X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Fdescriptives.c;h=eefd0598e31a6dca612c84ebd5339a77ef7eb05b;hb=refs%2Fheads%2Fmatrix2;hp=32a979d205fa413853c447c848f851f78b5f9752;hpb=5a1090431ff8f38e3983345d0924f614efff1a73;p=pspp diff --git a/src/language/stats/descriptives.c b/src/language/stats/descriptives.c index 32a979d205..eefd0598e3 100644 --- a/src/language/stats/descriptives.c +++ b/src/language/stats/descriptives.c @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000, 2009, 2010, 2011 Free Software Foundation, Inc. + Copyright (C) 1997-2000, 2009-2014 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 @@ -16,12 +16,14 @@ #include +#include #include #include #include #include "data/casegrouper.h" #include "data/casereader.h" +#include "data/casewriter.h" #include "data/dataset.h" #include "data/dictionary.h" #include "data/transformations.h" @@ -36,7 +38,7 @@ #include "libpspp/i18n.h" #include "libpspp/message.h" #include "math/moments.h" -#include "output/tab.h" +#include "output/pivot-table.h" #include "gl/xalloc.h" @@ -74,6 +76,10 @@ struct dsc_trns size_t var_cnt; /* Number of variables. */ enum dsc_missing_type missing_type; /* Treatment of missing values. */ enum mv_class exclude; /* Classes of missing values to exclude. */ + const struct variable *filter; /* Dictionary FILTER BY variable. */ + struct casereader *z_reader; /* Reader for count, mean, stddev. */ + casenumber count; /* Number left in this SPLIT FILE group.*/ + bool ok; }; /* Statistics. Used as bit indexes, so must be 32 or fewer. */ @@ -130,26 +136,17 @@ struct dsc_var double stats[DSC_N_STATS]; /* All the stats' values. */ }; -/* Output format. */ -enum dsc_format - { - DSC_LINE, /* Abbreviated format. */ - DSC_SERIAL /* Long format. */ - }; - /* A DESCRIPTIVES procedure. */ struct dsc_proc { /* Per-variable info. */ + struct dictionary *dict; /* Dictionary. */ struct dsc_var *vars; /* Variables. */ size_t var_cnt; /* Number of variables. */ /* User options. */ enum dsc_missing_type missing_type; /* Treatment of missing values. */ enum mv_class exclude; /* Classes of missing values to exclude. */ - int show_var_labels; /* Nonzero to show variable labels. */ - int show_index; /* Nonzero to show variable index. */ - enum dsc_format format; /* Output format. */ /* Accumulated results. */ double missing_listwise; /* Sum of weights of cases missing listwise. */ @@ -160,6 +157,9 @@ struct dsc_proc unsigned long show_stats; /* Statistics to display. */ unsigned long calc_stats; /* Statistics to calculate. */ enum moment max_moment; /* Highest moment needed for stats. */ + + /* Z scores. */ + struct casewriter *z_writer; /* Mean and stddev per SPLIT FILE group. */ }; /* Parsing. */ @@ -200,19 +200,18 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) /* Create and initialize dsc. */ dsc = xmalloc (sizeof *dsc); + dsc->dict = dict; dsc->vars = NULL; dsc->var_cnt = 0; dsc->missing_type = DSC_VARIABLE; dsc->exclude = MV_ANY; - dsc->show_var_labels = 1; - dsc->show_index = 0; - dsc->format = DSC_LINE; dsc->missing_listwise = 0.; dsc->valid = 0.; dsc->bad_warn = 1; dsc->sort_by_stat = DSC_NONE; dsc->sort_ascending = 1; dsc->show_stats = dsc->calc_stats = DEFAULT_STATS; + dsc->z_writer = NULL; /* Parse DESCRIPTIVES. */ while (lex_token (lexer) != T_ENDCMD) @@ -243,18 +242,15 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) lex_match (lexer, T_EQUALS); while (lex_token (lexer) != T_ENDCMD && lex_token (lexer) != T_SLASH) { - if (lex_match_id (lexer, "LABELS")) - dsc->show_var_labels = 1; - else if (lex_match_id (lexer, "NOLABELS")) - dsc->show_var_labels = 0; - else if (lex_match_id (lexer, "INDEX")) - dsc->show_index = 1; - else if (lex_match_id (lexer, "NOINDEX")) - dsc->show_index = 0; - else if (lex_match_id (lexer, "LINE")) - dsc->format = DSC_LINE; - else if (lex_match_id (lexer, "SERIAL")) - dsc->format = DSC_SERIAL; + if (lex_match_id (lexer, "LABELS") + || lex_match_id (lexer, "NOLABELS") + || lex_match_id (lexer, "INDEX") + || lex_match_id (lexer, "NOINDEX") + || lex_match_id (lexer, "LINE") + || lex_match_id (lexer, "SERIAL")) + { + /* Ignore. */ + } else { lex_error (lexer, NULL); @@ -274,7 +270,15 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) else if (lex_match_id (lexer, "DEFAULT")) dsc->show_stats |= DEFAULT_STATS; else - dsc->show_stats |= 1ul << (match_statistic (lexer)); + { + enum dsc_statistic s = match_statistic (lexer); + if (s == DSC_NONE) + { + lex_error (lexer, NULL); + goto error; + } + dsc->show_stats |= 1ul << s; + } lex_match (lexer, T_COMMA); } if (dsc->show_stats == 0) @@ -288,7 +292,7 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) else { dsc->sort_by_stat = match_statistic (lexer); - if (dsc->sort_by_stat == DSC_NONE ) + if (dsc->sort_by_stat == DSC_NONE) dsc->sort_by_stat = DSC_MEAN; } if (lex_match (lexer, T_LPAREN)) @@ -299,7 +303,8 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) dsc->sort_ascending = 0; else lex_error (lexer, NULL); - lex_force_match (lexer, T_RPAREN); + if (! lex_force_match (lexer, T_RPAREN)) + goto error; } } else if (var_cnt == 0) @@ -367,6 +372,8 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) /* Construct z-score varnames, show translation table. */ if (z_cnt || save_z_scores) { + struct caseproto *proto; + if (save_z_scores) { int gen_cnt = 0; @@ -386,6 +393,20 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) } } } + + /* It would be better to handle Z scores correctly (however we define + that) when TEMPORARY is in effect, but in the meantime this at least + prevents a use-after-free error. See bug #38786. */ + if (proc_make_temporary_transformations_permanent (ds)) + msg (SW, _("DESCRIPTIVES with Z scores ignores TEMPORARY. " + "Temporary transformations will be made permanent.")); + + proto = caseproto_create (); + for (i = 0; i < 1 + 2 * z_cnt; i++) + proto = caseproto_add_width (proto, 0); + dsc->z_writer = autopaging_writer_create (proto); + caseproto_unref (proto); + dump_z_table (dsc); } @@ -417,7 +438,7 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) dsc->vars[i].moments = moments_create (dsc->max_moment); /* Data pass. */ - grouper = casegrouper_create_splits (proc_open (ds), dict); + grouper = casegrouper_create_splits (proc_open_filtering (ds, false), dict); while (casegrouper_get_next_group (grouper, &group)) calc_descriptives (dsc, group, ds); ok = casegrouper_destroy (grouper); @@ -475,6 +496,7 @@ free_dsc_proc (struct dsc_proc *dsc) free (dsc_var->z_name); moments_destroy (dsc_var->moments); } + casewriter_destroy (dsc->z_writer); free (dsc->vars); free (dsc); } @@ -494,7 +516,7 @@ try_name (const struct dictionary *dict, struct dsc_proc *dsc, for (i = 0; i < dsc->var_cnt; i++) { struct dsc_var *dsc_var = &dsc->vars[i]; - if (dsc_var->z_name != NULL && !strcasecmp (dsc_var->z_name, name)) + if (dsc_var->z_name != NULL && !utf8_strcasecmp (dsc_var->z_name, name)) return false; } return true; @@ -522,7 +544,7 @@ generate_z_varname (const struct dictionary *dict, struct dsc_proc *dsc, /* Generate a synthetic name. */ for (;;) { - char name[8]; + char name[16]; (*z_cnt)++; @@ -553,37 +575,38 @@ generate_z_varname (const struct dictionary *dict, struct dsc_proc *dsc, static void dump_z_table (struct dsc_proc *dsc) { - size_t cnt = 0; - struct tab_table *t; + struct pivot_table *table = pivot_table_create ( + N_("Mapping of Variables to Z-scores")); - { - size_t i; + pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Names"), + N_("Source"), N_("Target")); - for (i = 0; i < dsc->var_cnt; i++) - if (dsc->vars[i].z_name != NULL) - cnt++; - } - - t = tab_create (2, cnt + 1); - tab_title (t, _("Mapping of variables to corresponding Z-scores.")); - tab_headers (t, 0, 0, 1, 0); - tab_box (t, TAL_1, TAL_1, TAL_0, TAL_1, 0, 0, 1, cnt); - tab_hline (t, TAL_2, 0, 1, 1); - tab_text (t, 0, 0, TAB_CENTER | TAT_TITLE, _("Source")); - tab_text (t, 1, 0, TAB_CENTER | TAT_TITLE, _("Target")); + struct pivot_dimension *names = pivot_dimension_create ( + table, PIVOT_AXIS_ROW, N_("Variables")); + names->hide_all_labels = true; - { - size_t i, y; + for (size_t i = 0; i < dsc->var_cnt; i++) + if (dsc->vars[i].z_name != NULL) + { + int row = pivot_category_create_leaf (names->root, + pivot_value_new_number (i)); - for (i = 0, y = 1; i < dsc->var_cnt; i++) - if (dsc->vars[i].z_name != NULL) - { - tab_text (t, 0, y, TAB_LEFT, var_to_string (dsc->vars[i].v)); - tab_text (t, 1, y++, TAB_LEFT, dsc->vars[i].z_name); - } - } + pivot_table_put2 (table, 0, row, + pivot_value_new_variable (dsc->vars[i].v)); + pivot_table_put2 (table, 1, row, + pivot_value_new_user_text (dsc->vars[i].z_name, -1)); + } + + pivot_table_submit (table); +} - tab_submit (t); +static void +descriptives_set_all_sysmis_zscores (const struct dsc_trns *t, struct ccase *c) +{ + const struct dsc_z_score *z; + + for (z = t->z_scores; z < t->z_scores + t->z_score_cnt; z++) + *case_num_rw (c, z->z_var) = SYSMIS; } /* Transformation function to calculate Z-scores. Will return SYSMIS if any of @@ -599,7 +622,50 @@ descriptives_trns_proc (void *trns_, struct ccase **c, struct dsc_trns *t = trns_; struct dsc_z_score *z; const struct variable **vars; - int all_sysmis = 0; + + *c = case_unshare (*c); + + if (t->filter) + { + double f = case_num (*c, t->filter); + if (f == 0.0 || var_is_num_missing (t->filter, f, MV_ANY)) + { + descriptives_set_all_sysmis_zscores (t, *c); + return TRNS_CONTINUE; + } + } + + if (t->count <= 0) + { + struct ccase *z_case; + + z_case = casereader_read (t->z_reader); + if (z_case) + { + size_t z_idx = 0; + + t->count = case_num_idx (z_case, z_idx++); + for (z = t->z_scores; z < t->z_scores + t->z_score_cnt; z++) + { + z->mean = case_num_idx (z_case, z_idx++); + z->std_dev = case_num_idx (z_case, z_idx++); + } + case_unref (z_case); + } + else + { + if (t->ok) + { + msg (SE, _("Internal error processing Z scores. " + "Please report this to %s."), + PACKAGE_BUGREPORT); + t->ok = false; + } + descriptives_set_all_sysmis_zscores (t, *c); + return TRNS_CONTINUE; + } + } + t->count--; if (t->missing_type == DSC_LISTWISE) { @@ -609,19 +675,18 @@ descriptives_trns_proc (void *trns_, struct ccase **c, double score = case_num (*c, *vars); if (var_is_num_missing (*vars, score, t->exclude)) { - all_sysmis = 1; - break; + descriptives_set_all_sysmis_zscores (t, *c); + return TRNS_CONTINUE; } } } - *c = case_unshare (*c); for (z = t->z_scores; z < t->z_scores + t->z_score_cnt; z++) { double input = case_num (*c, z->src_var); - double *output = &case_data_rw (*c, z->z_var)->f; + double *output = case_num_rw (*c, z->z_var); - if (z->mean == SYSMIS || z->std_dev == SYSMIS || all_sysmis + if (z->mean == SYSMIS || z->std_dev == SYSMIS || var_is_num_missing (z->src_var, input, t->exclude)) *output = SYSMIS; else @@ -635,11 +700,15 @@ static bool descriptives_trns_free (void *trns_) { struct dsc_trns *t = trns_; + bool ok = t->ok && !casereader_error (t->z_reader); free (t->z_scores); + casereader_destroy (t->z_reader); assert((t->missing_type != DSC_LISTWISE) ^ (t->vars != NULL)); free (t->vars); - return true; + free (t); + + return ok; } /* Sets up a transformation to calculate Z scores. */ @@ -658,7 +727,7 @@ setup_z_trns (struct dsc_proc *dsc, struct dataset *ds) t->z_score_cnt = cnt; t->missing_type = dsc->missing_type; t->exclude = dsc->exclude; - if ( t->missing_type == DSC_LISTWISE ) + if (t->missing_type == DSC_LISTWISE) { t->var_cnt = dsc->var_cnt; t->vars = xnmalloc (t->var_cnt, sizeof *t->vars); @@ -670,6 +739,11 @@ setup_z_trns (struct dsc_proc *dsc, struct dataset *ds) t->var_cnt = 0; t->vars = NULL; } + t->filter = dict_get_filter (dataset_dict (ds)); + t->z_reader = casewriter_make_reader (dsc->z_writer); + t->count = 0; + t->ok = true; + dsc->z_writer = NULL; for (cnt = i = 0; i < dsc->var_cnt; i++) { @@ -678,17 +752,17 @@ setup_z_trns (struct dsc_proc *dsc, struct dataset *ds) { struct dsc_z_score *z; struct variable *dst_var; + char *label; dst_var = dict_create_var_assert (dataset_dict (ds), dv->z_name, 0); - var_set_label (dst_var, - xasprintf (_("Z-score of %s"),var_to_string (dv->v)), - false); + + label = xasprintf (_("Z-score of %s"),var_to_string (dv->v)); + var_set_label (dst_var, label); + free (label); z = &t->z_scores[cnt++]; z->src_var = dv->v; z->z_var = dst_var; - z->mean = dv->stats[DSC_MEAN]; - z->std_dev = dv->stats[DSC_STDDEV]; } } @@ -706,8 +780,11 @@ static void calc_descriptives (struct dsc_proc *dsc, struct casereader *group, struct dataset *ds) { + const struct variable *filter = dict_get_filter (dataset_dict (ds)); struct casereader *pass1, *pass2; + casenumber count; struct ccase *c; + size_t z_idx; size_t i; c = casereader_peek (group, 0); @@ -739,10 +816,18 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, dsc->valid = 0.; /* First pass to handle most of the work. */ + count = 0; for (; (c = casereader_read (pass1)) != NULL; case_unref (c)) { double weight = dict_get_case_weight (dataset_dict (ds), c, NULL); + if (filter) + { + double f = case_num (c, filter); + if (f == 0.0 || var_is_num_missing (filter, f, MV_ANY)) + continue; + } + /* Check for missing values. */ if (listwise_missing (dsc, c)) { @@ -771,6 +856,8 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, if (x > dv->max) dv->max = x; } + + count++; } if (!casereader_destroy (pass1)) { @@ -785,6 +872,13 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, { double weight = dict_get_case_weight (dataset_dict (ds), c, NULL); + if (filter) + { + double f = case_num (c, filter); + if (f == 0.0 || var_is_num_missing (filter, f, MV_ANY)) + continue; + } + /* Check for missing values. */ if (dsc->missing_type == DSC_LISTWISE && listwise_missing (dsc, c)) continue; @@ -806,6 +900,15 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, } /* Calculate results. */ + if (dsc->z_writer && count > 0) + { + c = case_create (casewriter_get_proto (dsc->z_writer)); + z_idx = 0; + *case_num_rw_idx (c, z_idx++) = count; + } + else + c = NULL; + for (i = 0; i < dsc->var_cnt; i++) { struct dsc_var *dv = &dsc->vars[i]; @@ -839,8 +942,17 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, dv->stats[DSC_MAX] = dv->max == -DBL_MAX ? SYSMIS : dv->max; if (dsc->calc_stats & (1ul << DSC_SUM)) dv->stats[DSC_SUM] = W * dv->stats[DSC_MEAN]; + + if (dv->z_name && c != NULL) + { + *case_num_rw_idx (c, z_idx++) = dv->stats[DSC_MEAN]; + *case_num_rw_idx (c, z_idx++) = dv->stats[DSC_STDDEV]; + } } + if (c != NULL) + casewriter_write (dsc->z_writer, c); + /* Output results. */ display (dsc); } @@ -871,63 +983,53 @@ static algo_compare_func descriptives_compare_dsc_vars; static void display (struct dsc_proc *dsc) { - size_t i; - int nc; - struct tab_table *t; - - nc = 1 + (dsc->format == DSC_SERIAL ? 