1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 2017 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "matrix-reader.h"
23 #include <libpspp/message.h>
24 #include <data/casegrouper.h>
25 #include <data/casereader.h>
26 #include <data/dictionary.h>
27 #include <data/variable.h>
30 #define _(msgid) gettext (msgid)
31 #define N_(msgid) msgid
35 This module interprets a "data matrix", typically generated by the command
36 MATRIX DATA. The dictionary of such a matrix takes the form:
38 s_0, s_1, ... s_m, ROWTYPE_, VARNAME_, v_0, v_1, .... v_n
40 where s_0, s_1 ... s_m are the variables defining the splits, and
41 v_0, v_1 ... v_n are the continuous variables.
45 The variables ROWTYPE_ and VARNAME_ are of type A8,
46 the variables s_x are of type F4.0 (although this reader accepts any type),
47 and v_x are of any numeric type.
49 The values of the ROWTYPE_ variable are in the set {MEAN, STDDEV, N, CORR, COV}
50 and determine the purpose of that case.
51 The values of the VARNAME_ variable must correspond to the names of the varibles
52 in {v_0, v_1 ... v_n} and indicate the rows of the correlation or covariance
57 A typical example is as follows:
59 s_0 ROWTYPE_ VARNAME_ v_0 v_1 v_2
61 0 MEAN 5.0000 4.0000 3.0000
62 0 STDDEV 1.0000 2.0000 3.0000
63 0 N 9.0000 9.0000 9.0000
64 0 CORR V1 1.0000 .6000 .7000
65 0 CORR V2 .6000 1.0000 .8000
66 0 CORR V3 .7000 .8000 1.0000
67 1 MEAN 9.0000 8.0000 7.0000
68 1 STDDEV 5.0000 6.0000 7.0000
69 1 N 9.0000 9.0000 9.0000
70 1 CORR V1 1.0000 .4000 .3000
71 1 CORR V2 .4000 1.0000 .2000
72 1 CORR V3 .3000 .2000 1.0000
78 const struct dictionary *dict;
79 const struct variable *varname;
80 const struct variable *rowtype;
81 struct casegrouper *grouper;
83 gsl_matrix *n_vectors;
84 gsl_matrix *mean_vectors;
85 gsl_matrix *var_vectors;
87 gsl_matrix *correlation;
88 gsl_matrix *covariance;
91 struct matrix_reader *
92 create_matrix_reader_from_case_reader (const struct dictionary *dict, struct casereader *in_reader,
93 const struct variable ***vars, size_t *n_vars)
95 struct matrix_reader *mr = xzalloc (sizeof *mr);
98 mr->varname = dict_lookup_var (dict, "varname_");
99 if (mr->varname == NULL)
101 msg (ME, _("Matrix dataset lacks a variable called %s."), "VARNAME_");
106 mr->rowtype = dict_lookup_var (dict, "rowtype_");
107 if (mr->rowtype == NULL)
109 msg (ME, _("Matrix dataset lacks a variable called %s."), "ROWTYPE_");
115 const struct variable **dvars = NULL;
116 dict_get_vars (dict, &dvars, &dvarcnt, DC_SCRATCH);
119 *n_vars = dvarcnt - var_get_dict_index (mr->varname) - 1;
124 *vars = xcalloc (sizeof (struct variable **), *n_vars);
126 for (i = 0; i < *n_vars; ++i)
128 (*vars)[i] = dvars[i + var_get_dict_index (mr->varname) + 1];
132 /* All the variables before ROWTYPE_ (if any) are split variables */
133 mr->grouper = casegrouper_create_vars (in_reader, dvars, var_get_dict_index (mr->rowtype));
141 destroy_matrix_reader (struct matrix_reader *mr)
145 bool ret = casegrouper_destroy (mr->grouper);
152 next_matrix_from_reader (struct matrix_material *mm,
153 struct matrix_reader *mr,
154 const struct variable **vars, int n_vars)
156 struct casereader *group;
158 gsl_matrix_free (mr->n_vectors);
159 gsl_matrix_free (mr->mean_vectors);
160 gsl_matrix_free (mr->var_vectors);
161 gsl_matrix_free (mr->correlation);
162 gsl_matrix_free (mr->covariance);
164 if (!casegrouper_get_next_group (mr->grouper, &group))
167 mr->n_vectors = gsl_matrix_alloc (n_vars, n_vars);
168 mr->mean_vectors = gsl_matrix_alloc (n_vars, n_vars);
169 mr->var_vectors = gsl_matrix_alloc (n_vars, n_vars);
171 mm->n = mr->n_vectors;
172 mm->mean_matrix = mr->mean_vectors;
173 mm->var_matrix = mr->var_vectors;
175 mr->correlation = NULL;
176 mr->covariance = NULL;
180 for ( ; (c = casereader_read (group) ); case_unref (c))
182 const union value *uv = case_data (c, mr->rowtype);
184 for (col = 0; col < n_vars; ++col)
186 const struct variable *cv
187 = vars ? vars[col] : dict_get_var (mr->dict, var_get_dict_index (mr->varname) + 1 + col);
188 double x = case_data (c, cv)->f;
189 if (0 == strncasecmp ((char *)value_str (uv, 8), "N ", 8))
190 for (row = 0; row < n_vars; ++row)
191 gsl_matrix_set (mr->n_vectors, row, col, x);
192 else if (0 == strncasecmp ((char *) value_str (uv, 8), "MEAN ", 8))
193 for (row = 0; row < n_vars; ++row)
194 gsl_matrix_set (mr->mean_vectors, row, col, x);
195 else if (0 == strncasecmp ((char *) value_str (uv, 8), "STDDEV ", 8))
196 for (row = 0; row < n_vars; ++row)
197 gsl_matrix_set (mr->var_vectors, row, col, x * x);
199 if (0 == strncasecmp ((char *) value_str (uv, 8), "CORR ", 8))
201 if (mr->correlation == NULL)
202 mr->correlation = gsl_matrix_alloc (n_vars, n_vars);
203 for (col = 0; col < n_vars; ++col)
205 const struct variable *cv
206 = vars ? vars[col] : dict_get_var (mr->dict, var_get_dict_index (mr->varname) + 1 + col);
207 double x = case_data (c, cv)->f;
208 gsl_matrix_set (mr->correlation, crow, col, x);
212 else if (0 == strncasecmp ((char *) value_str (uv, 8), "COV ", 8))
214 if (mr->covariance == NULL)
215 mr->covariance = gsl_matrix_alloc (n_vars, n_vars);
216 for (col = 0; col < n_vars; ++col)
218 const struct variable *cv
219 = vars ? vars[col] : dict_get_var (mr->dict, var_get_dict_index (mr->varname) + 1 + col);
220 double x = case_data (c, cv)->f;
221 gsl_matrix_set (mr->covariance, crow, col, x);
227 casereader_destroy (group);
229 mm->cov = mr->covariance;
230 mm->corr = mr->correlation;