X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Fdescriptives.c;h=1ffe861552b6aaae990816bb08097a2e4db7f617;hb=df2f1ae30c2835a39f4bc6b0c14f10a2090d9401;hp=adedd5e67f600aeb651eafa68316d615d5d73ef0;hpb=9ade26c8349b4434008c46cf09bc7473ec743972;p=pspp diff --git a/src/language/stats/descriptives.c b/src/language/stats/descriptives.c index adedd5e67f..1ffe861552 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 @@ -16,14 +16,16 @@ #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/procedure.h" #include "data/transformations.h" #include "data/variable.h" #include "language/command.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. */ + 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. */ @@ -160,6 +166,9 @@ 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. */ @@ -213,6 +222,7 @@ cmd_descriptives (struct lexer *lexer, struct dataset *ds) 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) @@ -367,6 +377,8 @@ 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 +398,20 @@ 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 +443,7 @@ 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 +501,7 @@ 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 +521,7 @@ 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; @@ -578,7 +605,7 @@ dump_z_table (struct dsc_proc *dsc) 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_get_name (dsc->vars[i].v)); + 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); } } @@ -586,6 +613,15 @@ dump_z_table (struct dsc_proc *dsc) 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_data_rw (c, z->z_var)->f = SYSMIS; +} + /* Transformation function to calculate Z-scores. Will return SYSMIS if any of the following are true: 1) mean or standard deviation is SYSMIS 2) score is SYSMIS 3) score is user missing and they were not included in the original @@ -599,7 +635,48 @@ 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")); + t->ok = false; + } + descriptives_set_all_sysmis_zscores (t, *c); + return TRNS_CONTINUE; + } + } + t->count--; if (t->missing_type == DSC_LISTWISE) { @@ -609,19 +686,18 @@ 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; - 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 +711,15 @@ 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. */ @@ -670,6 +750,11 @@ 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 +763,17 @@ 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)), - dict_get_encoding (dataset_dict (ds)), 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 +791,11 @@ static void calc_descriptives (struct dsc_proc *dsc, struct casereader *group, struct dataset *ds) { + 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 +827,18 @@ 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 +867,8 @@ calc_descriptives (struct dsc_proc *dsc, struct casereader *group, if (x > dv->max) dv->max = x; } + + count++; } if (!casereader_destroy (pass1)) { @@ -785,6 +883,13 @@ 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 +911,15 @@ 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_data_rw_idx (c, z_idx++)->f = count; + } + else + c = NULL; + for (i = 0; i < dsc->var_cnt; i++) { struct dsc_var *dv = &dsc->vars[i]; @@ -839,8 +953,17 @@ 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_data_rw_idx (c, z_idx++)->f = dv->stats[DSC_MEAN]; + case_data_rw_idx (c, z_idx++)->f = dv->stats[DSC_STDDEV]; + } } + if (c != NULL) + casewriter_write (dsc->z_writer, c); + /* Output results. */ display (dsc); } @@ -914,18 +1037,19 @@ display (struct dsc_proc *dsc) size_t j; nc = 0; - tab_text (t, nc++, i + 1, TAB_LEFT, var_get_name (dv->v)); - tab_text_format (t, nc++, i + 1, 0, "%g", dv->valid); + tab_text (t, nc++, i + 1, TAB_LEFT, var_to_string (dv->v)); + tab_text_format (t, nc++, i + 1, 0, "%.*g", DBL_DIG + 1, dv->valid); if (dsc->format == DSC_SERIAL) - tab_text_format (t, nc++, i + 1, 0, "%g", dv->missing); + tab_text_format (t, nc++, i + 1, 0, "%.*g", DBL_DIG + 1, dv->missing); for (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); + tab_double (t, nc++, i + 1, TAB_NONE, dv->stats[j], NULL, RC_OTHER); } - tab_title (t, _("Valid cases = %g; cases with missing value(s) = %g."), - dsc->valid, dsc->missing_listwise); + tab_title (t, _("Valid cases = %.*g; cases with missing value(s) = %.*g."), + DBL_DIG + 1, dsc->valid, + DBL_DIG + 1, dsc->missing_listwise); tab_submit (t); } @@ -942,7 +1066,7 @@ 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];