2 : 1); - for (i = 0; i < DSC_N_STATS; i++) + struct pivot_table *table = pivot_table_create ( + N_("Descriptive Statistics")); + pivot_table_set_weight_var (table, dict_get_weight (dsc->dict)); + + struct pivot_dimension *statistics = pivot_dimension_create ( + table, PIVOT_AXIS_COLUMN, N_("Statistics")); + pivot_category_create_leaf_rc ( + statistics->root, pivot_value_new_text (N_("N")), PIVOT_RC_COUNT); + for (int i = 0; i < DSC_N_STATS; i++) if (dsc->show_stats & (1ul << i)) - nc++; + pivot_category_create_leaf (statistics->root, + pivot_value_new_text (dsc_info[i].name)); if (dsc->sort_by_stat != DSC_NONE) sort (dsc->vars, dsc->var_cnt, sizeof *dsc->vars, descriptives_compare_dsc_vars, dsc); - t = tab_create (nc, dsc->var_cnt + 1); - tab_headers (t, 1, 0, 1, 0); - tab_box (t, TAL_1, TAL_1, -1, -1, 0, 0, nc - 1, dsc->var_cnt); - tab_box (t, -1, -1, -1, TAL_1, 1, 0, nc - 1, dsc->var_cnt); - tab_hline (t, TAL_2, 0, nc - 1, 1); - tab_vline (t, TAL_2, 1, 0, dsc->var_cnt); - - nc = 0; - tab_text (t, nc++, 0, TAB_LEFT | TAT_TITLE, _("Variable")); - if (dsc->format == DSC_SERIAL) + struct pivot_dimension *variables = pivot_dimension_create ( + table, PIVOT_AXIS_ROW, N_("Variable")); + for (size_t i = 0; i < dsc->var_cnt; i++) { - tab_text (t, nc++, 0, TAB_CENTER | TAT_TITLE, _("Valid N")); - tab_text (t, nc++, 0, TAB_CENTER | TAT_TITLE, _("Missing N")); - } - else - tab_text (t, nc++, 0, TAB_CENTER | TAT_TITLE, "N"); + const struct dsc_var *dv = &dsc->vars[i]; - for (i = 0; i < DSC_N_STATS; i++) - if (dsc->show_stats & (1ul << i)) - { - const char *title = gettext (dsc_info[i].name); - tab_text (t, nc++, 0, TAB_CENTER | TAT_TITLE, title); - } - - for (i = 0; i < dsc->var_cnt; i++) - { - struct dsc_var *dv = &dsc->vars[i]; - size_t j; + int row = pivot_category_create_leaf (variables->root, + pivot_value_new_variable (dv->v)); - nc = 0; - tab_text (t, nc++, i + 1, TAB_LEFT, var_to_string (dv->v)); - tab_text_format (t, nc++, i + 1, 0, "%g", dv->valid); - if (dsc->format == DSC_SERIAL) - tab_text_format (t, nc++, i + 1, 0, "%g", dv->missing); + int column = 0; + pivot_table_put2 (table, column++, row, + pivot_value_new_number (dv->valid)); - for (j = 0; j < DSC_N_STATS; j++) + for (int j = 0; j < DSC_N_STATS; j++) if (dsc->show_stats & (1ul << j)) - tab_double (t, nc++, i + 1, TAB_NONE, dv->stats[j], NULL); + { + union value v = { .f = dv->stats[j] }; + struct pivot_value *pv = (j == DSC_MIN || j == DSC_MAX + ? pivot_value_new_var_value (dv->v, &v) + : pivot_value_new_number (dv->stats[j])); + pivot_table_put2 (table, column++, row, pv); + } } - tab_title (t, _("Valid cases = %g; cases with missing value(s) = %g."), - dsc->valid, dsc->missing_listwise); - - tab_submit (t); + int row = pivot_category_create_leaves ( + variables->root, N_("Valid N (listwise)"), N_("Missing N (listwise)")); + pivot_table_put2 (table, 0, row, pivot_value_new_number (dsc->valid)); + pivot_table_put2 (table, 0, row + 1, + pivot_value_new_number (dsc->missing_listwise)); + pivot_table_submit (table); } /* Compares `struct dsc_var's A and B according to the ordering @@ -942,7 +1044,7 @@ descriptives_compare_dsc_vars (const void *a_, const void *b_, const void *dsc_) int result; if (dsc->sort_by_stat == DSC_NAME) - result = strcasecmp (var_get_name (a->v), var_get_name (b->v)); + result = utf8_strcasecmp (var_get_name (a->v), var_get_name (b->v)); else { double as = a->stats[dsc->sort_by_stat];