s%03d", temp, idx);
+ if ((lex_token (lexer) != T_ID || dict_lookup_var (dataset_dict (ds), lex_tokid (lexer)) == NULL))
+ {
+ return 2;
+ }
+
+ if (!parse_variables_const (lexer, dataset_dict (ds),
+ &group_vars, &n_group_vars,
+ PV_NO_DUPLICATE | PV_NO_SCRATCH) )
+ {
+ free (group_vars);
+ return 0;
+ }
}
- return new_name;
+ return 1;
}
+
/* Parse the [/rank INTO var1 var2 ... varN ] clause */
static int
-parse_rank_function(struct cmd_rank *cmd UNUSED, enum RANK_FUNC f)
+parse_rank_function (struct lexer *lexer, struct dictionary *dict, struct cmd_rank *cmd UNUSED, enum RANK_FUNC f)
{
- static const struct fmt_spec f8_2 = {FMT_F, 8, 2};
int var_count = 0;
-
+
n_rank_specs++;
rank_specs = xnrealloc(rank_specs, n_rank_specs, sizeof *rank_specs);
rank_specs[n_rank_specs - 1].rfunc = f;
+ rank_specs[n_rank_specs - 1].destvars = NULL;
- rank_specs[n_rank_specs - 1].destvars =
- xcalloc (sc->crit_cnt, sizeof (struct variable *));
-
- if (lex_match_id("INTO"))
+ rank_specs[n_rank_specs - 1].destvars =
+ xcalloc (case_ordering_get_var_cnt (sc),
+ sizeof (struct variable *));
+
+ if (lex_match_id (lexer, "INTO"))
{
struct variable *destvar;
- while( token == T_ID )
+ while( lex_token (lexer) == T_ID )
{
- ++var_count;
- if ( dict_lookup_var (default_dict, tokid) != NULL )
+
+ if ( dict_lookup_var (dict, lex_tokid (lexer)) != NULL )
{
- msg(ME, _("Variable %s already exists."), tokid);
+ msg(SE, _("Variable %s already exists."), lex_tokid (lexer));
return 0;
}
- if ( var_count > sc->crit_cnt )
+ if ( var_count >= case_ordering_get_var_cnt (sc) )
{
- msg(ME, _("Too many variables in INTO clause."));
+ msg(SE, _("Too many variables in INTO clause."));
return 0;
}
- destvar = dict_create_var (default_dict, tokid, 0);
- if ( destvar )
- {
- destvar->print = destvar->write = f8_2;
- }
-
- rank_specs[n_rank_specs - 1].destvars[var_count - 1] = destvar ;
+ destvar = create_rank_variable (dict, f, src_vars[var_count], lex_tokid (lexer));
+ rank_specs[n_rank_specs - 1].destvars[var_count] = destvar ;
- lex_get();
-
+ lex_get (lexer);
+ ++var_count;
}
}
- /* Allocate rank variable names to all those which haven't had INTO
- variables assigned */
- while (var_count < sc->crit_cnt)
- {
- static int idx=0;
- struct variable *destvar ;
- const struct variable *v = dict_get_var(default_dict,
- sc->crits[var_count].fv);
-
- char *new_name;
-
- do {
- new_name = new_variable_name(v->name, f, idx);
-
- destvar = dict_create_var (default_dict, new_name, 0);
- if (!destvar )
- ++idx;
-
- } while( !destvar ) ;
-
- destvar->print = destvar->write = f8_2;
-
- rank_specs[n_rank_specs - 1].destvars[var_count] = destvar ;
-
- ++var_count;
- }
-
return 1;
}
static int
-rank_custom_rank(struct cmd_rank *cmd )
+rank_custom_rank (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, RANK);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, RANK);
}
static int
-rank_custom_normal(struct cmd_rank *cmd )
+rank_custom_normal (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, NORMAL);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, NORMAL);
}
static int
-rank_custom_percent(struct cmd_rank *cmd )
+rank_custom_percent (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, NORMAL);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, PERCENT);
}
static int
-rank_custom_rfraction(struct cmd_rank *cmd )
+rank_custom_rfraction (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, RFRACTION);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, RFRACTION);
}
static int
-rank_custom_proportion(struct cmd_rank *cmd )
+rank_custom_proportion (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, PROPORTION);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, PROPORTION);
}
static int
-rank_custom_n(struct cmd_rank *cmd )
+rank_custom_n (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, N);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, N);
}
static int
-rank_custom_savage(struct cmd_rank *cmd )
+rank_custom_savage (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- return parse_rank_function(cmd, SAVAGE);
+ struct dictionary *dict = dataset_dict (ds);
+
+ return parse_rank_function (lexer, dict, cmd, SAVAGE);
}
static int
-rank_custom_ntiles(struct cmd_rank *cmd )
+rank_custom_ntiles (struct lexer *lexer, struct dataset *ds, struct cmd_rank *cmd, void *aux UNUSED )
{
- if ( lex_force_match('(') )
+ struct dictionary *dict = dataset_dict (ds);
+
+ if ( lex_force_match (lexer, '(') )
{
- if ( lex_force_int() )
+ if ( lex_force_int (lexer) )
{
- lex_get();
- lex_force_match(')');
+ k_ntiles = lex_integer (lexer);
+ lex_get (lexer);
+ lex_force_match (lexer, ')');
}
else
return 0;
else
return 0;
- return parse_rank_function(cmd, NTILES);
+ return parse_rank_function (lexer, dict, cmd, NTILES);
}
-
+/*
+ Local Variables:
+ mode: c
+ End:
+*/